Amazon Wins Cloud Storage Shootout, Microsoft Places Second
Amazon Web Services edged out 16 cloud storage providers in a 26-month stress test that measured scalability, availability, stability and performance.
The company's Simple Storage Service (S3) was one of only six that made the cut, with Microsoft's Windows Azure coming in second. The tests were conducted by Nasuni, a provider of premises-based network attached storage (NAS) gear that uses cloud storage providers for primary storage backups and/or disaster recovery.
While this benchmark is based on one vendor's assessment aligned with its own criteria and service-level requirements, it is the first I have come across that has measured cloud storage providers over a prolonged period of time and publicly disclosed its findings.
In addition to Amazon and Microsoft, AT&T, Nirvanix, Peer1 Hosting and Rackspace all passed Nasuni's stress test. The company declined to name the 10 that didn't make the cut, noting that as those providers mature their offerings, they stand a good chance of passing the tests in the future.
"The large providers certainly have a leg up with regard to economies of scale and tenure of performance," said Jennifer Sullivan, Nasuni's VP of marketing, in an e-mail. "We'll continue to monitor a variety of cloud providers, and as adoption of the cloud increases (e.g., different use cases for the use of cloud in organizations emerge), this will help shape what cloud storage has to become to be adopted and embraced by the enterprise."
Nasuni has maintained that the cloud is merely a component of an overall storage solution, particularly enterprises with distributed locations. Nasuni's on-premise storage controller, which leverages the cloud as a target for data, provides added security and access control.
When offering its solution, Nasuni chooses a cloud storage provider for a customer that will meet the service-level agreements at any given time. "We choose the cloud provider and we can also migrate providers if we feel that one provider offers better performance," Sullivan said, likening the process to computer makers that choose hard disk drives for customers. "We dedicate the cycles to evaluating the providers so our customers don't have to."
Here are some findings from the report:
- Writing large files: Windows Azure had the highest average speed at 2.38 MB per second. Nirvanix was close behind at 2.32 Mbps. The remainder of the six had similar speeds except for Peer1, whose average write speed was 1.49 Mbps.
- Reading large files: Nirvanix was fastest at 13.3 Mbps, with Windows Azure coming close behind at 13.2 Mbps. Amazon posted 11.28 Mbps.
- Writing medium-sized files: Windows Azure led at 2.1 Mbps, followed closely by Amazon S3 at 2.0 Mbps. The remainder came in 28 to 70 percent slower.
- Reading medium-sized files: Amazon significantly outpaced everyone else at 9.2 Mbps. Coming in second was Microsoft, though 28 percent slower at 6.6 Mbps.
- Reading small files: Amazon S3, at 387 files-per-second, was 41 percent faster than its nearest rival, AT&T.
- Writing objects: Windows Azure led with 154 files per second, with Amazon S3 coming in second at 135 files per second. AT&T came in third with 98 files per second. The remaining three were much slower.
- Outages: Amazon had the fewest, with only 1.4 per month, and the average duration was not significant, giving it an uptime of nearly 100 percent. (Those who experienced some of its major disruptions earlier this year, including April's four-day outage, may beg to differ.) Microsoft had 11.1 outages per month with an overall uptime of 99.9 percent. Peer1 had 6.8 outages, Rackspace experienced 10.3, AT&T averaged 10.4 (though posted uptime of 99.5 percent) and Nirvanix was less fortunate with 332, though the outages apparently were not significant since its uptime still came in at 99.8 percent.
Sullivan said it will be interesting to see if Amazon holds the top spot, noting Microsoft has a good chance of taking the lead. "Time will tell," she noted. A copy of the report is available for download here.
Posted by Jeffrey Schwartz on December 13, 2011 at 11:59 AM