WAN Optimization Distorts Network and Application Performance
Monitoring and Reporting

By design, WAN Optimization devices apply various optimization techniques on the application packets and hence affect monitoring and reporting tools to varying degrees. The impact on IT organizations can be significant and may include loss of visibility into important metrics, inability to quantify the benefits of WAN Optimization, and difficulty in troubleshooting bottlenecks.

Most WAN Optimization solutions “confuse” network and application performance monitoring systems by changing packet header data, which severely impacts the measurement of application performance in an enterprise network. Optimization techniques that cause this include:

• Tunnel-based architecture (Layer-3): Most WAN Optimization solutions do not preserve the client and server original IP addresses that are needed by monitoring systems to perform end-to-end application performance reporting. Instead they tunnel the client-server traffic through a set of dedicated TCP ports between the Optimization devices. This tunnel can easily be verified by taking a trace on either side of the WAN Optimization devices to check the source/destination IP address. The following diagram illustrates how a monitoring system may lose granular visibility into each individual application flow because all flows disappear into a single WAN Optimization port/tunnel.

• A closer look at the three segments issue: Most WAN Optimization solutions consist of two peer devices on either side of the WAN. They optimize the TCP connection between a client and server by breaking a single TCP connection into three segments as shown in the illustration below:

1. Client to WAN Optimization device (WOD) at the edge
2. WAN Optimization edge device to WAN Optimization core device
3. WAN Optimization core device to server

This obscures critical information from monitoring systems that assume the existence of a single TCP connection and, therefore, only report on a single segment (typically the data center LAN segment). A monitoring system not designed for WAN Optimization creates the false impression that client to server application response time is “lightning” fast when all it is actually reporting is the response time from the server to the WAN Optimization core device. The charts below shows this effect.

Another look at this data in the following report illustrates the danger in relying on existing response time monitors to measure the impact of WAN Optimization. It shows end-to-end response time measurements broken down into their various components, including network round-trip time, retransmission delay, data transfer time, and server response time. On the left of the chart are response times before WAN Optimization, showing most of the latency is due to data transfer. The center of the chart illustrates how the segmentation of WAN optimized TCP connections causes performance monitors to report erroneous “lightning fast” performance, because only the data center LAN response times are being measured. The right side of the chart shows how SuperAgent, using the integrated WAAS management interface, accurately reports the impact of WAN Optimization by measuring latency across all three segments of the client-to-server connection.

Integrated NetQoS and Cisco WAAS Offering Preserves Accurate Monitoring and Reporting

To ensure accurate reporting after a Cisco WAAS deployment and to protect customers’ existing investments, NetQoS and Cisco took the initiative to deliver the first and only solution for accurate and comprehensive End-to-end Response Time Monitoring for WAN Optimization. As a result, NetQoS is the first network performance management vendor to provide accurate reports that quantify the latency improvements delivered by WAN Optimization.

Download a Tech Brief on the NetQoS Cisco WAAS solution now or schedule a demonstration.

Loading...