Advanced Computing in the Age of AI | Thursday, August 18, 2022

Oracle Cranks Up The Cores To 32 With Sparc M7 Chip 

Say what you will about Oracle co-founder and CEO Larry Ellison, but when the software giant bought Sun Microsystems more than four years ago, for $7.4 billion, he said that he had seen the light and wanted Oracle to control its own hardware and created specialized machinery tuned up to run Oracle's software stack. Despite falling sales in the early years, which have now stabilized and are now growing, Oracle has continued to invest in hardware.

This week at the Hot Chips conference, its chip designers are showing off the forthcoming Sparc M7, the biggest and baddest Sparc processor that either Sun Microsystems or Oracle has ever created. Stephen Phillips, senior director of Sparc architecture, gave the presentation at Hot Chips on the Sparc M7, and John Fowler, executive vice president of systems, spoke to EnterpriseTech about the feeds and speeds at a system level and what it will mean for customers when it ships sometime next year.

With over 10 billion transistors on the die, the Sparc M7 is a whopper and will be, in terms of transistor count, the most dense processor on the market – bar none – when it ships sometime in 2015. The chip will have 32 cores, which is larger than a lot of four-socket servers had only a few years ago and which by any measure would have constituted a supercomputer two decades ago. It will be etched using Taiwan Semiconductor Manufacturing Corp's 16 nanometer FinFET 3D transistor manufacturing node (which is sometimes also called 20 nanometer by some customers) on a 13 metal layer design.

The M7 will be the sixth processor that Oracle has brought to market since the Sun acquisition closed in January 2010, and it is based on the fourth generation of Sparc CMT cores (short for Chip Multi-Threading) designed by Sun and Oracle. The Sparc T3 chip for entry and midrange servers from four years ago was based on the S2 cores, the Sparc T4 and T5 chips for similar sized boxes as well as the high-end Sparc M5 servers had chips based on the S3 cores. The Sparc M6-32 system announced last year and nicknamed the Big Memory Machine was based on the Sparc M6 chip, which also used the S3 cores. Like Intel, Oracle stretches a core design over several processor generations, making a few architectural tweaks between products and changing the performance profile of each chip by changing core counts, clock speeds, cache sizes, or altering other features like system interconnects.

The S4 core has a dual-issue, out-of-order execution unit and it has dynamic threading that ranges from one to eight virtual threads per core. The Sparc chips have had eight threads for a long time, but only in recent years with the S3 cores were the threads able to be dynamically allocated. What this means is that if a workload needs very high single-thread performance, the processor can allocate the chip resources to a single thread on a core, thereby allowing its work to complete more quickly than if it had to share resources across many threads on a core. (Dynamic threading is also available on IBM's Power7, Power7+, and Power8 chips and the latter has eight virtual threads per core just like the S3 and S4 cores from Oracle. Intel's Xeon processor have HyperThreading, which is also dynamic and which provides two virtual threads per core.) The S4 core has two Arithmetic Logic Units, one Load Store Unit, one Floating Point & Graphics Unit, one Branch Unit, and one Stream Processing Unit.

The S4 cores each have 16 KB of L1 instruction cache and 16 KB of L1 data cache. They are organized into clusters of four. The S4 core has a new L2 cache architecture, which allows for the same cycle count for accessing caches but allows for that cache to be 50 percent larger and still get the accesses done in the same compute time. The four cores share a 256 KB L2 instruction cache, which sits at the heart of the four-core cluster, with four independent interfaces to each core that deliver in excess of 128 GB/sec of bandwidth. (Oracle is not saying how much because it would reveal the clock speed of the processor, but Fowler did confirm that the Sparc M7 will spin faster than the current 3.6 GHz of the Sparc T5, M5, and M6 chips.) Each pair of cores shares a 256 KB L2 writeback data cache that also has interfaces to the core that provide more than 128 GB/sec of bandwidth. By having two L2 data caches, the bandwidth into the S4 core is twice as high as was the case with the S3 core. The overall L2 cache bandwidth on the four-core module is more than 1 TB/sec, which makes L2 cache bandwidth more than 8 TB/sec for the entire 32-core processor.

Stepping up the next level in the memory hierarchy is the L3 cache. Each cluster of four S4 cores on the Sparc M7 chip can push data into the L3 cache at 70 GB/sec and pull it out at 140 GB/sec. On the other side of the cache sits the on-chip network (OCN), which has two ports running at 64 GB/sec that move data into and out of L3 cache and onto the OCN to be sent around the interconnect to other caches as S4 cores in the chip request data that is not in their own local L3 cache segment. That L3 cache weighs in at 64 MB across all eight segments, which is 33 percent higher than the cache on the Sparc M5 and M6 chips. The local latency in this L3 cache has been reduced by 25 percent compared to the Sparc M6 processor, says Oracle, and at 1.6 TB/sec of aggregate L3 cache bandwidth across the Sparc M7, that is 2.5X that of the Sparc T5 and 5X that of the Sparc M6. One neat new feature: accelerators on the die and outside of the core can reach in over 256 GB/sec links on the OCN to allocate data into one of the eight L3 cache segments on the die. That OCN has 512 GB/sec of bisection data bandwidth and uses rings to request data, point-to-point links to get responses and a mesh to move data. The four 256 GB/sec ports (two up, two down) also go out to four DDR4 main memory controllers.

With such a beast of a chip, you might be thinking that Oracle would add a L4 cache layer somewhere in the system, and for all we know, it will. What we do know is that the Sparc M7 will have sixteen memory channels across its four DDR4 controllers and will support memory running at 2.1 GHz, 2.4 GHz, and 2.67 GHz. Using 2.1 GHz memory, a single socket will have 160 GB/sec of memory bandwidth, which is 2X that of the Sparc T5 and 3X that of the Sparc M6. Each DDR4 interface has two channels, and the memory links coming off them out to buffer chips on the DDR4 memory cards runs at between 12.8 Gb/sec and 16 Gb/sec, depending on the memory speed. The memory controller allows for DIMM retirement without having to stop the system in the event that a chip fails in the field, and memory lanes can failover with full CRC protection. The memory will top out at 2 TB per socket, twice that of the Sparc M6 system.

The Sparc M7 chip has four PCI-Express 3.0 links that support more than 75 GB/sec of bandwidth, which is twice as much as the Sparc T5 and M6 chips.

For improved power management, the Sparc M7 chip has an on-die power estimator on each core that takes a whack at guessing what the power draw is every 250 nanoseconds, and this feeds data into a power controller that tries to figure out what the power draw is for the entire chip. This is accurate to within a few percent of measured power for the chip, Phillips said in his presentation, and is used to adjust the frequency and voltage of each quadrant of the chip as workloads rise and fall. Each quadrant can be power gated separately.

Both the S4 core and the circuitry wrapped around it have various kinds of accelerators and other functions added to them specifically to enhance the performance of the Solaris Unix operating system, Java middleware and applications, or Oracle's relational databases.

On the Java front, the Sparc M7 has new memory protection features and virtual address masking that will make Java garbage collection easier and more deterministic, according to Fowler. More generically, the Sparc M7 has features to try to get threads that are sharing data to get onto the same S4 core cluster on the die and similarly has features to try to keep LDom logical partitions from thrashing the L3 cache.

The S4 core, for instance, has special instructions to ensure application data integrity, which is done in real-time and which safeguards against invalid or stale memory references and buffer overruns for both Solaris running C and C++ applications and the Oracle database. The Sparc M7 also has database query offload engines and accelerators for in-memory compression and decompression algorithms.

"Decompression can be driven at memory bandwidth rates, so there is no downside to using compression," says Fowler. The on-chip compression leaves the S4 cores leftover capacity to do useful work.


