Mary-Jane Foley has a very good column on Microsoft which I recommend. Do have a look at this latest article on ZDNet about BPOS cloud customers hit by multi-day email outage.
Some statements just stop me:
"Users are reporting there have been intermittent, multi-hour Exchange Online outages on May 10, 11 and 12 at different times in North America, Europe and Asia."
“There have been various outages over last 48 hours. Not a good situation,” said one Microsoft partner with whom I spoke. “Microsoft does not have a formal response ….if they had isolated the problem they would have issued a response. When this happened previously they responded quickly with the cause and SLA (service level agreement) remedy.”
"The Microsoft spokesperson said the team also declined to say whether the Exchange Online problems were related in any way to Microsoft moving from BPOS to Office 365,the successor to BPOS. The Office 365 launch has been rumored to be set for early June 2011."
It would not be appropriate to bash Redmond's based clouds by position to Armonk's ones or Amazon's one or any of those operated by professional system integrators. It seems however that there is every month a major cloud issue which is so bad that it reaches mainstream media.
There are however two important general points to make:
1) Clouds will only be accepted if those who run them agree on user-centric service level agreements. The discussions that we have with vendor-based clouds still tend to indicate that running data centers with appropriate measurements is not part of their core competencies although professional outsourcers and system integrators have done this daily for years (or at least those we work with).
2) Frameworks, may they come from Microsoft, IBM, HP or Open Source community have limited notion of end-user experiences. While such SW tools are essential for Infrastructure Management, their usage limits the understanding of available metrics to objectively quantify quality of services. These softwares measure quality of service by IT metrics which are of little value by users like CPU availability instead of Application availability or percentage of mail routed below agreed timeframes etc.
Cloud will only take off when these notions are clear to CIOs and vendors alike. Things are improving far too slowly and the credibility of the concept of cloud could be seriously in danger.