Amazon Web Services, seen by many as the market-leading infrastructure cloud computing provider, has a pretty full shelf of virtual machine (VM) image sizes for customers to spin up in its cloud -- 17 separate instance VM sizes are listed on the company's website, in fact.
But a growing trend among some smaller cloud providers allows customers to create their own virtual machine sizes, specified by however much RAM, CPU and memory they want. In an increasingly busy market, vendors, particularly smaller ones, are looking to differentiate themselves from the mega-players -- like the Amazons, Googles and HPs of the world -- and this is one way to do it.
[ CLOUD SHOWDOWN: Amazon vs. Rackspace (OpenStack) vs. Microsoft vs. Google
NEWS: What's going on with VMware? ]
"You're pretty much talking about the biggest advantage we bring to the table," says Michael Higgins, chief enterprise solutions officer for European-based IaaS cloud provider CloudSigma, which also offers its services in the U.S. "If you are a high CPU user you won't have to purchase bundled RAM which you don't need!" the CloudSigma site explains. "If you need a lot of storage but less CPU and RAM likewise no problem. There are no 'standard' server size on the CloudSigma cloud."
CloudSigma isn't alone. ProfitBricks is another DIY-type cloud that allows customers to set their own VM instance sizes -- up to 48 cores per server, 196GB of RAM and 16 virtual drives with up to 16TB each of storage. ProfitBricks makes a big deal out if its use of InfiniBand, which it says allows for improved performance. It also offers per-minute billing and allows "live scaling," or the ability to change VM image sizes without having to reboot as well. "Why force customers to stumble through some long list of choices with meaningless names they then have to associate with some amount of capacity, none of which may meet their needs without waste?" says Pete Johnson, ProfitBricks' evangelist, who recently joined the company after a 19-year career at HP.
Gartner analyst Lydia Leong, who tracks the IaaS market, says customizable VM sizes is somewhat of a trend in the market now -- many VMware vCloud Powered public cloud providers, for example, have the feature. "Allowing arbitrary VM sizes allows customers to size VMs that fit the needs of their workloads as exactly as possible," she wrote in an emailed response to a question.
There could be some downsides to using this model, though, she warns. "Customers typically feel that this is the most cost-efficient approach, although whether or not it turns out to be any cheaper, in the end, is a function of the VM variants offered by a given provider, actual VM performance, and the cost of those VMs," Leong notes.
Price comparisons between the providers are not apples-to-apples. CloudSigma, for example, offers a recommended "small" instance size with 1.1GHz of CPU and 1.7GB of memory for $59 a month, according to its quote calculator. Amazon's default "small" instance type with 1.7GB of memory is $0.06 per hour for a Linux VM.
Customers also need to figure out exactly what customized VM size they want or need. Higgins, from CloudSigma, says for many test and development customers, this isn't a problem -- they're technical enough to know what they need and spin it up themselves in CloudSigma's website. Higgins runs a division within the company that works with customers to customize their VMs.
There is also a growing ecosystem of third-party vendors that will help customers optimize their cloud usage. Many of these vendors focus on specific public cloud service providers -- Newvem and Cloudyn, for example, help customers optimize their Amazon resources. But Cloudability is a similar tool that works across more than a dozen providers.