Amazon High I/O Option Tackles Tough Cloud Workloads
High I/O Instance virtual server relies on solid-state memory for the kind of high throughput needed by online game suppliers, medical researchers, and intensive database users.
Amazon Web Services has introduced the High I/O Instance, a new virtual server type, into its widening pantheon of server instances. It's designed to yield high throughput for a workload on the EC2 compute cloud.
The high I/O server is designed to take advantage of and is backed by 2 TB of low-latency flash storage close to or "local" to the server CPU. The flash storage is capable of delivering 120,000 read input/output operations per second. The storage is also capable of executing 80,000 write operations per second. The flash memory is visible as two 1 TB volumes.
"AWS customers are bringing their most demanding workloads onto the cloud ... workloads that require dedicated and high performance I/O for which we are now introducing the High I/O Quadruple Extra Large" virtual server, wrote CTO Werner Vogels Thursday in his All Things Distributed blog.
A quadruple extra-large server has eight virtual CPUs with 3.5 EC2 compute units each. That's awkward Amazon parlance for saying each virtual CPU that has been assigned processor cycles equivalent to 3.5 Xeons of the 1-GHz, 2007 vintage. In effect, the quadruple extra-large server has the equivalent of 35 Xeon cores running at 1-GHz clock speeds.
The virtual server also has access to 60.5 GB of RAM.
Those dimensions appear to appeal to some potential new users of EC2. A researcher who does genome mapping said in the Hacker News forum that his project's "indexes won't entirely fit in memory" of his university cluster.
"It would be handy to spin up a few of these instances, load the indexes from an EBS volume [Amazon's Elastic Block Store service] onto the local solid-state disks, then run for a couple of hours or so. This is a very I/O intensive job that we need to run about once a week," he wrote.
Amazon is charging $3.10 an hour for the I/O intensive instance from its northern Virginia data center. If run continuously over the course of a month, the bill might top $2,232. But many I/O jobs, like the genome researcher's, need to run for only a limited amount of time. "We've toyed with the idea of adding SSDs to our in-house cluster, but couldn't quite justify the costs. This might actually ... get our lab to migrate to EC2 as opposed to our university cluster," he wrote.
The high I/O instances could also be used by enterprises for periods of intense database transactions or by social gaming providers to handle hundreds of thousands of gaming interactions.
Adrian Cockcroft, director of architecture at Netflix, a major EC2 user, said his firm streams films and video to customers on EC2 using the NoSQL system, Cassandra. It already uses high-memory servers. The prospect of running the same system on high I/O solid-state servers appeals to Netflix. "The ability to leverage SSD-backed instances, which provide tens of thousands of I/Os per second, will significantly increase the performance of our Cassandra clusters and overall service capability," he said in Amazon's announcement Thursday.
High I/O instances are available in Amazon's northern Virginia data center (US-East) and in its Dublin, Ireland data center. If activated in Dublin, they cost $3.41 an hour. AWS spokesmen say high I/O instances will become available in other regions later this year.
Multicloud Infrastructure & Application ManagementEnterprise cloud adoption has evolved to the point where hybrid public/private cloud designs and use of multiple providers is common. Who among us has mastered provisioning resources in different clouds; allocating the right resources to each application; assigning applications to the "best" cloud provider based on performance or reliability requirements.
Join InformationWeek’s Lorna Garey and Mike Healey, president of Yeoman Technology Group, an engineering and research firm focused on maximizing technology investments, to discuss the right way to go digital.