Pancake Podcast: Which Comes First, Kubernetes or the Service Mesh?
There are still so many questions about service mesh. So, so many questions, as a curious, spirited and hungry KubeCon + CloudNative Con North America 2019 attendees raised in The New Stack’s early morning pancake breakfast panel discussion, sponsored by VMware. Which comes first, Kubernetes or a service mesh? How many sidecars per app is the right number of sidecars? Should a small company start with a service mesh, or just bring it in later when business booms and ops grow?
Luckily, we had assembled a savvy panel of service mesh experts from across the industry to address these queries, namely:
- Jean Atelsek / Analyst / 451 Research
- Vivian (Wei) Fu / Engineering Manager / Uber
- Fuyuan Bie / Software Engineer / Pinterest
- Lita Cho / Senior Software Engineer / Lyft
- Ines Envid / Product Manager / Google Cloud Platform
- Pere Monclus / Vice President and Chief Technology Officer, Network and Security / VMware
To the answer of how many sidecars is enough: It’s a trade off between resource use and management efficiency, Cho and Bie told the audience. Lyft has five sidecars, while Pinterest has 10. Which should come first, Kubernetes or service mesh? It really depends on the enterprise’s requirements, Envid advised, though Cho mentioned that in Lyft’s case, the service mesh came first, and it helped making Kubernetes onboarding a lot easier.
Overall, Atelsek concluded, there’s a lot of trepidation among businesses about running a service mesh and having a unified management layer, but a lot of cloud native companies have it in production. And they are enjoying the benefits, as a service mesh provides end-to-end connectivity, improving the security and management of microservices, Monclus noted.
Listen to the podcast to hear in full!
— The New Stack (@thenewstack) November 20, 2019