The highlight of Google's coming-out party for Compute Engine's general availability as supported infrastructure was the pricing on its virtual servers and storage. It shows a heretofore absent determination on Google's part to compete with the market leader, Amazon Web Services, and secondarily, Microsoft.
Amazon has made it difficult for competitors to attract customers by repeatedly lowering prices on commonly used virtual servers and storage. Microsoft has matched its moves, particularly on storage.
With its announcement this week, Google has come close to matching AWS's virtual server prices while, in a clever move, it has flipped prices around on storage to undercut Amazon. It cut the price for storage attached to working virtual machines by 60%, making it cheaper than Amazon's Elastic Block Store. It made up for that big cut by leaving prices on its long term storage -- the equivalent of Amazon's S3 -- at a little more expensive level. It has chosen the sometimes finicky Amazon Elastic Block Store as the Achilles' heel of the market leader, which it might or might not be.
For example, both Amazon and Google were charging 10 cents per GB per month for Elastic Block Store, or what Google calls Provisioned Storage under its Persistent Storage products. Either one is the disk volume that's attached to a running virtual machine, as opposed to the temporary disk assigned to it on the server rack. VMs are automatically assigned a chunk of temporary storage when they're commissioned, but it, like the user's data, goes away as soon as an instance is shut down. To preserve the configuration of the instance and its data, an Amazon user needs to sign up for EBS. When working with Compute Engine, the user needs what Google calls "provisioned space" on its Persistent Disk service.
[ Want more on cloud pricing? Read Why Cloud Pricing Comparisons Are So Hard. ]
Google cut the price of provisioned space from 10 cents per GB a month to four cents per GB a month. Cloud storage services have frequently been the focus of price cutting in the past. But to put Google's unprecedented 60% cut into perspective, Amazon reduced S3 long-term prices in March 2012 from 12.5 cents to 11.5 cents for the first GB, a drop of 8%. Google followed suit within two days, cutting its long-term storage from 13 cents to 12 cents, a drop of 7.7%. On Nov. 26, 2012, Google dropped storage prices 20% as Amazon opened its first Re:Invent user group show in Las Vegas. Two days later Amazon announced it would implement a 25% cut. Microsoft matched the cuts at the end of the month with a 28% cut on Azure cloud storage.
Now Google has dropped the price of block storage by 60%, going from 10 cents per GB per month to four cents. At the end of two days, Amazon and Microsoft remained silent -- one is tempted to say it has maintained a stunned silence.
Long-term storage, Amazon S3 versus Google Cloud Storage, is a different matter. Amazon charged 9.5 cents per GB a month for snapshot storage or the storage of a file containing many data objects. Google charges 12.5 cents per GB a month for snapshot storage, although the difference is less than the 3-cent price differential would seem to indicate. To get a guaranteed level of disk I/O, Amazon increases the price to 12.5 cents. Google appears to guarantee user-desired I/O levels at no additional charge.
Developers always looking for the lowest-cost infrastructure are sure to take notice. IT managers with responsibility for outsourcing enterprise workloads are likely to consider their options as well. The result of Google's cuts might eventually be a changed competitive lineup in cloud services. Google's 60% reduction in block-storage cost has the direct effect of bringing down the total cost of running a workload.
Amazon might have more vulnerability in block storage than generally perceived. One of the most frequent complaints about Amazon's operation is the sometimes-unpredictable nature of its EBS service. Virtual machines that normally run fine sometimes appear to slow down and take longer to complete a task, with EBS being the most likely wild card in Amazon's lineup of services. The "noisy neighbor" problem is often a reflection of waiting for data from EBS. This happens when one customer's busy virtual machine on a shared host uses up I/O channels and puts CPUs on idle, slowing output for another customer.
Google also heightened the competition on another front with its 10% cut in virtual-server pricing. It listed no examples, but working from Wednesday's posted prices, a Google n1-standard server with a single virtual CPU and 3.75 GB of RAM is available at 10.4 cents an hour, down from 11.5 cents an hour. A roughly equivalent Amazon m1.medium server with one virtual CPU and 3.75 GB of RAM goes for 12 cents an hour. Direct comparisons, however, are difficult. The virtual CPU on the Amazon machine is the equivalent to two 2007 Xeon cores running at 1 GHz. The Google virtual CPU is half of a double-threaded, modern Xeon core running at a higher clock speed.
Storage is the more easily compared service. There Google has chosen to strike in its bid to be taken seriously as a supplier of infrastructure-as-a-service.
Buying power and influence are rapidly shifting to service providers. Where does that leave enterprise IT? Not at the cutting edge, that's for sure: Only 19% are increasing both the number and capability of servers, budgets are level or down for 60%, and just 12% are using new micro technology. Find out more in the 2014 State Of Server Technology report. (Free registration required.)