AWS beefs up cloud for super-fast data processing

Latching onto the trend toward in-memory storage for real-time computing, Amazon Web Services (s amzn) has added a new type of virtual server. The new option — the 10th such available on the EC2 offering — is called the High-Memory Cluster Instance and includes 88 EC2 Compute Units of compute capacity (running on two Intel (s intc) Xeon E5-2670 processors, two 120 GB solid-state drives of instance storage and 244 GB of RAM.
It’s designed with speed in mind for uses such as in-memory analytics (including on SAP’s(s sap) popular HANA platform) and certain scientific workloads that require data delivery to keep up with processing speed. The faster applications can read and write data — and doing so from an in-memory cache or solid-state drives is much faster than doing so from hard drives — the sooner that processors can compute it.
And because the new instance is part of AWS’s Cluster Compute family, multiple instances are connected via a 10 GbE network for speedy server-to-server data transfer. In benchmark tests from a site called CloudHarmony in 2010, Cluster Compute instances far outperformed anything else on the market (GigaOM Pro subscription req’d) at the time. They’ve also been used to spin up clusters that can compete with traditional supercomputers in terms of sheer performance — reaching No. 102 on the lastest Top500 list with a peak speed of 354.1 teraflops.
Although, it should be noted, AWS isn’t the only game in town for users wanting this type of beefy core in order to handle their real-time data processing needs. Liquid Web’s Storm cloud service, for example, offers some high-memory, SSD-powered servers of its own at nearly $1.50 per hour less than what AWS charges (albeit with fewer cores and absent the 10 GbE backbone and list of features that comes along with the AWS platform).
Whatever the cloud, though, ever-higher-performing instances mean new classes of workloads and more business for cloud providers that offer them. Especially as big data and analytics applications pick up steam and move from batch to real-time, clouds that can handle demanding users are in a good position.
Feature image courtesy of Shutterstock user ssguy.