The query accelerator for the Oracle 12c database's in-memory columnar data store does in-memory format conversions, value and range conversions, and set membership lookups. These on-chip database functions were developed in conjunction with the Oracle database team and reside on eight off-core query accelerator engines.


In one example, Phillips showed a query where a SQL command running against an Oracle database was told to find the records in a database of cars where the make was a Toyota and the year of the car was greater than or equal to 2010. The query accelerator fuses both the decompression of the database data and filtering of that data, and the thread where that query is running is stopped precisely long enough by the thread manager in the chip to allow the results to pop up precisely when the thread resumes and needs t5he results of the offloaded query. This is all done in hardware and at memory bandwidth speed, and a Sparc M7 system was able to run that query about 4.7X faster than a Sparc T5 baseline system.

Here is where Oracle thinks the Sparc M7 processor will stack up compared to the processor Sparc M6 over a bunch of different workloads:


All things being equal, you would expect around a 2.7X bump just moving from 12 cores in the Sparc M6 to 32 cores in the Sparc M7. The microarchitecture and clock speed increase probably make up for most of the rest, since the increased caches and bandwidths are used to keep these cores fed. If half the remaining performance increase comes from architecture, then the Sparc M7 will probably clock in at around 4 GHz, given this performance data above, which averages around a 3.25X bump.

It's All About The System

Way back when, four and a half years ago when Oracle put out its first Sparc roadmap, it said that it would be able to deliver monster Sparc machines in the 2015 timeframe and it showed the steps it was going to take over the years to get there. To its credit, Oracle has by and large stuck to this roadmap, although it has scaled back on the NUMA scalability of its systems even as it did create the "Bixby" interconnect, which during the Sparc M5 and M6 generations could have scaled up to 96 sockets in a single system image with 96 TB of main memory.

In August 2010, Oracle put these stakes in the ground to demonstrate its good faith to those who did not believe Oracle believed in hardware:


This roadmap showed both the Sparc T and M series of processors and their overall throughput and single-threaded performance gains compared to their prior generations of T and M chips. Oracle was promising four times the cores, 32 times the threads, 16 times the memory capacity, and 40 times the database transaction processing from its base T and M machines to the final machine on the roadmap, which had neither a T nor an M on it. The M Series machines were supposed to scale up to 64 sockets, but the commercial version never did reach higher than 32 sockets even though, as we pointed out above, the Bixby interconnect can scale three times as far as what the Sparc M6-32 system can do today.

The late 2014 or early 2015 product, you will notice, never did have a T or an M next to it, just Sparc, and everyone was a little perplexed by this. But, as it turns out, Oracle will be discontinuing the T Series chips, Fowler tells EnterpriseTech, and building future Sparc machines on the M7 processors solely. So it looks like the Sparc T6, which was originally slated around the end of this year into early next year is being cut. The good news is that Oracle delivered the Sparc M6 early and the M7 is right on time and able to handle all kinds of work. It also seems to suggest that Oracle wants a better chip with which to compete against Xeon E5 and E7 processors from Intel and Power processors from IBM, and a low-core count T Series chip was not cutting it. It is also cheaper to make one chip than two – particularly if, like Oracle, a whole line of clustered systems is available using Intel Xeon E5 and E7 processors. Oracle needs to pick its fights, and the Sparc M7 is being aimed at heavy, big memory workloads on scale-up systems.

And as EnterpriseTech has pointed out before, Oracle has made some pretty compelling arguments about why such scale-up systems are better than scale-out clusters. Network latencies between nodes using the first-generation Bixby interconnect for the Sparc M6 are on the order of 150 nanoseconds, 100X lower than for a 10 Gb/sec Ethernet link between nodes in a loosely coupled cluster of Sparc T2 servers with approximately the same performance as the M6-32 system. The M6-32 has 24 TB/sec of bandwidth, about 37X that of the cluster with 16 servers with four 10 Gb/sec ports each. Both setups cost a little more than $1.2 million. For workloads that are bandwidth and latency sensitive, the NUMA machine is better than the cluster.


The Sparc M7 has on-chip NUMA interconnect circuits that will allow eight sockets to be snapped together into a single system image without extra glue chips, just as was the case with the Sparc M5 and M6 chips. An eight-socket machine will have a maximum of 256 cores, 2,048 threads, and 16 TB of main memory, with a bisection bandwidth across the interconnect of more than 1 TB/sec.

Fowler confirmed that the Sparc M7 will use a second-generation Bixby interconnect, and that initially it will scale to 32 sockets. This interconnect will have two links connecting every processor socket to every Bixby interconnect switch; there are a total of twelve switches to link four-socket NUMA nodes together, which is six for one set of links and six for the other. The bi-section bandwidth on this switch interconnect is 5.2 TB/sec, which is four times that of the first-generation Bixby interconnect used in the M6-32 system. This M7-32 machine, as we presume it will be called, will have 1,024 cores, 8,192 threads, and up to 64 TB of main memory.

Interestingly, the updated Bixby NUMA interconnect will allow for cache coherent links between the Sparc four-socket nodes and will also allow for non-coherent links. This will allow a big Sparc M7 machine to function as a cluster for Oracle RAC database clustering software, using the Bixby interconnect instead of 40 Gb/sec InfiniBand as the Exadata database clusters do. The Bixby interconnect and SMP glue chips also have low-level message passing protocols etched on them specifically to pass messages used in database clusters.

When asked about what the performance advantage would be comparing an InfiniBand or Ethernet cluster running Oracle RAC and the Sparc M7-Bixby setup using the non-coherent memory clusters, Fowler said that the difference "would not be subtle."

The Sparc M7 chips are in the late stages of testing now and will come out in systems in 2015. Solaris 11 will be supported on the machines, with Solaris 10 being supported inside of LDoms and Solaris 8, 9, and 10 being supported inside of Solaris containers for customers who have vintage workloads.

34 Responses to Oracle Cranks Up The Cores To 32 With Sparc M7 Chip

  1. Phil Dunn says:

    Great in-depth article! Well done. Would be great if you could add some un-biased competitive comparisons on SPARC M7 to Power8 and Xeon (expected in early 2015 of course)…

    FYI. theres a typo that seems to contradict the advantage of the accelerator:

    “Decompression can be driven at memory bandwidth rates, so there is now downside to using compression,” says Fowler.

    “now” should be “no”


  2. Very interesting article, thanks for the write-up Timothy.

    A few comments: why do you think an L4 cache would carry any benefit? Where would it go?

    Also regarding the “memory protection features and virtual address masking that will make Java garbage collection easier and more deterministic”, what are these features precisely and how will they assist GC? Perhaps that needs an article all of its own.

    Do you know when we’re likely learn the clock speed?

    • Mark Funk says:

      Mark –

      Since this article was referring to the concept of an L4 cache within a paragraph describing the chip’s memory controller, I suspect that Timothy was alluding to the possibility that an L4 cache could be included outside of the memory chip much as is being done on IBM’s POWER8. Check out and search down for Centaur. Aside from faster access to data still in the L4, this tends to speed how quickly data being cast-out of a processor’s cache makes its way back to memory.


  3. Ted Codd says:

    The roadmap shows M7 going into 1-way systems. It will be interesting to see if Oracle makes high volume 1 and 2 way boxes. With the right software bundle they could sell quite a lot of them.

  4. Brett Murphy says:

    It is a good article but it does lack analysis of Oracles continued claims, what they deliver and the real impact to customers. They claimed 64 sockets but M5 & M6 have just 32 sockets. These servers are 192 & 384 cores respectively vs 1024 cores in the M7. With Oracle you have to listen to everything they say as they are a marketing company first that seem like masters of communication (I wanted to say deception but that sounds so harsh). The article claims a 2.7X performance bump but going from 384 cores to 1026 is 2.66666X increase. A cynic may say that is 2.7X. Thus, the actual increase is just .1X at best.

