Forrester's James Staten discusses some of the hurdles that enterprise users face in getting to cloud computing.
“All clouds are not created equal,” said Forrester’s James Staten in a talk in San Jose last week. Enterprise buyers will need to be aware of what their monitoring, security and service-level agreement needs are before they’ll be able to evaluate different services.
For good or for ill, cloud computing is not one thing. If you strip away the pretenders and re-labelers -- those who have just cloud-washed their product and given it a new name -- the field is still a bewildering array of suppliers with long lists of services and prices.
But the term "cloud" has gained currency because many people see a new distribution model of computing power emerging, offering improved economics for the end user. I should say "the possibility of improved economics," because it's not been proven in studies that cloud computing is any cheaper than private data center operation. I believe it will be, but the economic argument may apply best to certain applications, not across the board.
Forrester analyst James Staten expounded on these points from his own point of view at a talk hosted by the managed service provider Savvis at the San Jose Marriott on Oct. 14. Savvis operates 31 data centers for enterprise IT outsourcing customers around the world, with 18 equipped to offer its Symphony private cloud services.
Staten pointed out a key difference between your workload in the cloud versus the private data center. An Oracle or IBM application server, running in the data center, offers a 30-millisecond response time 99.999% of the time, he noted. If you send your workload to the cloud, chances are it’s going to run as a virtual machine in a data center “somewhere on the West Coast,” and that guaranteed minimal latency will give way to a longer and more variable wait time in application operations. At best, the cloud will give you a 99.95% guarantee of service at a higher latency, and that means your system must be able to tolerate up to 15 minutes of downtime a year, he said.
“There is no guarantee of quality of service in the cloud. If your application doesn’t perform well, you have no recourse,” he said. In addition, your application and its data may be located on the same server or stored in the same disk as your competitors. You need to accept that possibility to proceed with cloud computing on a plain vanilla infrastructure as a service provider, Staten said.
To me, one of the main differences between the cloud and enterprise data center is how system reliability, or high availability, gets kicked upstairs into software. Amazon, EBay, Google and Yahoo have all figured out how to build big clusters of x86 servers, and then run software on them with a tolerance for hardware failure. The workload you send off to the cloud will have no such tolerance for hardware failure unless you have architected it into it. Internally, most enterprises still operate hardware systems that contain redundant components or try to warn of a failure before it occurs. Reliability still lies on the hardware side.
SaaS As Innovation Driver?Software as a service is the clear No. 1 way enterprises consume cloud. InformationWeek's SaaS Innovation Survey reveals three tips to get the most from SaaS: Make it a popularity contest. Have an escape plan. And remember that identity is the new perimeter.