We are in the process of discussing the issue of how does one optimize WAN performance in a cloud-based world in a Thought Leadership Discussion at Webtorials. As a part of that discussion, we've been chatting with Thierry Grenot, CTO of Ipanema Networks. (Ipanema offers a suite of products for optimizing the use of network facilities.)
The gist of the discussion centers around the simple fact that while cloud-based AaaS (Anything as a Service) can bring great economies of scale and open up applications and services that might not otherwise be available, the very fact that they are "in the cloud" means that there is also an inherent loss of control.
This loss of control has happened in a gradual progression over the past 20 or so years. First, we moved from "dedicated transmission services" such as private lines, T-1/E-1, etc. to packet-based services such ranging from frame relay to MPLS. The pay-off was simple. By sharing bandwidth dynamically, there was an immediate financial upside, and the network could be engineered to guarantee a certain expected level of service.
We then moved in many cases to Internet-based services. Again, the savings were sufficient that we were willing to sacrifice deterministic response for what was statistically "pretty good."
But now we have yet another variable. Any time that we are using a cloud-based service, the odds are pretty good (for now) that the access to those public cloud services will be via an Internet connection. So the actual performance now has two significant "unknowns" - the performance of the network AND the performance of the service itself. Consequently, having an SLA that can be taken seriously is almost impossible.
We see this as an imperative to be addressed, and that is the focus of the Thought Leadership Discussion. We cordially invite you to explore these ideas with us and to participate in the discussion.
Sign up for MIS Asia eNewsletters.