    In 2010 was T3 with 4 sockets x 16 cores with 8 threads per core totaling 512 threads. The article says M7 will be 32 sockets with 32 cores per socket with 8 threads totaling 8192 threads – that is not 32X. Another Oracle overstatement!

    Cores increase 2.66X while memory increases just 2X from M6 to M7 or 32 TB to 64 TB. So, for so many cores they actually decrease the amount of memory per core. I read it would have 4 memory controllers per socket with 160 GB/s of bandwidth. Compare this to the M6-32 which has 12 cores per socket with 2 memory controllers with a per socket bandwidth of 60 GB/s. Cores go up 2.7X (I’m rounding up 2.666666) from 12 to 32, memory only increases 2X while per socket memory bandwidth increases 2.6666 or 2.7X. Where is the improvement (and this is with DDR4 memory at that)?

    There seems to be no increase in per core performance or other benefits. Don’t dispute the *potential* benefits from some of their accelerators but I’ll wait to analyze that when and if they come to market with M7 based solutions. Could go on with other comments but customers should realize at the end of the day they need servers as part of solutions that meet business requirements at the lowest cost…that cost isn’t aways the server but everything from OS, virtualization, maintenance, Oracle database, Oracle annual maintenance cost, etc.

    • kebabbert says:

      If you compare the largest M7 server with 32 sockets, 1024 cores, 8192 threads, 64TB RAM, to the largest IBM POWER8 server, there is no competition.

      There are rumours that the POWER8 has scalability problems and the maximum will be 16-socket POWER8 servers:
      “…Depending on how many customers are hitting the performance ceiling on the Power 795, IBM could skip putting out a 32-socket Power8 machine and just got with the 16-socket machine with 16 TB of memory….”

      Because the POWER8 is only ~2-2.5x faster than the POWER7 we see that this 16 socket POWER8 server matches the largest IBM P795 server with 32-socket POWER7 cpus. So you choose, an 32-socket P795 or an 16-socket POWER8 server – they will have roughly the same performance.

      If we talk about the SPARC M6, it is faster than the POWER7 cpu. But this SPARC M7 is 3-3.5x faster than SPARC M6. So a small ordinary 8-socket M7 server corresponds to an 24-28 socket M6 server. And such a big M6 server corresponds to a 32-socket P795 or 16-socket POWER8 server.

      Ergo, a small 8-socket M7 SPARC server, corresponds to a 16-socket POWER8 server, or a 32-socket P795 server. Just look at the specs:
      -Oracle 8-socket M7 server: 256 cores, 2048 threads, 16TB RAM
      -IBM P795 server: 256 cores, 1024 threads, 16 TB RAM
      -IBM 16-socket POWER8 server: 192 cores, 1536 threads, 16TB RAM

      We see that the Oracle server is best in class. So what do you choose? A small cheap 8-socket M7 server, or a hugely expensive and large and power hungry 32-socket P795 server?

      BTW, I dont understand why IBM did release the POWER8 cpu at all, as it doesnt even beat existing older cpus such as the SPARC T5. POWER8 can not even beat the newest Intel x86 cpus.

      • Brett Murphy says:

        You made a lot of assumptions with nothing to back up any of your statements. Making the claim does not substantiate the claim.

        Oracle achieves world records and impressive numbers by one means and one means ONLY. They produce servers with excessive sockets and cores then aggregate the values for the given product and claim superiority. That is weak engineering, disingenuous marketing with diluted value to the customer.

        First proofpoint – Oracle sells software, primarily by core (yes, other means are available). Thus, it stands to reason that performance per core is crucial. Look at the SAP benchmark where there are Power7 795, Power8 S824 and the M6-32 results.

        M6-32 with 32 sockets & 384 cores deliver 793,930 SAPS or 2,067 SAPS per core. P7 795 with 32 sockets and 256 cores deliver 688,630 SAPS or 2,690 SAPS per core. With these two, which delivers the greatest performance per core? The entry level Power8 S824 server with 2 sockets and 24 cores deliver 115,870 SAPS or 4,827 SAPS per core. For additional comparison purposes, here is the 8 socket 124 core T5-8 that deliver 220,950 SAPS or 1,726 SAPS per core.

        Just about every SAP landscape that I see have a total SAPS requirement but but that isn’t a singular value. Also, the ability of the server to utilize the full SAP value of that platform is also critical – this is dependent on the ability of the hypervisor, server technology and OS to be able to drive utilization that is useful. What this means is the likelihood of a Power8 24 core server being able to deliver 115K SAPS is quite likely. The likelihood of a x86 server being able to deliver it’s full SAPS is not likely. Given the weak hypervisor used in M6 & T5 along with the less capable SPARC chipset which includes the less efficient CMT (threads) makes it less likely for these servers to be driven as high as Power. I’ll hold off on critiquing Domains and LDOMs pending your desire to engage in that discussion as I look forward to that.

        Looking at the per core results of M6-32 I deliver more per core with the 4 year old P7 795 server and IBM’s entry level 2 socket 24 core S824 delivers 2.8X higher performance per core over the “Solves World Hunger” Oracle SPARC M5-8 and 2.3X over the M6-32 “King of the Hill” server. To put this in perspective, if we normalize the performance of the S824 with its 24 cores, that is equivalent to approximately 67 T5-8 SPARC cores or 56 x M6-32 SPARC cores.

        If you are a SPARC customer then you have fallen for their smoke and mirrors marketing that performance increase comes by doubling the cores. That simply doubles your software costs. If you work for Oracle then I expect you to sing from the Larry song book 🙂 What is impressive with IBM’s Power8 technology when you see that it is 2X the performance over Power7 and x86 is that they are doing that on a per core basis. Using the Oracle sizing method we claim we have 3X the performance over Power7 and x86 but that is because we are going from 8 cores in Power7 to 12 cores in Power8. That means nothing unless you are pay software licenses by the socket then per core and # of cores per socket is very important.

        Hopefully you see the error of your ways. I am a IT business partner, Architect and evangelist on Power technologies. I also worked at Sun for 10 years where I was a cluster and storage specialist (also StorageACE) and a instructor in the Army on SunOS/Solaris servers. I say this to say that i am a fan of Solaris. I love and miss Sun – great people and culture. Plus, I have experience on the platform. I am not speaking here as a seller trying to persuade you or the reader that I’m right because of talking points. I continue to challenge Oracle to a live, face to face technical debate that we can record and publish for all to see and hear the discussions. We both can have a whiteboard to aid in our discussions. We should have a panel of non-biased industry experts who can award points based on winning each topic. Cause we know there are some people who will never accept defeat even if proven beyond the shadow of a doubt.

        • Kebabbert says:

          As you kindly have answered to this post here:

          Where you answered me like this:
          “Spewing more talking points and FUD. You aren’t a “fanboi” as that implies loyalty to technology that may or may not be the best. You are dazed and confused…actually I would say you work for Oracle and simply defending the Larry’s honor or you work in their marketing department. Maybe you are the new guy and this is part of your hazing – “Get our there and get your butt kicked while saying this and this and don’t forget this!”. You can’t be taken seriously when everything you say is an obvious attempt at disparaging competitive platforms with no substantial data. “Rumours say …” – whatever!”

