A comment today by Google’s top cloud executive, taking a subtle jab at Amazon Web Services, isn’t sitting well with Microsoft for a different reason.
With last week’s Amazon Web Services outage still fresh in the minds of many people in business and technology, Google senior vice president Diane Greene took a few moments at the Google Cloud Next conference this morning to address the issue of reliability.
“I’m really proud — I just learned yesterday that we were recognized as having the highest availability of any cloud over the course of 2016,” Greene said on stage. She paused before adding with a chuckle, “I think 2017 will be promising, too.”
Greene didn’t cite a source, but a Google representative confirmed afterward that she was referring to new numbers from CloudHarmony, a unit of the Gartner research firm. Tech news site The Information reported on the CloudHarmony numbers in a story Tuesday, with a chart showing that Google Cloud had 74 minutes of “total time lost” in 2016, compared with 270 minutes for Microsoft Azure and 108 for Amazon Web Services.
Here’s the issue Microsoft is raising: Comparing downtime alone doesn’t take into account the larger number of regions operated by Azure, which Microsoft says would provide a more accurate picture of cloud reliability. Although each company defines its regions differently, Microsoft says it has 34 regions, AWS says it has 16 regions, and Google Cloud Platform says it has six.
Microsoft issued this statement: “Microsoft has 34 Azure regions online worldwide, more than any other provider. When looking at average uptime across regions, rather than total downtime across a disproportionate amount of regions for each provider, Azure reliability is in line with that of the other cloud providers measured and in fact has consistently had global uptime upwards of 99.9979% for Compute in the past 12 months alone. What we hear from our customers is that uptime is a more useful measure of availability.”
The report by The Information’s Kevin McLaughlin and Mike Sullivan did note, “To be sure, Microsoft has been expanding its cloud business by opening new regions with multiple data centers in the past year. That may have increased its potential for outages, particularly if newer facilities have kinks to be worked out.”
Jason Read, the Gartner research vice president who founded CloudHarmony, agreed that basic comparisons of total downtime can be misleading, because a single extended outage can skew the comparisons; some outages have a bigger impact than others; and more regions and data centers create more potential for outages.
“I think a more appropriate method of comparing cross vendor availability at a high level is average downtime per region – i.e. total downtime/# regions,” Read said via email. “Additionally, I think it is useful to break out US and EU regions where a majority of their users would be impacted.
Amazon hasn’t commented publicly on the issue. A Google representative pointed to an additional CloudHarmony statistic showing that Google’s cloud storage service, specifically, had the least amount of downtime since the beginning of 2015 — just four minutes in that period. That’s relevant because Amazon’s outage last week, lasting a few hours, involved its Simple Storage Service (S3).
CloudHarmony’s Read confirmed, “Google Cloud Storage has historically been very reliable based on our availability checks with only a few small outages in regional buckets in 2015.”
RELATED: Google Cloud touts new customers, reliability and machine learning — says it’s ‘here for real’
“Google Compute Engine on the other hand, has experienced a fair number of outages,” he wrote. “It was the least available compute platform in both 2015 and 2016, and also experienced the only global outage I’ve observed for any cloud vendor on April 11 2016 due to a bad network configuration rollout. This type of outage is in the absolute worst case scenario because even multi-region fault tolerance would not have mitigated it.”
He added, “Almost all downtime in the past 90 days for both Azure and Amazon is due to the recent multi-hour outages in Azure/West US 2 and Amazon/us-east-1. Both services were highly reliable in 2016 – 0 downtime for Amazon in US/EU regions and 315 seconds per region average for Azure.”
See Greene’s comments and the rest of the Google Cloud Next Day 1 keynote in the archived live stream below.
