Service mesh technologies are being used in production just six months after the KubeCon North America keynotes that explored what a service mesh is, and its benefits to both organizations and developers. Many are still attempting to truly grasp what a service mesh offers, particularly as observability and compatibility concerns are raised. Joining us on this episode of The New Stack Makers podcast, from KubeCon + CloudNativeCon, are Dan Berg, distinguished engineer at IBM who is working Istio service mesh and Ivan Kruglov, principal developer at Booking.com to explore the overall management of a service mesh and how it relates to microservices.
“A lot of times I’ve seen that the proxies are awfully static. You define them once and you build your app, you never interact with the proxy all that much. What that means is that you still have to deal with the complexity of building microservices. What we see with the mesh is, we see the proxy as ubiquitous across the entire mesh. As a developer, I can remove a lot of logic out of my app and let the mesh take care of it,” said Berg.
Kruglov went on to note that the benefits of service mesh outweighed the impact of potential latency and cost, with the impact to overall server latency quite minimal, as the proxy is kept close to the client with a sidecar proxy with the use of a service mesh. “What we see is the cost is very acceptable. The increase is about 1MS, and in most of the cases, we won’t notice in our graph. It’s absolutely affordable for the benefits which Envoy gives us.”
Other proxies that were looked at were a node level proxy, with sidecar proxies running alongside the application, whereas node proxies run on a virtual machine in a container. At the end of the day, Berg explained that node proxies don’t have as much information about the application, and lose that information while sidecar proxies do not.
In this Edition:
3:08: What has happened to the sophistication of APIs that has led to the service mesh being used at Booking.com?
7:40: What does consolidation do for performance?
14:43: Exploring what use cases service mesh is applicable to
21:15: Clarity in the development process and continuing to define it within one’s service mesh
24:31: What role the networking layer plays in the stability of service mesh overall
27:24: What resources need to be invested in the community around Istio?