          To this I answer:
          “…Oracle achieves world records and impressive numbers by one means and one means ONLY. They produce servers with excessive sockets and cores then aggregate the values for the given product and claim superiority. That is weak engineering, disingenuous marketing with diluted value to the customer. M6-32 with 32 sockets & 384 cores deliver 793,930 SAPS or 2,067 SAPS per core. P7 795 with 32 sockets and 256 cores deliver 688,630 SAPS or 2,690 SAPS per core. With these two, which delivers the greatest performance per core?…”

          Who cares about cores? We are discussing which cpu or server is the fastest. And who has the higher score? Which server is more powerful? Which cpu is fastest? Sure, if we are going to discuss “who has the fastest core” – but we are not. We are discussing which cpu is best. Which server is most powerful. And Oracle SPARC wins hand down and humiliates the POWER6, POWER7 and POWER8.


          “….Spewing more talking points and FUD. You aren’t a “fanboi” as that implies loyalty to technology that may or may not be the best. You are dazed and confused…actually I would say you work for Oracle and simply defending the Larry’s honor or you work in their marketing department. Maybe you are the new guy and this is part of your hazing – “Get our there and get your butt kicked while saying this and this and don’t forget this!”….”

          Again, I have never worked at Oracle nor Sun. I have always worked in Finance, and right now I am a researcher in algorithmic trading. I tried to get my huge company to buy more Solaris and SPARC, but failed. But you on the other hand, work at IBM, right? Are you getting paid to write this errorneous stuff?


          “…You can’t be taken seriously when everything you say is an obvious attempt at disparaging competitive platforms with no substantial data. “Rumours say …” – whatever!…”

          I always post links as you can see, I would never post claims without links as you do. Just read the link I posted regarded “rumours say”. A mathematician needs to prove his claims, and I do that with links. Hardly anything I say are my own conclusions, I just reiterate and post official benchmarks and other links that other people have written. You on the other hand, makes up lot of weird stuff without any links. Your talk about “one core is faster, therefore the cpu is faster” – how in earth can you arrive at such a wrong conclusion? Have you not learned to reason and think critically at your university? Obviously not, as you display such a crippled logic.

          • Brett Murphy says:

            The claim that the server with the most “resources” wins and is the most powerful is a fools errand. If vendor A can produce a 1 core server with the performance of 1000 cores vs Vendor B that produces a server using 1000 cores with comparable performance. Which one is more impressive? Which one has the lowest TCO?

            Yes, Vendor A would have the superior solution even if that server had a higher cost. If we use Oracle Enterprise Edition which is $47,500 per core (list price) as the cost per core we can see why performance per core is what matters first, followed by other factors.

            If Vendor A has a licensing factor of 2.0 for their 1 core server which cost $10M vs Vendor B’s licensing factor of .5 for their 1000 core server which cost $1M. Yes, do you see how the deck is stacked against Vendor A? I am doing this to prove the point that performance per core is what customers should look at first and not be distracted by snake-oil salesmen who claim leadership results from excessive configurations.

            Vendor A
            Server cost: $10,000,000
            # of cores: 1.0
            License factor: 2
            # of Licensed cores: 2
            Cost per Oracle Core: $47,500
            Cost of Oracle EE: $95,000
            1st Years SW Maint %: 22%
            1st Years SW Maint: $20,900
            Total 1st year cost for Vendor A’s Solution: $10,115,900

            Vendor B
            Server cost: $1,000,000.00
            # of cores: 1000.0
            License factor: .5
            # of Licensed cores: 500
            Cost per Oracle Core: $47,500.00
            Cost of Oracle EE: $23,750,000.00
            1st Years SW Maint %: 22%
            1st Years SW Maint: $5,225,000.00
            Total 1st year cost for Vendor A’s Solution: $29,975,000

            I used this exaggerated configuration to show the impact for both high performing cores and high number of cores. So, who cares about cores? Customers do and should care about cores! Larry and you – well, you are master magicians whose job is to distract customers from what they should look at to make their businesses successful and only interested in what makes Oracle successful. I cite the situation in Oregon (“Cover Oregon” lawsuit against Oracle). The state of Oregon is accusing Oracle of fraud and lying claiming Oracle solely had their interests in mind and purposefully advised Oregon they could do some things which they are accused of not being able to do as well as not needing a Systems Integrator which Oracle is now claiming they do need. It appears Oregon has records that show Oracle claiming a SI is not needed. The courts will sort it out but if you ask 1000 random Oracle customers I would bet $1000 that 900+ would say negative things about Oracle, Oracle sales people, Oracle sales tactics, Oracle cost and that a majority of customers would move away from Oracle if their business or application wasn’t so dependent on their products.

            You can’t claim a CPU is best when that CPU isn’t out. We can critique what it looks like it will or won’t do. There has been plenty of analysis at this and “The Register” article you cite. I encourage readers to see those comments as they “schooled” you with data & facts, not hyperbole and wild statements.

            You are trying the trick of accusing me of what you are doing as a technique to confuse readers and call into question my analysis. I cited SAP benchmark results. Does ANY reader of this site not able or capable of knowing where to find SAP benchmarks? Here it is for those unsure: I stated that oracle licenses their products by core (implying Enterprise products). It shouldn’t be necessary for us to have to qualify if we are talking about the lower tier editions like Std Ed, Std Ed One, Express Ed, etc. If we do discuss these then most people cite them appropriately. Oracle Standard Edition is licensed by socket for solutions that include a total of 4 sockets. 4 x 1 socket, 2 x 2 socket or 1 x 4 socket server(s). For the pricing, it is available in the Oracle Store at,RP,2:PROD_HIER_ID:4509958287721805720011. Readers will see options core core and named user for perpetual or 1-5 year “annual” licenses. Some products will have different terms. I have to be literal with you or you will complain.

            The only link you provided was a link to another ET on Power8 which was very positive. You wrote 300 words in 7 paragraphs only citing that link yet you say “I always post links as you can see, I would never post claims without links as you do.”

            You say you do not work for Oracle, fine – you don’t. You are arguing for the sake of arguing because you thrive on drama and controversy. You mislead readers doing them a disservice. I don’t care if you or any other customer buy Oracle software and hardware because that is what you want – more power to you and I am truly happy for you. I like competition and choice as that drives innovation and technology in general. IBM innovates with their technology while Oracle innovates their marketing message. But, you don’t get to in the same vain claim they are the best just because you like it.

            I do not work for IBM. I work for a business partner. I worked for IBM for 4 years, 10 years at Sun before that and 8 years in the US Army. That is my resume. I love Solaris, AIX, HP-UX and Linux as OSes – each have attributes and qualities. I love Power and respect Fujitsu SPARC64. I respect IBM’s X6 and Fujitsu x86’s Primergy servers. I respect Cisco’s UCS product for their approach although inferior to traditional servers. I respect Intel’s Nehalem class and newer processors as they responded to the AMD threat and made real advances. They have however hit a wall and now scaling out like Oracle SPARC so their performance claims ring hollow to me. I am VERY impressed by IBM’s DB2 and respect Oracle. The more I learn about DB2 the more it blows me away and wonder why more customers aren’t using it over Oracle. I love IBM’s FlashSystems and SVC. I like their storage. I respect and like EMC VMAX. I respect VNX. I’m not impressed with XtremIO. Dell Compellent & HP 3PAR – not impressed. I really like and respect HDS VSP storage – damn good stuff. I could go on but want the reader to see that I try to evaluate each product on the basis of what it delivers, not what the vendor says it can do. I may respect some vendors but don’t like or recommend them because they can’t respect boundaries in a data center. If I partner with EMC I have to worry every day they will try to introduce a VCE to the customer. VCE is a farce (to me) but EMC will push it because of their sales incentives and quotas. Nobody respond to my VCE comment as that isn’t the point of this thread and I’m using it to explain to “YouKnowWho” where I stand on technology.

  5. YouKnowWho says:

    Is this the same Brett from, an IBM VAR?

