What Eucalyptus Brings To HP’s Helion Cloud
With Hewlett-Packard being such an eager supporter of the OpenStack cloud controller in the past couple of years, and increasingly so as it tries to build up its cloud portfolio to take on the public cloud operators that are moving away from its ProLiant gear and towards custom machines, it might have seemed a bit odd that the company would acquire Eucalyptus Systems. But the deal makes perfect sense for both companies.
Eucalyptus is one of the first cloud controllers to hit the market and get commercialized, and stems from work done at Rice University and the University of California at Santa Barbara more than a decade ago, Eucalyptus, the company, was founded with $5.5 million in venture funding from Benchmark Capital back in 2009, the same year that the cloud controller was included with the Ubuntu Server distribution of Linux from Canonical. The components of the system are written in Java and C, with C used where performance is crucial – the cluster controller and node controller modules, to be specific.
The interesting bit about Eucalyptus is that it was the first big open source cloud controller, and shortly after it was bundled with Ubuntu it had well over 10,000 installations where people were playing around with it, testing it to see how it might be used to convert their server farms into something that looked and felt more like the cloudy compute and storage capacity on Amazon Web Services. To this day, Eucalyptus prides itself on offering the broadest compatibility with the APIs and features of AWS, although it does offer additional services not available on AWS such as support for VMware's ESXi hypervisor on clouds. For the past two years, Amazon and Eucalyptus have had a technology agreement whereby Eucalyptus gets help ensuring AWS compatibility, and while this is not the same thing as having an actual chunk of AWS to run as a private cloud, using Eucalyptus is about as close as you can get without breaking into an Amazon datacenter with a pickup truck.
In a real sense, OpenStack, the current cloud controller darling of the open source world, got its start because NASA was unhappy with the scalability of Eucalyptus and with the fact that some of the components of Eucalyptus were not open sourced. NASA was spearheading the cloud efforts of the US government at the time, and among other things was keen on having a completely open source tool for managing its clouds and one that it could have more of a hand in directing. When OpenStack took off, Canonical, like all of the other Linux distributors as well as other companies that wanted to mainly focus on peddling OpenStack distributions or servers, all jumped on the OpenStack bandwagon, leaving Eucalyptus to itself. This suited Eucalyptus just fine because it hunkered down and stick to its AWS-compatible knitting. OpenStack has pretty much ignored AWS compatibility, and that is understandable given that Rackspace Hosting, the other big founder of OpenStack, wanted to have an alternative to AWS, not create a clone of it.
Eucalyptus still gets tens of thousands of downloads per month of its cloud controller, as best as the company can figure. One of the biggest Eucalyptus installations is at Nokia Solutions and Networks, which was on pace to swell to 100,000 cores when EnterpriseTech reported on it last fall. An unnamed financial services firm has over 50,000 servers in a cloud that it is managing using Eucalyptus that has hundreds of thousands of cores. Marten Mickos, the CEO at Eucalyptus Systems, tells EnterpriseTech that the company has fewer than one hundred enterprises that are paying for support contracts and add-ons for Eucalyptus, but the business has been growing steadily because AWS compatibility is important for many customers building private clouds and some service providers building public or hybrid cloud services.
Mickos was CEO of MySQL, which he sold to Sun Microsystems in January 2008 for $1 billion as the latter company was trying to get a software strategy together. At the time, MySQL had over 100 million downloads and was the M in the LAMP stack that was ubiquitous for Web applications. It is pretty certain that Hewlett-Packard did not pay anywhere near that for Eucalyptus, but it could be on the order of $250 million if HP paid five times the venture to get its hands on it. (The company raised $55.5 million in three rounds of funding Benchmark, New Enterprise Associates, Institutional Venture Partners, and e.ventures.) Terms of the deal were not disclosed, but HP said that it expected the Eucalyptus deal to close in the fourth quarter of its fiscal 2014, which ends on October 31. If it was a sizable chunk of change, it will report it in its financial filings with the Securities and Exchange Commission.
Hewlett-Packard introduced its $1 billion investment in OpenStack-based clouds under its Helion brand back in May, including its own distribution of OpenStack as well as managed clouds based on Helion and professional services to help customers install Helion. HP has ramped up its OpenStack programming efforts and says it will be the top contributor to the "Juno" release that is expected in mid-October. At the time, Martin Fink, who was running HP Labs as well as being the company's chief technology officer, was also steering HP's cloud efforts. But once this deal closes, Mickos will take over the HP Cloud business, which spans groups and divisions, and will report directly to CEO Meg Whitman.
Why did Eucalyptus do the deal with HP? Mickos explained the situation to EnterpriseTech in this manner: "There has been a lot of interest in Eucalyptus ever since I joined. And we always had an ambition to build an independent business as we have been doing. But we did learn in the past few years that this market is changing in a way that does not benefit a small, independent cloud software vendor. Partly because so many software companies are being acquired and partly because customers are becoming so much more verticalized. The prefer buying a whole lot of stuff from one vendor, and we were open to the acquisition because it is in the interests of the customers that we are serving. We had a lot of customers asking us to do managed clouds, others were asking us for hardware recommendations. These were signals to us that it would not be a bad idea to have an end-to-end solution ready."
This vertical integration, which to a certain extent was demonstrated so brilliantly by Apple consumer products is an idea that "the world is in love with now," according to Mickos, but IT could shift back towards more horizontal, best-of-breed in another decade or so if innovation stagnates. But for now, Eucalyptus has to deal with the vertical forces that are here, and hence the HP offer was taken.
So, in addition to getting an executive to run its cloud business, what else does HP get? Mickos tells EnterpriseTech it is two key things. The first is not just AWS compatibility for a private cloud, but the ability to run Eucalyptus and OpenStack side-by-side, in the same datacenter, and have them work together in a useful manner. Eucalyptus was already bringing in support for Ceph Swift object storage and Riak NoSQL data stores that are commonly deployed with OpenStack, and integration with Midokura SDN software was underway. The Eucalyptus stack was being modularized so that elements of OpenStack could be swapped in and pieces of Eucalyptus swapped out as need be. "We won't run as a layer on top of OpenStack, that just creates problems and layers of abstraction that we don't need," Mickos adds, just to be clear. "But we will be able to exist in the same deployment, and customers with OpenStack clouds will be able to carve out a portion of it to be Amazon compatible using the Eucalyptus components."
The other thing that HP gets with Eucalyptus is a user interface and cloud controlling experience that Mickos says is better than the fit and finish that OpenStack currently offers. "I don't think that any cloud platform has the level of ease of use that we have," he brags. Whether or not this is true is up to customers to judge, but in the past several releases of OpenStack there was plenty of grumbling about ease of use and the community has slowed down and worked on these issues rather than just focusing on raw features exclusively. In the long run, this OpenStack-Eucalyptus hybrid will need a unified management console that masks the differences between the two. This is work that Eucalyptus would have started with its 4.1 and 4.2 releases, which remain as they are on the roadmap, but further out, this will be decided by the unified HP-Eucalyptus team.