It’s difficult to avoid the subject of the cloud and its anticipated impact on email, unified communications, mobility, storage, CRM or any of a variety of other services that could be delivered via a hosted / SaaS / cloud model.  The benefits of the cloud are varied, but generally focus on lower total cost of ownership, more flexible service delivery, greater employee productivity and more efficient use of in-house IT staff.

Could anything convince decision makers that cloud computing might not be the best model for delivering the services they need, or at least limit the number of cloud vendors that they consider? While much has been made of the recent T-Mobile Sidekick debacle in which service was down for an extended period and substantial amounts of customer data was lost, these types of outages and data loss scenarios are relatively rare and, in my opinion, will have relatively little long-term impact on decisions to adopt the cloud model. Instead, I believe that what might kill the cloud, or at least slow adoption of the cloud model, will be day-to-day, sub-par service delivery.

Imagine a scenario in which your email is hosted by a cloud provider.  You log onto the Web portal to access your email and it takes five seconds for the login page to appear.  You type in your username and password, click Submit, and it takes another 30 seconds until you see your inbox.  Click on a message to read it and it takes another 15 seconds before the message opens. Click reply – wait another five seconds – type your message – wait another 10 seconds.  And so on.  Now, imagine that you just arrived at the airport and you have a short layover, and that you have 15 or 20 messages that you need to read and to which you need to reply before your next flight.  For those whose patience is more limited than they might like, waiting 30 seconds or more to process each message in your cloud-based email system might be less than tolerable.  If you’re a senior executive who makes decisions about such things, you might not be pleased.

This problem is by no means theoretical. Amazon.com, for example, has determined that every 100-millisecond delay in processing Web page views on its site costs 1% in revenue.  While Amazon’s is clearly a different model than cloud-based email, I believe the principal applies to virtually every cloud-based service there is or will be: the longer it takes to process page views, user clicks and the like, the less amenable decision makers will be use to cloud computing, or at least to use providers whose performance does not meet their expectations.

So, what do you do about it?  If you’re a potential customer of a cloud-based service provider, do your homework and ask for a track record of performance. Take advantage of trials so that you can test performance during real-world scenarios.  Evaluate the provider’s infrastructure and see if they operate multiple data centers so that your users can access one that is geographically close so as to improve performance.  If you’re a cloud-based provider, deploy a sufficiently robust infrastructure so that performance will be acceptable during times of peak demand. Test your infrastructure using a service like SOASTA’s CloudTest to determine if your service levels will be acceptable. Talk to users and get their impressions of your service over time.

Cloud computing offers a number of important advantages, whether the cloud is public or private.  However, given that it is replacing largely local/on-premises capabilities that today offer reasonably snappy performance, cloud services will have to compete on performance as well as their other benefits.

OR Commentary for Messaging Wire

Week of October 26, 2009