    • Brett Murphy says:

      This is one and the same Brett Murphy, aka “Powerman” 🙂 I do work for SIS (Software Information Systems at We are a multi-vendor VAR but I focus on IBM’s Power technology and overall I like to think of us as an IBM VAR although my Dell, NetApp, Cisco, and other counterparts may disagree. I formerly worked at IBM and Sun for 10 years and in the US Army prior to that.

      Do you agree or disagree with my analysis? I am intrigued to know who you are “YouKnowWho”. I’m guess my twitter buddy “Phil Dunn”. I like technology and prefer RISC/Unix over x86 by and large but I also believe in data and the facts. Lots of FUD as well as propagation of overstating capabilities.

  6. YouKnowWho says:

    Here’s the trick. You need 256 licenses of Oracle for 795 while you need 192 licenses of Oracle for M6-32. As a result your IBM solution is already going to cost 33% for software licensing alone and that too for a poorer performance (SAPS.) With Oracle’s linear pricing (yes, I can vouch for that having bought servers across the entire spectrum of Sparc offerings) the Sparc hardware is cheaper as well compared to the 795. Don’t forget the BS price I’ve to pay for memory enablement (whoever thought of charging us not only for memory DIMMs but also pay for AIX license is ludicrous.)

    I’m going to throw up if I hear once more about per core performance, nobody cares. What we care is the total price/performance. Same goes for rPerf, it is a classic YMMV number to wave in front of naïve CIOs and technology managers.

    • YouKnowWho says:

      ‘33% more for’, not ‘33% for’

    • Brett Murphy says:

      Ah, there is “Trick #1” for Oracle and x86 sellers. Power has a licensing factor of 1.0 while SPARC and x86 are .5 so they are twice as much! Done, when can I expect the PO. What I like about the comment sections for these articles – and by the way, TPM – I love your writing although I don’t always agree with your analysis. You deliver tremendous value to your readers!

      Anyway, what I like about the comment section is that it lets the reader, who I hope is made up of mostly customers hear truth & facts from technologists like myself but also see and hear the gimmicks which some vendors spew.

      You may as well remain in perpetual “throw up” mode because what matters is per core performance, then efficiency of how to use those cores, flexibility to use the systems resources, the security and reliability of the underlying platform. Total Cost of Acquisition is important but TCO is what really matters as that is what determines if a businesses budgets are strained vs returning dollars back to the business to be reinvested.

      The per core SAP value for the M6 which was announced in 2013 vs the 795 announced in 2010 and to be replaced “soon” shows the 795 has roughly 25% per core greater performance. That puts the 256 core 795 at 192 cores. Further, with SMT4 which is highly efficient and allows for real parallelism vs CMT8 which is inefficient by comparison plus the efficiency of PowerVM to manage the cores & threads vs either Domains or LDOMs which both have limitations not to mention severe limitations. Happy to provide Oracles own documents as references if you want me to cite. What this translates to is that you will need to have many more cores (>2X to overcome the .5 SPARC licensing factor) than Power. I am speaking from my experience here and as Kevin Closson says you can believe me or not. I typically size SPARC servers to Power Servers where the SPARC servers have 3 – 4X more cores. So, that may look like 384 SPARC cores (with 192 licenses) vs 96 – 128 Power cores with the same number of licenses. Now, add in the cost of the Oracle software for a server with 96 – 128 cores vs a server with 192 cores and the cost of the server doesn’t matter.

      795 with 128 cores:
      Server cost: $2,000,000
      # of cores: 128
      Licensing factor: 1.0
      # of licenses: 128
      Oracle cost per core: $47,500
      Oracle license cost: $6,080,000
      22% Oracle SW maint cost: $1,337,600
      Total cost of the solution: $9,417,600

      M6-32 with 384 cores:
      Server cost: $1,000,000
      # of cores: 384
      Licensing factor: 1.0
      # of licenses: 192
      Oracle cost per core: $47,500
      Oracle license cost: $9,262,500
      22% Oracle SW maint cost: $2,037,750
      Total cost of the solution: $12,300,250

      Notice I made the 795 twice the price of the M6-32 just to give it the advantage. Even with Oracle favoring SPARC servers or better said being punitive against competitors the software will (ie WILL) cost less on Power at acquisition. The annual maintenance bill starting with year 1 and EVERY year thereafter is 45% more at $2M per year.

      I’m happy to compare a sales proposal for a decked out 795 vs a decked out M6-32 if you want. Remember, I’ve challenged you and Oracle (since they are not your employer 😉 to a technical debate done in public, recorded with a panel of industry experts to score each topic. You bring whoever you want. I’ll bring my backup set of dry-erase markets, macbook and laser pointer ready to battle. I’ll probably have my wife and kids in the audience so they can cheer me on as I rock your (whoever is on the Oracle side). There is surely a Oracle DE, marketing person or some uber fan willing to take me up on this. The Hilton hotel at Chicago O’Hare is ideal for anybody to fly into. We can host it there. I’ll ensure we both have 2 whiteboards, a butcher block notepad with the tape backside to put on the wall. I’ll provide food and beverage during the day and the Oracle team can buy dinner and drinks so we can hopefully have a cordial meal afterward.

      • YouKnowWho says:

        To summarize your response, only you are the authority on both Power and Sparc. Even the latest version of Sparc, even after you’ve left Sun. No one else counts. I’d never say that about myself. I’m convinced I have flaws in my thinking which leads me to adjust my views based on the market. It is never holy to me.

        There is not a substantiated claim that you make in your latest response. Just that PowerVM is better (said who?), SMT4 is better than CMT (I guess so), AIX (the one I’ve to pay for memory usage) is better at managing threads (yet no benchmarks with large thread count, always LPAR’d), … SAP benchmark says 795 is slower than M6, yet you claim to deliver the same performance with 128 Power cores. Makes no difference what year you published your benchmark, until you publish your latest that’s all I’ve to go by (I wonder if anyone is left at IBM STG to even run new benchmarks.) Most of AIX developers have already bolted. I hope you’re not promising new features in AIX to your customers. It’s your reputation you’re staking, IBM has no reservations about moving on. As someone once said “Only the middle class has morality, rich don’t care, and the poor can’t afford it”

        5 years back IBM was beating the drum that Oracle would sell Sun hardware business. Look who’s selling stuff to the Chinese? IBM is innovating so much in semiconductor, they’re willing to give money to a buyer to take it off their plate. v7000 is next on the chopping block. You must admit that irony is not lost on most of us. At least those that have been around the block a while.

        I remember charts from IBM in 2009 saying how Power6 based p560(?) was 3.5x faster than UltraSparc based M5000. After 4 iterations (over which IBM reduced their prices by a whopping 70% with different configurations of the solution) we chose to go with what IBM claimed was a inferior configuration of M5000 (a configuration that Sun never changed once and cost lower than the last IBM bid). IBM team assured us that we’d have performance problems. After 5 years, I can safely say there hasn’t been a single performance issue. Not once, nada, zero, … I’m supposed to trust my IBM team, right?

        BTW, how’s your buddy John Mayer @ IBM? He visited us to preach me the Power business value 3 years back. I still have his marked up copy of the Sparc roadmap that he put in front of me and other unsuspecting tech executives. I chuckle every time I look at it.

        We both need to get to our day jobs. I’m sure you’ll sell Power gear and there’s nothing wrong with that. Last I checked, it’s a free country. Nobody can be stopped from buying expensive gear.

