Cloud Performance Testing Gives Slight Edge to Azure over AWS
Enterprises seeking a cloud services provider might want to choose either Amazon Web Services (AWS) or Microsoft Azure, according to a new report from Nasuni.
In fact, Azure just edged market leader AWS in recent performance testing, according to the third biennial cloud storage report from cloud storage vendor Nasuni.
Google, the only other cloud provider with a large enough global scale that could be compared with the two by Nasuni's standards, came in a distant third place.
It's important to keep in mind that this is one benchmark by a single provider with its own requirements -- primarily using a large global cloud provider as a NAS target. But Nasuni performs the tests to determine which services it should use to provide as a storage target and claims it's not wedded to any one player, unless a customer specifies one. Nasuni first began sharing its benchmarks in 2012 when AWS had an overwhelming edge, though that was before Microsoft had a mature infrastructure as a service available.
Today, depending on the service, Nasuni primarily distributes its workloads between AWS and Azure and is always willing to add or shift to other suppliers. Nasuni currently prefers Microsoft Azure Blob Storage and Message Queue though it uses AWS's Dynamo database and EC2 compute instances, said John Capello, Nasuni's VP of product strategy. The primary test Nasuni conducted between October and February for the report evaluated a variety of read-write and delete scenarios, according to Capello.
"For our purposes, which is to write files for mid-sized to large enterprises to the cloud, Microsoft Azure Blob storage is a better target for us than Amazon or Google," he said. "Amazon is a very, very close second. Amazon and Microsoft seem to be, as many others have said, the real two competitors in this space in providing cloud services in general, but specifically with storage, they're very, very close in terms of both their speed, availability and their scalability."
According to the report, which is available for download if you're willing to give up your contact information, is that Microsoft outpaced Amazon and Google when it comes to writing data to a target 13 of the 23 scenarios of varying thread counts or file counts. When it came to reading files, Microsoft constantly performed better, though not to the extent it did in the write tests. Microsoft was twice as fast as Amazon when it came to deleting files and five times as fast as Google.
For system availability, Amazon's average response time of 0.1 seconds slightly edged Microsoft's 0.14 seconds, while Google was roughly five times slower. Nasuni also measured scalability and when writing 100 million objects to look at the number of read and write misses, "Microsoft had, by far, the largest write variance, which was more than 130 times larger than Google's, who had the smallest variance." Read and write errors were almost non-existent, according to a summary of the report. "Only Amazon showed any misses at all: five write errors over 100 million objects, which gives an error rate of .00005 percent."
Nasuni omitted several key players from the test, notably IBM's Softlayer, which was undergoing system upgrades and led to frequent periods of planned downtime during the testing period, according to Capello. HP was also initially in the test, though Capello said Nasuni chose to leave the company out this time because of HP's announced plans of changes in cloud strategy. "Before we decided we weren't going to continue testing them, they actually did surprisingly well, in some cases -- better than Amazon and Microsoft in some of the read-write and delete benchmarks," he said. "If we had run the full test, it would be interesting to see where they came out. "
Jeffrey Schwartz is editor of Redmond magazine and also covers cloud computing for Virtualization Review's Cloud Report. In addition, he writes the Channeling the Cloud column for Redmond Channel Partner. Follow him on Twitter @JeffreySchwartz.