Is Your Network WAN Optimization Really Helping?

David Swenson works at Infineta Systems, a provider of WAN optimization systems for Big Traffic, where he is responsible for technical marketing research and communications.

david-swenson

When considering WAN optimization solutions, the two most common metrics for measuring effectiveness are the data reduction rate, which looks at data compression on the network, and throughput, which looks the rate of successive packet delivery. Both metrics are excellent for point-in-time analysis, and they provide clear numbers that are easy to understand and compare.

WAN Optimization’s Compound Effect

There is a third metric, however, that is often overlooked. It is the compound effect, a measure of the combined effects of both reduction and acceleration. The ideal vantage point for observing the compound effect is at the endpoints generating the WAN traffic being optimized. Given both reduction and improved throughput, a solution’s real value is its impact on endpoint performance and whether it helps the organization achieve the lowest possible RTOs (Recovery Time Objectives) and RPOs (Recovery Point Objectives) , and thus meet or improve their BCDR (Business Continuity and Disaster Recovery) goals.

RTO-and-RPO

Image Courtesy of Infineta

Un-confounding Compounding

So, what exactly is the compound effect? Take the example of SRDF replication between two data centers connected by a 1Gbps MPLS WAN with 80ms RTT. Burgeoning storage demands and unyielding RPOs can turn the inter-data center WAN into a bottleneck. After deploying a good WAN optimization solution, one could expect to see an average of 5x or more data reduction for high-speed replication flows. At the same time, TCP acceleration could improve throughput by about 4x, for a total performance improvement of 20x. Due to the combined benefits of reduction and acceleration, applications that were running at 250Mbps would instead run at 5Gbps over the same 1Gbps link. In other words, not only is the link now fully utilized and thus carrying four times as much traffic, the traffic it is carrying has been condensed by a factor of five. When delivered to the endpoint on the other side of the WAN, the condensed traffic will be restored and the endpoint will receive it at the rate of 5Gbps.

WAN-Optimization

Before WAN optimization, 10 SRDF connections running on a 1Gbps MPLS WAN link with an 80ms RTT were able to achieve a throughput of 25Mbps per connection. After WAN optimization, throughput increases to 100Mbps per connection, i.e., link capacity. At the same time, 5x reduction decreases the size of the TCP packets, which frees up bandwidth that is then consumed by the connections until the link is filled with reduced packets. Whereas the endpoints were running at 250Mbps before optimization, they are able to run at 5Gbps after optimization - a sustained 20x performance increase. Graphic courtesy of Infineta.

Surplus Bandwidth

Now let’s consider a scenario where the WAN is not a bottleneck that is throttling replication performance. In this case, the compound effect is expressed solely through reduced bandwidth consumption. Using the example above, you could expect to see the bandwidth required for the same amount of replication to drop by 5x, or to 1/5 of previous levels.

Conclusion

When comparing WAN optimization solutions from different vendors, continue to track data reduction rates and throughput. But don’t forget to check the endpoints to get the complete picture. It is the compound effect, after all, that reflects the true value of the solution and its benefit with regards to RPO and RTO.

Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating. View previously published Industry Perspectives in our Knowledge Library.