        • Brett Murphy says:

          We can both agree on your last paragraph! I’ll start from the bottom or your rant. I do not not John Mayer – he may or may not be my buddy. If he’s lazy, doesn’t know the technology failing to challenge the competition then we probably wouldn’t be buddies. If he is hard charging, aggressive and loyal to his partners (like me) then we could be buds – I think you and I could be buds. I’m the conservative right minded thinker and you are the crazy from left field liberal. 🙂

          So, if you are a customer then see for yourself – let’s set up a meeting. On my dime I will visit you. Bring in your team and hear my story. Give me an hour to present the “story” then you *all* can go to town on me – rapid fire. I love it! Steve Sibley of IBM is somebody I look up to as he is a master of objection handling.

          I’m happy to have a Power6 conversation with you but there is no need to talk about a 2007-Feb 2010 technology. You will see when I am not restricted by brief comments on websites that I freely discuss pro’s and con’s of everything – Power, Mac’s, Hybrid cars, Chick-fil-a (they don’t have any cons btw 🙂 )

          We’ve gone back and forth on Power7 and Power8 vs SPARC M6 & M7. Don’t need to repeat myself. I do not intend to misstate or overstate anything – if I do I own up to it and will correct the record. Credibility and losing is more important than lack of integrity and winning. I serve a boss greater than I who I strive to please.

          Amazing that it was 5 years now. IBM was this ” ” close to buying Sun. Wow, Solaris on Power. AIX is great but Solaris has a strong and loyal following. It would have been interesting to see what IBM would have chosen if presented with two UNIX OSes on Power.

          This is the platform to discuss PowerVM vs LDOMs nor SMT vs CMT. Maybe a blog – or if you and Oracle take me up on that technical debate.

          I am an authority on Power technology! Doesn’t mean I do not make mistakes or misstate things. I am a Power Champion and self-proclaimed Evangelist – I speak of Power with “fire and brimstone”! 🙂 There are many other Champions and subject matter experts out there. Most focus on direct discussions with customers while others like myself step into the public domain to clarify and defend Power technologies and if given the chance promote it. I am not an authority on SPARC like I recall Brian Wong and many others in my Sun days. However, because I do know Solaris and SPARC from my many years of experience as well as the underlying technology from VME based servers to the release of T1000 servers. Due to my direct work and training I do understand it enough to know what it is doing, what it claims, what it overstates and having been at Sun then IBM a perspective to be critical of marketing hype. Again, if you ask me to be critical of IBM technologies I can do it but these forums are the place to air dirty laundry. Every vendor has their successes and shortcomings.

          Lastly, I am glad you do not consider yourself an authority on Power or SPARC as I can attest that your thinking is flawed! 🙂 Come on now, you tossed me that softball!

  7. DM says:

    Seeing as this discussion has completely veered off topic and on to comparing competing products, I’d like to ask Brett Murphy ( or anyone else here ) what they think about the added costs everywhere in the form of additional licenses, enablements and activations ( VET codes ) etc on POWER and AIX solutions unlike on Solaris where you buy the server and most everything else comes included in the cost of the server. Below are some examples comparing AIX software/features/solutions that need to be licensed or purchased separately compared to Solaris ones that do not incur additional costs.

    ( Not saying they are all like for like, but they would be the closest related competing technology. )

    PowerHA – Sun Cluster
    GPFS – ZFS
    LPAR ( PowerVM Express, Standard, Enterprise ) – Oracle VM SPARC ( LDOM )
    VIO ( PowerVM Standard ) – Oracle VM SPARC ( LDOM )
    Partition Mobility ( PowerVM Enterprise ) – Live Migration
    WPAR – Solaris Zones
    XL C/C++ Compilers – Sun Studio
    Express, Standard or Enterprise editions of AIX – Solaris ( You get the lot )
    PowerVC, HMC, Systems Director etc – Oracle Enterprise Manager

    • Brett Murphy says:

      Good questions. Your questions about licensing are focused on the enterprise servers. Entry & Mid-range servers come with all procs activated and a choice of what & how many OS licenses for AIX, IBM i or Linux. The VET code is specific to PowerVM or virtualization which is again a enterprise thing. Entry & Mid-range servers have a choice of no-virtualization, or 3 levels of it depending on feature requirements. Enterprise servers offer the flexibility to have activated resources and COD (ie dark) resources for cores and memory. In order to support COD you have to separate out the the license to activate from the dark resource. Otherwise, if you have active resources it becomes a challenge to fence that resource from other activated resource customers are entitled to. Along with activating cores, the new or COD resources then can be licensed with different OSes and other software products. Power enterprise servers let you have 16 active cores and only 1 AIX license (implying just 1 core is licensed for AIX). The other 15 could be used for Linux for example or maybe some virtualization features. Similarly, If I have 16 active cores with 8 AIX licenses I could have just one PowerHA license. 1 on Server A and 8 on Server B totaling 9 PowerHA licenses to support a Active / Passive cluster. Competitors like VCS tend to license by class and # of sockets of a server if not all cores on both sides – their choice but it gives IBM power software products a licensing advantage.

      So, your statement that Solaris comes with everything is similar to the entry level and mid-range servers (ie 710 – 750 in the Power7+ family and the new Power8 S814 & S824 servers).

      I accept your list of product comparisons. Some are more comparable than others. The first one is PowerVM vs LDOM’s. There is no comparison as PowerVM is far more robust being flexible, scalable, secure and efficient. I’ve been following the Solaris Live Migration feature. I’d like to talk with you more on how effective it is. LPM (Partition Mobility) is very flexible across server generations like using it from a Power6 to a Power8 server and vice versa. AIX with it’s WPARs like Solaris Zones also offers Application Mobility which lets users move a WPAR (ie a Solaris Zone) from one VM or Server to another – live!
      AIX Exp & Std Editions are the same – it was just targeted at different tiers of servers. Enterprise Edition included quite a bit of extra software bundled with AIX – it is a excellent value for customers who are using those features vs buying ala carte. Your association of PowerVC, HMC & Systems Director are not quite accurate but I see what you are saying. Sure, they are infrastructure products to support the compute solution.

      Some (ie most) of these products can be purchased a core at a time. That’s good because I can have hundreds of VM’s on a server, all having different needs. So, being able to license just the products needed ensures customers they don’t overpay. If it is included in every server for all cores then the cost to develop and support is built into something.

      As you have seen me respond to YouKnowWho, it sounds interesting that Solaris can say they bundle all of these things with the OS but those distract from the big picture which is both the cost of acquisition and cost of ownership.

      Examples of this could be Power server running Oracle. This bothers SPARC cheerleaders but far fewer Power servers & cores are needed for workloads than on x86 or SPARC. Fewer servers, fewer cores, fewer DB licenses, less maintenance equals lower cost. Doesn’t mean SPARC & Solaris are bad and I guess I shouldn’t expect Oracle & SPARC sellers to admit that or that Power/AIX is better. It’s Ford vs Chevy. Some buy the best and other buy idealogy.

      Did I answer your questions? To your satisfaction?

    • YouKnowWho says:

      Exactly my point when I said ‘memory activation…’

      • Brett Murphy says:

        You realize SPARC servers have a history of activating cpu & memory right? They have had COD? Look at this COD Users Guide for the M9000. You remember that server – the one Sun, then Oracle OEM’d from Fujitsu because they made a better SPARC server than Sun could – faster, more reliable and less expensive.

        If that is too “Legacy” for you then how about the CPU COD feature for the Fujitsu M10 server?

        I’m guess they don’t have COD for memory because the architecture limits it. The inference is that you license (ie activate) cores individually with all subsequent software products & maintenance as appropriate.

