Exclusive interview with Liz Rice, chief open source officer at Isovalent on the Launch of Cilium Service Mesh.
VMblog: Why does the cloud-native community need another service mesh, and what
is Cilium Service Mesh's unique value to Kubernetes platform teams?
Liz Rice: Lots of platform teams would like the features of a service mesh for
their Kubernetes infrastructure, but many report that sidecars bring
additional administrative complexity and resource overhead. Cilium
Service Mesh gives platform teams a sidecar-less option that is much
lower latency and more resource efficient, and avoids the complexity of
injecting sidecars.
VMblog: What are the main advantages of bringing
service mesh capabilities closer to the Linux kernel? And how does
Cilium Service Mesh accomplish this?
Rice: Cilium Service Mesh is
able to achieve these performance breakthroughs because it's based on
eBPF. eBPF is an operating system abstraction that allows dynamic
changes to kernel behavior, and the Cilium project has long been using
eBPF to provide a highly efficient networking solution for Kubernetes.
We've now extended Cilium's capabilities to add Service Mesh
functionality in the latest 1.12 release.
In the sidecar model,
every single packet between two pods has to traverse two userspace
proxies, making the network path very convoluted. When users choose the
sidecarless option in Cilium Service Mesh, the network path becomes much
shorter and more efficient. It also avoids duplicating the memory
needed to run a proxy in every pod in the sidecar model, instead running
one proxy per node (see: How eBPF will solve Service Mesh - Goodbye
Sidecars).
VMblog: Who are the main contributors and maintainers involved with Cilium and this Cilium Service Mesh launch?
Rice: Cilium was originally created by the team at Isovalent, but we
contributed it to the CNCF last year to cement its status as a community
project. Major contributors include Datadog, F5, Form3, Google,
Isovalent, Microsoft, Seznam.cz, and The New York Times.
VMblog: Where do you see cloud-native infrastructure evolving most rapidly today? What are its hardest unsolved problems?
Rice: If we consider Kubernetes to be the distributed operating system, then
we need distributed tools for networking, observability and security.
eBPF is a great platform for building these tools, and we'll see the
ability to move more functionality into the kernel creating great
strides in performance of cloud-native infrastructure tooling.
VMblog: Tell VMblog readers about Isovalent, the company. Who is involved, and
why is this a company to watch in cloud-native infrastructure?
Rice: All of the core maintainers of the Cilium project share a deep focus on
low-level network and security infrastructure, right down to the
operating system kernel. Cilium creator and CTO and co-founder at
Isovalent, Thomas Graf, had a long background as a kernel maintainer
with eBPF and Open vSwitch. Daniel Borkmann from our team is one of the
two eBPF maintainers for the Linux kernel. Isovalent uses our broad
experience to create an enterprise distribution of Cilium that's being
used by major brands in some highly scaled deployments, such as Adobe,
Bell Canada, and IKEA as well as many managed Kubernetes platforms
including products from Google Cloud and AWS.
##