With the data that Catchpoint provides us, we were able to quickly and easily identify an issue that was actually with a vendor rather than us. In Catchpoint, you can dig into these charts to see sources and destinations, addresses, and a lot more data that becomes valuable when troubleshooting.
In order to monitor performance for any asset request, Fastly focuses on five major components of performance: DNS Lookup, Connect Time, TLS Handshake, Wait Time, and Load Time.
DNS lookup and response time can obviously have a significant impact on overall page performance. By using Catchpoint’s DNS observability solution, Fastly can do a deep dive into their DNS performance whenever they see a spike in performance.
Using this feature, the Fastly operations team can see in-depth DNS data such as resolvers, intermediaries, source/destination information, and resolved addresses. They can then quickly and easily troubleshoot and share insights gained with their DNS providers around the world.
All told, Catchpoint’s DNS observability functionality allows Fastly to observe and address issues in a matter of hours or even minutes rather than days.
The second part of this process takes place with the connect time, or three-way TCP handshake. Fastly needs to monitor connect time at both the network layer and the TCP layer, so the ability to isolate the two becomes critical when troubleshooting.
Fastly terminates TLS at the edge, which can reduce both TCP and TLS handshake costs for their customers. In order to provide this advantage to their customers, Fastly has to keep their TLS stack optimized at all times, which requires regular testing.
In Catchpoint, they can view the full process of the TLS handshake and the complete round-trip data to ensure that they are hitting their goals. Since Catchpoint is capable of isolating the performance of the TLS handshake, Fastly is able to monitor the effectiveness of their TLS stack and ensure secure, high-quality experiences for end users.
Catchpoint’s ability to measure wait time helps Fastly monitor the speed at which requested objects are delivered. From an end user perspective, wait time is the time between when the request was sent and the first byte of the response is received. For cached objects, wait time is primarily a function of storage efciency, caching effciency, and network round-trip time. Since connect time measures the network components, wait time becomes a great indicator of storage and caching efciency.
Naturally, the ability to measure when end users’ browsers are able to start downloading the actual page elements is a very valuable metric, and to ensure that they’re meeting expectations, Fastly holds a general philosophy to keep their Connect and Wait times as close to each other as possible. Fastly uses Catchpoint to evaluate the performance of their customers to see whether these two metrics diverge for cached objects. If so, Fastly can then use the insight provided by Catchpoint to troubleshoot quickly.
For example, at one point, there was a wide gap between connect time (yellow line) and wait time (red line) indicating that a configuration issue meant objects were being fetched from origin instead of being cached at the edge. Using Catchpoint, Fastly was able to quickly address the Cache-Control headers being set by the customer at origin. With the help of Fastly, the customer addressed the issue, resulting in converged connect and wait times which indicate optimal performance for object delivery.
The final component of the transaction is load time, which is essentially the time it takes from when an end user receives the first byte of the response to when the transaction is completed. Load time is a vital component when measuring the performance of larger objects and is generally a function of bandwidth and TCP efciency. By evaluating load time, Fastly ensures their edge locations can deliver objects optimally, while enabling them to adjust TCP optimizations based on each customer’s specific workload and requirements.
Thanks to Catchpoint’s in-depth analytical capabilities, Fastly is able to isolate each of these metrics in every performance test that they run. As a result, they are able to catch and troubleshoot issues faster and more accurately.
"With the data that Catchpoint provides us, we were able to quickly and easily identify an issue that was actually with a vendor rather than us. In Catchpoint, you can dig into these charts to see sources and destinations, addresses, and a lot more data that becomes valuable when troubleshooting," says Jason Evans, director of product at Fastly.