        Now, Oracle has their T-series only which is a evolution of the Niagara chipset starting with the T1000 products. They were over-hyped entry level servers until the S3 core was added putting it on par with IBM Power5 & Power6 servers. The current T5 & M5/M6 are nothing more than bigger versions of entry level servers. They may have lots of cores & memory but they are not in the same class as a Fujitsu M10 or the SPARC M-class servers – especially the M8000 / M9000. I conclude they do not have COD for two reasons 1) Can’t let customers consume more cores if they aren’t easily available. When you sell software by the core with a hungry audit team you want customers to “over consume” so you can either sell more or collect via software audits. 2) Since the technology originates from lightweight entry level servers the architecture doesn’t support it. I’m assuming the latter is correct but happy to have one of you tell me why a customer would want a T5-8 or M6-32 to come fully activated.

        • YouKnowWho says:

          @Brett, you need to calm down a little bit. Marketplace has spoken about Power, a whopping 50% erosion in 8 quarters. Sell those Power boxes and beat Sparc into submission/extinction. Fair challenge.

          I thought I was a bigot/fanboi, man you make me look like Mother Teresa.

          ‘Nuff said.

          • Brett Murphy says:

            Thank you for your concern. Unlike SPARC, Power servers deliver more performance per core while lowering the price. Power technology actually consumes itself as a result. Needs to expand the market which is why you are seeing such a push into the Linux space. Get a clue!

          • YouKnowWho says:

            How about that monster asset sale and the continued downward momentum for Power? Market has spoken, you can pout all you want about the value; we’ve moved on.

          • Brett Murphy says:

            Yes, there is the Oracle marketing machine spreading the FUD. Amazing how IBM is credited with creating the term some 30+ years ago while Oracle seems to have hijacked it these days.

            If you are referring to the sale of x86 to Lenovo and IBM’s Chip business to Global Foundry I see what you are trying to imply but as usual it falls flat.

            IBM didn’t survive for 100+ years by holding on to unprofitable products or businesses. They shed tabulators, punch cards, printers, disk drives, x86 and now their chip fabs because of changing market times. x86 is commodity requiring vast volume sales which is better suited for a company like Lenovo vs IBM. Companies like TSMC and Global Foundry specialist in chip manufacturing or “chip making for hire” so they have an efficient supply chain and business model. IBM still requires the output from these plants so essentially outsource them to somebody who specializes while you still own the IP to drive the profitable parts of the business like Infrastructure, Software and Services.

            I understand your need to diminish Power any way you can. Why wouldn’t you when it delivers performance roughly 3X more than your M6 server. The per core performance of an E870 for the SAP S&D 2 tier is 996 Users per core or 5425 SAPS per core. Compare that to the M6-32 with 2067 SAPS per core and 364 Users per core. The T5-8 is just 1726 SAPS with 312 Users per core. Even the massive 640 core Fuji M10-4S delivers just 1319 SAPS per core and 239 Users per core.

            I know, I know….per core, per core. You are tired of hearing per core. Truth hurts. A 8 core E870 delivers 79,750 Users and 436,100 SAPs which is 2X of the 128 core T5-8, on par with a 256 core M10-4S and just under a 192 core M5-32 at 472,600/85,050 respectively.

            Yes, your 1024 core M7 using 10 Billion (Beellion) transistors is just around the corner (oh my) and I’m sure it will just eviscerate Power8 by making up for the 3-4X performance gap and taking it to the next level by 2-3X or at least Oracle Marketing will say it will.

        • YouKnowWho says:

          Wait a minute, are you saying COD is the same as activation fees? COD is having capacity on standby, it isn’t even visible to OS. Activation fees is: first you buy memory at full price and then pay AIX activation fees to use the memory.

          In Solaris, you see it and it can use it. I don’t need to write another check to my AIX sales person. Same goes for processors activation fees. Reminds me of Netapp pricing which goes something like “you mean you really want to use the NFS to access your storage, I thought sftp is all you need to run your database”

          BTW, have you heard of ‘psradm’ command? Using it you can turn off processors all together. I can also use LDOMs to partition the processor into separate units that can be licensed independently at a guest level. So you can buy a server with 256 cores and license only 12 cores for DB2 LUW (see Solaris can run even your database ;-)) or Oracle database which you seem to so dearly love since you can’t stop talking about it in your comments. I’ve to congratulate you since you recognize a good database when you see one.

          • Brett Murphy says:

            Were you dropped at birth? How in the world you come up with some of your comments is baffling. Yes, Dark resources on enterprise Power servers are by definition COD resources. Buy the activations and licenses for just what you need to get all of the *other* enterprise features. If you need more you can use them by the minute, by the day, even temporarily for up to 30 days at no cost or you can also permanently activate them. I can also transfer the “active” activations to other servers live and Power servers can *share* COD credits. Once again, Power delivers flexibility while SPARC regresses by not offering it anymore.

            LDOMs are the equivalent of “pink slime” to Power’s PowerVM “Kobe Steak”. PowerVM can create as many Shared Processor Pools isolating cores to each as there are cores in the server. It can then share and move cores between them dynamically. Each core can have up to 20 distinct and isolated VM’s running AIX, Linux AND IBM i. Power can deliver a Quality of service to each and if needed it can share unused resources with others and consume more than entitlement if available. PowerVM can add/remove cpu, memory and I/O dynamically for AIX, Linux and IBM i.

            When LDOM’s get to be 70% ground beef then come talk to me otherwise they are a chihuahua biting at PowerVM’s ankles.

            I’m happy DB2 runs on SPARC. It also runs on x86, Mainframe and others. Woohoo! IBM treats each vendor the same based on that platforms capabilities. Not only could I run with 1 license of DB2 on a 256 core 795 using your example of 12 SPARC cores, if that was running on Power8 I would need 3 cores to run the same DB2. And – DB2 will not only be less expensive but offer more features like BLU Acceleration. Where is your owner Chihuahua? I need to go put up some “Found, Lost Chihuahua” signs in the neighborhood.

            Lastly, Oracle is a very good database. DB2 is just better. Furthermore, as good as the database is, it is owned by Oracle which is akin to the mafia….just legal – unless you live in Oregon!

  8. DM says:

    I was intending on staying out of this, but having quite an extensive amount of experience on the ground working with both POWER & SPARC as opposed to having studied the features in sales brochures, I feel compelled to add my thoughts here. I’m not so sure you’ve actually used LDOMs @Brett after those comments at least not anytime recently ?

    Yes they came much later than PowerVM and LPARs (IBM go way back with virtualization with mainframes and you will see this heritage in AIX and Power), but these days you’re not missing anything that you really need on SPARC when it comes to virtualisation, the whole point of which is to increase your hardware’s utilisation while keeping workloads separate and secure. Power/AIX possibly make things a bit easier to use, better management tools and generally bit more consistent between generations of hardware, but feature wise, it’s most certainly all possible on SPARC too, if you’re a competent sysadmin. You can’t run Linux or IBMi of course. Otherwise it is “flexible” enough to even let you configure yourself into less than ideal and possibly un-supported configurations. Power/AIX is more controlled, bit more idiot-proof, where I can entrust a lot more with junior staff without additional work on my part.

    Cores are distributed not shared between LDOMs. With the large number of cores SPARC boxes come with these days, there’s certainly no need to share them out among LDOMs as you’d do with Power LPARs. You don’t need to be concerned about noisy neighbors that share the same share processor pool either as you would on Power. And you can’t argue there will be LESS virtualisation overheads doing it this way. That is, a core either belongs to an LDOM or it doesn’t. No additional work required by the hypervisor. You can use zones to further drive up utilisation if you need to and there’s much more fine grained resource controls available here. Much more than AIX. In fact AIX has been late to the OS level resource controls party by several years compared to Solaris and this brings me to Zones/Containers.

    When it comes to Zones vs WPARs, Zones have a longer history, is more mature and is much more widely adopted than WPARs in any environment I look at and I support thousands of servers across multiple clients. And AIX & WPARs can’t hold a candle to what you can do with a ZFS/Zones combination.

