informa
/
News

Amazon Adds Supersize General-Purpose Servers

Amazon Web Service expanded its virtual server offerings again with M4s, a much larger version of its general-purpose M3s.
7 Data Center Disasters You'll Never See Coming
7 Data Center Disasters You'll Never See Coming
(Click image for larger view and slideshow.)

Amazon Web Services has added a new general-purpose virtual server, the M4 series, to its EC2 lineup. It's meant to be used as a standard or in-memory relational database server, a gaming server, or a large SAP application or Microsoft SharePoint server.

With its balance of CPU, memory, and networking resources, it could also be used as a caching server, said Matt Garman, VP of Amazon's Elastic Compute Cloud (EC2), in an announcement on June 11. That gives Amazon three levels of general-purpose computing virtual servers. The smallest are the T2s, followed by the M3s and the M4s.

Like their predecessor M3 instances, M4s run on Intel Xeon processors, but M4s require E5-2676 version 3 Haswell chips running at 2.4 GHz, while M3s use the prior year's (September 2013) E5-2670 version 2 Ivy Bridge chips, running at roughly the same speed. Both chips are etched with 22-nanometer circuits and both are double-threaded, meaning they can execute two processes simultaneously.

[Want to learn more about EC2 instances in the Amazon cloud? See Amazon Bulks Up Cloud Instances For Big Data.]

The smallest M4 virtual server, the m4.large, comes with two virtual CPUs and 8 GB of memory. That's roughly twice the size of the M3's entry-level, general-purpose offering, the m3.medium, which has a single virtual CPU and 3.75 GB of memory.

The next size up is the m4.xlarge, with four virtual CPUs and 16 GB of RAM. It's followed by the m4.2xlarge, with eight CPUs and 32 GB of RAM, and then the m4.4xlarge, with 16 virtual CPUs and 64 GB of RAM. The largest, the m4.10xlarge, is equipped with 40 virtual CPUs and 160 GB of RAM. In contrast, the largest M3 has eight virtual CPUs and 30 GB of RAM.

The M4s' pricing also follows a steep ascending curve. An on-demand m4.large costs 12.6 cents per hour; an xlarge, 25.2 cents/hour; a 2xlarge, 50.4 cents/hour; an 4xlarge, $1.008/hour; and a 10xlarge, $2.52/hour.

The smaller M3s, with their Ivy Bridge processors, range from the m3.large with one virtual CPU at 13.3 cents per hour up to the m3.2x large with eight virtual CPUs at 53.2 cents/hour.

The M4s, with their larger sizes, can be used for large, resource-intensive tasks. An enterprise caching server, for example, might stockpile large amounts of frequently accessed content in memory, and then rely on the multiple CPUs and Amazon's Enhanced Networking to serve that content out to thousands of website visitors. The networking between virtual CPUs and other resources works at 10 Gb per second, and "delivers up to four times the packet rate of instances without Enhanced Networking, while ensuring consistent latency, even when under high network I/O," the announcement said. That networking speed would also help ensure fast transactions and quick database reads and updates.

Amazon offers Placement Groups, where servers with the same or similar functions can be grouped in close proximity to each other within its EC2 infrastructure-as-a-service. "Within Placement Groups, Enhanced Networking reduces average latencies between instances by 50% or more," the announcement said.

The Amazon M4s join a lineup of virtual servers that is optimized for various tasks: compute-intensive applications, the C3s and C4s; memory-intensive virtual servers, the R3s; and storage-intensive instances, I2s and D2s. There are also virtual servers designed for graphics processing-intensive tasks, the G2s.

Rich Belanger, CIO at ProQuest, whose software aids research in government, corporate, and academic organizations, said in the announcement that his firm frequently employs Apache SOLR, an open source search engine, to optimize customer searches. ProQuest offers Summon Service for specialized library searches. "When running our SOLR workload to optimize search, we found that we could benefit from an instance that has a memory footprint right between what is offered on compute-optimized and memory-optimized instances ... With the new m4.4xlarge, we will be able to better optimize our resource usage to decrease our costs," he said.

The smallest M4 virtual server, the m4.large, comes with two virtual CPUs and 8 GB of memory. That's roughly twice the size of the M3's entry-level, general-purpose offering, the m3.medium, which has a single virtual CPU and 3.75 GB of memory.