Amazon.com should proactively offer relief to customers impacted by its outage over Easter, regardless of whether it is compelled to do so by the terms of service level agreements.
For the last week, the company has gone to ground following a major outage that took out its entire cloud computing service on Thursday night, with some customers hosted in its Virginia, US data centre still suffering today.
The only information available to customers has been via interpretations of Amazon’s service status page – no official response or apology has been offered.
In the past 24 hours I have attempted to contact both Amazon’s communications team and CTO Werner Vogels (pictured). Neither has responded.
Critically, customers are seeking a post-mortem more detailed than the information provided on the service status page to date. Amazon said this would be forthcoming once all customers are back online.
In its latest update, the company conceded that 0.07 percent of the storage volumes attached to the virtual servers it hosts will “never be recoverable”.
This will have a major impact on how the business community percieves the public cloud.
For Amazon, it poses a more immediate problem. It will be a difficult and delicate task to appease these customers, some of whom may choose to pursue a remedy by testing Amazon’s contract in courts of law.
Any such case would test whether cloud computing contracts are viewed as fair and reasonable.
As noted by Truman Hoyle partner Mark Vincent at a recent iTnews event on cloud contracts, Amazon’s standard terms state that it “will not be liable to you for any direct, indirect, incidental, special, consequential or exemplary damages”.
For the remainder of the thousands of Amazon customers impacted, the pressing question is whether the company will live up to its service level agreement (SLA) and offer service credits.
Amazon offers a 99.95 percent guarantee of uptime on its EC2 service – which means the cloud computing provider must pay out service credits for any more than 4.38 hours of downtime per year.
Gartner analyst Lydia Leong has speculated that the company might avoid paying up for the outage, according to her interpretation of the fine print in Amazon’s contract.
“Nobody’s going to be able to claim SLA credits,” she said, as Amazon’s SLAs are “narrow and specific”.
Technically, she said, Amazon’s virtual servers (the EC2 service) subject to the SLA did not fail. Rather it was the EBS (data storage service) and RDS (relational database service) attached to these instances that went down. Amazon doesn’t offer an SLA for EBS or RDS, Leong said.
But practically, the EC2 service was useless to customers during the outage, since many services hosted on EC2 absolutely require the attached storage or database services to operate.
Further, it has been argued by some observers that Amazon’s service level agreement is only applicable to customers with applications straddling multiple ‘availability zones’. Amazon could perhaps argue that all of its availability zones were only down for a few hours, after which time (for several days) the problem was limited to a single availability zone.
Again, this is a technicality. Practically speaking, customer services were unavailable for far longer than a couple of hours.
The question is, will Amazon use this fine print to weasel out of its responsibilities to customers?
“This is a test of any service provider’s ethical mettle,” commented Lorenzo Modesto, COO of enterprise hosting provider Bulletproof Networks, itself an Amazon customer – albeit unaffected by the outage.
He said that Bulletproof, which promises a higher level of service to enterprise customers than public cloud computes, would naturally feel compelled to issue service credits on the basis of “impact on the customer” rather than small technicalities, for fear of customers seeking an exit after an outage.
While Amazon doesn’t claim to be an enterprise-level provider, “it would be commercially motivated” to take a customers impact into account, he said.
Amazon will need to act decisively. Rather than wait for customers to send in forms and server request logs seeking service credits, the company should make a statement that helps it appear excessively generous in the face of this outage.
AWS may have been first to market, but it’s not the only cloud computing provider out there.