    I don’t want to go into per core performance as it’s been done to death already. Like YouKnowWho has said above, it’s the total price/performance that matters and per core performance doesn’t always apply to all situations.

    Going back to my earlier comment and your response about IBM’s model of using feature enablements, entitlements and licences, I do get how the licensing works. But in a time where a company credit card can get an end user what they need in the cloud on the fly with the least amount of hassles, this sort of complexity is not going to help IBM sell Power gear. After buying a server I don’t need IBM to tell me how I can use it and what I can or can’t install on it. You don’t have any of this head-ache with SPARC and Solaris.

    • Brett Murphy says:

      @DM – thanks for your comments. Was it the “pink slime” or “chihuahua” comments that prompted you to respond? 🙂 I was trying to push “YouKnoWho’s” buttons.

      Seriously though, and I hope YKW will chill on the back and forth as he doesn’t know how to yield a point or see’s a squirrel and introduces another topic from nowhere.

      I haven’t used LDOM’s but have done extensive reading on them since there release. I would add another feature of the hypervisor would be to control software licensing which PowerVM does. And, yes – all hypervisors will let you screw yourself….just because you can do it doesn’t mean you should 🙂

      I disagree with your characterization of “sharing” cores on PowerVM. On every server since Power5 you can have dedicated cores AND shared processor pools. Since Power6 you can have multiple SPP; as many as 1 per core or as few as one SPP for all cores (any combo in between). The cores are not physically isolated like in Domains but logically. However, PHYP separates the cores and cache lines while interacting with the hypervisor to pass TCP, disk (ie vSCSI & vFC) and other data between LPARs (by design to share I/O). Noisy neighbors only come in to play if you choose to put them together. I’d characterize it to say if the workloads compliment one another they will do well in SPP. If not, then give them dedicated cores alongside the SPP that can have other VM’s.

      For Zones vs WPARs. Yes, WPARs came much later..about 7-8 years I think. Interesting though, because of the sub-capacity for years at .1 of a core to a VM, OS virtualization has never been that critical. Since SPARC has always lacked anything besides Domains for years (til now w/LDOMs) that was the only way to get fine grain resource allocation of cores. I like WPARs (and Zones). They have a couple of features now that Zones do not. Just saying to say there are differences between a Ford & Chevy.

      Ultimately, features like reliability and security are critical influencers the price/performance are major decision points. YNW always wants to go to the 780/795 servers for price comparison. Because Power cores are so powerful they can do more with less. Thus, a 16 core Power server vs a 64 core SPARC. YNW and Oracle will try to compare a 64 co Power server to their 64 core and compare cost. It’s just not done that way. TCA of Power can be equal or better to any SPARC solution. User pricing is user pricing so yes there are some areas that any solution is comparable. But, when software is core based, Power will almost always have a better sw TCA and will dominate the TCO.

      If SPARC is so powerful then why does Oracle give it a pricing factor of .5? Either because it isn’t that powerful per core which is how they price software or they have to do it to give SPARC any chance to be TCO competitive. If you work with SPARC and current generations of Power then you know it is both.

      Not every Power option and feature is right for every situation. You mention a cloud environment – all of that is possible with Power. If you are making the argument that a customer should buy a 384 core M6-32 so a customer can have extra cores for end of month processing I want to sell you the computers then because you are overpaying. I am only talking about the enterprise servers now such as the 770, 780 & 795. Instead of paying for the OS and support on 384 cores you can pay for the resources that are need 365 days a year and for those resources only needed for 24 days (12 mo x 2 days for ex) they can be added using COD. A COD activation can be turned on immediately. COD features include all licensing & support costs for OS, virtualization, HA, etc. When used, just like SPARC the customer has to license and software like WebLogic, WebSphere, Oracle or DB2 appropriately. DB2, btw can also be used used as COD. Unlike Oracle where you buy the license with DB2 you can buy a daily license. I think with a bit of one on one discussion you would see how and when to use COD.

      Enjoyed your responses though.

      • YouKnowWho says:

        I must agree that you’re entertaining. Let’s see how the STG revenues stack up this quarter? Another 20% slide 😉 Before long SIS will be selling Chinese junk at discounted prices (is that an oxymoron?)

        BTW, I now know how IBM is moving Power8 boxes. You can’t wait to give them away for evaluations. Power will go the way of Lexmark, PC, x86 servers, ProtecTIER, SONAS, …

        • Brett Murphy says:

          More ‘hot air’ from YKW! As a BP, I don’t make any money to give a server away. What I am giving away is a opportunity for customers to reduce their Oracle licensing by 1/2 to 3/4th! Again you are wrong with your claims – but being a Oracle mouthpiece it is expected for one to say whatever is necessary – just look at “Cover Oregon” to watch Oracle at it’s best!
          I guess I could always switch to selling “Redwood City junk” 🙂

    • Brett Murphy says:

      Btw, I have a POC Power8 server I would be happy to let you test on to see both performance and features. We would provide consulting and training to go with it. I wouldn’t just drop it off and say ‘go’. It’s not that I like or want to say Solaris / SPARC or bad, I try to be pragmatic in showing where and how Power is better. Charts & words can say a lot. Testing on a solution is the proof point.

  9. Shadow_Cat says:

    @ Brett

    You like how Intel responded to the AMD threat?

    AMD is the one that brought computers into modern multi-core era, yet you are rooting for the criminals that is masquerading the technology forefront as theirs?

    You come off sounding very irresponsible in broadcasting quality information.

    You deliberately failed to write about Intel’s illegal monopolistic tactics that set AMD back for years to come.

    You failed to include how Intel paid Dell four billion-dollars, back in the year 2000, not to sell desktops with AMD chipsets.

    Hurting the frontrunners of the technology is hurting all mankind.

    Just like the zealous, envious, and jealous Greeks who realized that the Egyptians were over 3,000 years more advanced then them,they conspired to reset the clock from BC to AD; OK, now, let’s start with 0 AD.

    You support companies with no scruples, you have no morals.

  10. Fenton says:

    Well at least if you buy your oracle/DB2 license from SAP you only pay a proportion of your SAP license so cores don’t matter (DB2 is a lot cheaper)

    Performance per core does matter, especially if you have lots of heavy single threads (I know inefficient programming). But a lot of software these days still relies on single threaded performance.

    Where I see the large multi socket boxes showing their real strength is where you need the grunt for a single OS image that requires more processing power than your run of the mill 2/4 socket boxes can deliver. And that is pretty much a niche market.
    That is why 2/4 socket systems are king.
    Which is exactly where x86 sits.

    But back to Power vs Sparc.

    As somebody who has supported SAP customers for far too many years, my experience is that IBM based systems, generally require fewer people to feed and water than Solaris based systems and also tend to be the most stable.

    I’ve been bitten bad by oracle per core licensing, and the T2000 systems hurt one of my customers very badly.

    Solaris on x86 with containers is the biggest headache one of my customers has. Every new project now uses AIX/DB2 backend with cheap x86 application servers.

    And on a personal note, I’d rather distribute all of my workloads over multiple boxes, rather than stick em on one large box, because shit will happen and I’d rather loose a few mission critical apps whilst they fail over than all of them.

    Give me a rack full of Power/x86 blades or pizza boxes any day!

  11. Harold Cain says:

    My experience BTW with blades is that mtbf is not very high. You may have sliced and diced an app to the point the loss is not great but you spend all your time replacing them.

    Newer large system running ldoms can have so much redundancy that the likelihood of any of them going down and staying down is VERY low. If you compare the cost of all those single point of failure blades and pizza boxes to one
    large system, the virtual domains always win.

Add a Comment