Advanced Computing in the Age of AI | Friday, March 29, 2024

AWS Joins Cloud Memory Push 

Public cloud vendors are focusing their infrastructure upgrades on the growing number of memory-intensive enterprise workloads. Earlier this week, Google rolled out virtual machines that support persistent memory for in-memory workloads.

Amazon Web Services (NASDAQ: AMZN) countered with a new “high frequency” instance generally available on its Elastic Compute Cloud. The cloud giant said mid-week the new instance, dubbed “z1d” (numeral “one”) targets enterprise workloads requiring a bunch of memory along with single-threaded performance.

Workload examples include chip design, financial transaction simulations and relational databases.

AWS also filled out its latest memory push with its next generation of memory (R5) and instances with local storage (R5d). The memory instances are designed to improve price-per-gigabyte performance for memory-dependent applications, the company said.

AWS and other cloud vendors continue to add memory capacity as customers look for additional processing power to handle ever-larger data sets. AWS added it is seeing increased demand for single-thread performance in which a single set of instructions are executed over a specified amount of time.

Matt Garman, vice president of compute services at AWS, said the new instances deliver “sustained all core frequency” of 4.0 GHz, which the company claims is the fastest cloud instance currently available. That capability would decrease the number of cores required to handle memory-intensive enterprise applications, Garman added.

The new instances run on Intel Xeon Scalable processors (Skylake-SP) customized for AWS servers. Along with chip design and relational database workloads based on pricey per-core licensing fees, the new instance targets high-performance computing simulations such as financial risk analysis.

AWS said z1d instances are available in six sizes, with as many as 48 virtual CPUs and 384 GiB of memory along with up to 1.8 Tb of NVMe-based local storage. The new cloud instances are available now, with a bare metal version to follow in several weeks.

The new instances are based on the AWS Nitro System, a collection of hardware and software components aimed at reducing virtualization overhead. The cloud vendor said z1d, R5 and R5d instances can be used on demand, reserved or “spot” instances.

The new AWS E2C instances are available here.

Meanwhile, Google, too, said its in-memory push in the cloud is backed by Intel chips. The cloud rival announced a partnership earlier this weekwith the chip maker and SAP (NYSE: SAP) to add cloud virtual machines supporting the upcoming release of SAP HANA workloads running on Intel’s Optane DC persistent memory. Google (NASDAQ: GOOGL) said the cloud virtual machines would be powered by Intel’s (NASDAQ: INTC) next-generation Cascade Lake processor, the successor to its current generation Skylake scalable processor.

Google notes it was the first cloud provider to offer Skylake.

About the author: George Leopold

George Leopold has written about science and technology for more than 30 years, focusing on electronics and aerospace technology. He previously served as executive editor of Electronic Engineering Times. Leopold is the author of "Calculated Risk: The Supersonic Life and Times of Gus Grissom" (Purdue University Press, 2016).

EnterpriseAI