Facilitating the Spread of Knowledge and Innovation in Professional Software Development

Write for InfoQ


Choose your language

InfoQ Homepage News AWS Offers 4TB Memory Virtual Machines

AWS Offers 4TB Memory Virtual Machines

This item in japanese

Amazon has launched its largest EC2 machine: the x1e.32xlarge instance. This new instance offering has 4 TB of DDR4 memory, a quad socket Intel Xeon E7 8880 v3 Haswell processors and runs at 2.3 GHz (128 vCPUs) with large L3 cache. According to AWS official documentation, its network bandwidth should reach a maximum throughput of 25 Gbps.

Jeff Barr, AWS chief evangelist, announced on the September 13 launch that these instances are designed to run SAP HANA and other memory intensive, in-memory applications. Many AWS customers like Coca-Cola, Cisco and other large organizations are already running production SAP applications on the existing instance types. For example, the Kellogg company runs complex simulations to predict which of its promotional activities will be the most effective.

In order to optimize complex analytic processing, SAP HANA runs entirely in memory since accessing a disk is 10,000 to 1,000,000 times slower than accessing data stored in memory. This is the reason for the increasing demand of memory to handle larger workloads. For example, with Intel Ivy Bridge v2, SAP recommends a ratio of 256GB of memory per CPU.

In order to process and store far larger datasets, companies running SAP HANA workloads on smaller X1 instances can now migrate to a bigger VM and scale up their production systems.

Image: SAP HANA Cluster on AWS

This new type of machine is considered as a memory-optimized VM, just like X1 or R4 instances that are a good fit for in-memory databases, high-performance/real-time databases, data-mining applications, large-scale in-memory caches and Hadoop/Spark clusters.

Compared to other types of AWS VMs, memory-optimized machines are better suited when a production application's memory need is huge. The other types of machines could be categorized into compute-optimized, storage-optimized and accelerated computing instances.

For now, x1e.32xlarge instances can be launched in only four AWS Regions:

  • US East (Northern Virginia)
  • US West (Oregon)
  • EU (Ireland)
  • Asia Pacific (Tokyo)

Microsoft Azure's largest machine has 2 TB of memory and Google offers a max of 416 GB, so AWS has now the largest machine in terms of memory. However this comes with higher costs: e.g. US$26.688 per hour in the US East region and US$38.688 in Tokyo.

Rate this Article