The Amazon Web Services outage Monday underscores for solution providers the fact that the cloud is not bulletproof when it comes to service availability, and there needs to be a backup plan when working with cloud providers like Amazon.
"One of the biggest hurdles we have to overcome is educating our clients that redundancy is not built into the cloud and that there is a cost associated with being redundant," Kevin Chu, director of systems and infrastructure with digital marketing and technology company Digitaria, said in an email.
San Diego, Calif.-based Digitaria's clients were largely unaffected by the outage because Digitaria had service contracts for them with Amazon in two service areas. A database issue is still unresolved, but Digitaria is working around it, he said.
[Related: Top 10 Issues Eroding Cloud Confidence]
Tony Safoian, president and CEO of solution provider SADA Systems, of North Hollywood, Calif., also urges a cloud backup strategy. "The cloud is not a magical place where nothing bad happens," he said. "People are forgetting they need to plan for failover and redundancy."
Amazon Web Services went down in its Northern Virginia data center complex, lasting most of the day and causing website outages in an unknown number of companies, including Reddit, Pinterest and Airnb.
Amazon's service dashboard during the day showed service interruptions with its Elastic Beanstock services, Management Console for Elastic Beanstock Services, Relational Database Service, ElasticCache, Elastic Compute Cloud and CloudSearch.
But, Amazon pointed out Monday that just one service availability zone was affected.
Amazon experienced an outage on June 14, 2012, and another on April 21, 2012.
Chu said Amazon's service-level agreements (SLAs) tell users that they will reimburse solution providers for the cost of an outrage only if two or more service availability zones are affected. The message is clear that solution providers should contract for service in at least two areas, Chu said.
Chu said not all organizations need to be placed in two availability zones. "If a customer is a non-profit local provider in one city, it doesn't make sense to go with two contracts," Chu said. But if an organization stands to lose money or business with downtime, a second contract makes sense.
Chu said he still values Amazon as a cloud partner. "Being an AWS Advanced Consulting Partner + Channel Partner (aka Reseller) gives us access to enterprise account managers that help us ensure that our clients get top-notch service and support," he said.
Safoian, whose business provides services for Google Apps and Microsoft Office 365 and is a member of the Google Cloud partner program, said in some cases having one cloud provider is not enough to guarantee service availability.
He's planning to develop an automatic failover program between Amazon Web Service and Google Compute.
"Applications should be designed for multiple clouds or at least multiple services locations with the same provider," he said.
PUBLISHED OCT. 25, 2012
