WAN Optimization Survey Results: Visibility into optimized WANs, “Important” or “Very Important,” Say Nine in Ten.

During a recent NetQoS – Cisco WAN Optimization Seminar Series, and to some extent at Interop, NetQoS launched a survey regarding WAN Optimization. We found some interesting things:
Over 90% of the survey respondents said that it is either important or very important to be able to quantify accurately the results of WAN optimization.
Over 80% say it is important/very important to have integrated WAN optimization and performance reporting.
We also found that most respondents – 60% – believe that the most relevant measure of WAN optimization impact is in application and network latency. 30% believe that link utilization is the most relevant measure, and 10% believe that the two are equally important, with a handful of respondents saying that protocol distribution is the most important.
What this tells us is that many enterprise IT organizations do not want to deploy a WAN Optimization solution that is going to break their application and network performance monitoring. However, WAN Optimization devices perform local TCP ACKnowledgements, thereby confusing performance monitoring and making it near impossible to accurately quantify the results of WAN Optimization.
We go into this challenge more in “WAN Optimization’s Dirty Little Secret” and John Mao and Ben Erwin have uploaded some videos in Network Performance Daily’s Whiteboard Series to illustrate the points, but basically, many WAN Optimization devices (WOD) split a TCP/IP session between a client and server into three separate sessions – a client segment, a WAN segment (or “channel”), and a server segment. But data center-resident passive network monitoring tools assume only one TCP session between the client and server – which means that when WAN optimization is deployed, they only have visibility into the server segment response time – between the server and the data center WOD. So, visibility into all the components that make up end-user application response times, including network round trip time and data retransmission time, is lost.
If ultimately, IT is about application delivery – the reliability and speed with which you can deliver applications to the end-users – then forgoing application and network monitoring is not an option. We’ll have more on this problem next week. In the meantime we look forward to seeing you at Cisco Networkers in Anaheim.
————–
More information:
On Quantifying the results of WAN Optimization