Though the vast majority of enterprises would have lost nothing during Microsoft's Azure Services Platform beta outage last weekend, the event teaches IT managers a valuable lesson as they plan future cloud strategies.
IT shops need to have a disaster recovery plan for cloud services just as they would for their own data center. "There is no reason you won't need a disaster recovery site with cloud computing," said David Mickelson, the vice president and chief technical architect at Loomis Sayles & Co., a Boston-based financial services firm.
Azure, the cloud services platform introduced by Microsoft last fall at the company's Professional Developer's Conference, is an operating system and set of developer services. Still in trial, the technology went dark for 22 hours last weekend, leaving applications unreachable.
Mickelson does not use the service but said it's important for companies to have a primary and backup failover cloud vendor as well as an intermediary that allows them to switch back and forth. Other companies with cloud services, such as Amazon and Google, have also had major outages.
It's premature to know whether these early outages will have fallout in terms of lost IT manager confidence. The technology does not introduce new risks, but it does invalidate a lot of legacy remedies in the sense that IT experts are used to looking at application reliability and security in silos, said Tom Nolle, the president at CIMI Corp., an independent consulting firm.
"Cloud creates a variable and sometimes subliminal interdependence," Nolle said. "Anyone who has done failure analysis will tell you that calculating the mean time between failures is complicated by interdependency factors."
Cloud computing still has a long way to go in terms of enabling IT shops to manage resources and determine the operational status of their resources, as is possible with current client/server tools.
Many corporations are used to receiving services with some levels of quality assurance from their service providers, much those they receive from telecommunications carriers, Nolle said. As a best-effort service, the Internet offers no real service guarantees and no guarantees of customer service.
What should we expect from cloud computing? IT experts still aren't sure whether the cloud is an extension of the Internet or an extension of computing, "If it's an extension of computing, it may have to be more accountable, and if it's the Internet, well maybe not so much," he said. "The question is, 'Are we applying to the Internet-hosted applications a standard of availability we've never applied to the Internet before?'"
Certainly one call center hosting provider considers cloud services as a computing service. "The four or five nines [of uptime] is available in a telco -- unless your system is clustered or mirrored," said Marc Robinson, the vice president of IT at Centennial, Colo.-based Cloud 10 Corp. "People have to build that into their expectations." But Robinson said that even though telecommunications systems, such as a PBX, rarely go down, they tend to take longer to recover. Computing systems may go down more but downtime is minimized, he said.
Margie Semilof is the Executive Editor of SearchVirtualDesktop.com and SearchEnterpriseDesktop.com. Write to her at firstname.lastname@example.org.