Cisco today announced it plans to acquire Isovalent as part of an effort to advance adoption of an approach to networking that leverages extended Berkeley Packet Filtering (eBPF) within the microkernel of the operating system to process packets at levels of speed and scale that would not otherwise be possible.
Tom Gillis, senior vice president and general manager for Cisco Security, said the open source Cilium networking software based on eBPF will move a wide range of networking and security functions that today are generally handled today by hypervisors into the kernel of the operating system.
In addition to enabling much higher levels of performance at scale, that approach will also create an opportunity to embed the processing of those functions into a range of classes of processors to further improve performance at levels of unprecedented scale, he added.
At the same time, running those functions at the microkernel level will also provide greater visibility into processes that will ultimately improve cybersecurity, noted Gillis. In addition to a curated instance of Cilium for networking Kubernetes clusters dubbed Cilium Mesh, Isovalent also developed a security tool for enforcing runtime behaviors dubbed Tetragon.
By leveraging eBPF and Cilium, it becomes feasible to see the relationship between processes in a way that surfaces anomalous behavior without necessarily having to decrypt every network packet, noted Gillis.
Cilium has been gaining traction primarily in cloud-native application environments where it is deployed across Kubernetes clusters. In fact, Cilium is now one of the open source projects most frequently downloaded from GitHub, noted Gillis.
Once the deal closes next spring, Cisco plans to further broaden the reach of Cilium to include support for multiple types of applications running in both the cloud and on-premises IT environments, said Gillis. In addition, more advanced IT organizations are also starting to encapsulate legacy monolithic applications running on hypervisors that can run on top of Kubernetes clusters, he added. The overall goal is to make eBPF and Cilium an open, ubiquitous standard, said Gillis.
It’s still early days so far as the number of instances of operating systems that support eBPF; it’s only been generally available in the most recent releases of Linux. Microsoft, meanwhile, has pledged to support eBPF in Windows, an event that should make it much simpler to create network overlays across a heterogeneous IT environment. The adoption of eBPF itself is now being advanced by The eBPF Foundation, an arm of The Linux Foundation committed to enabling other platforms beyond Cilium to run as sandbox programs at the kernel level in any operating system. It may be a few more years before that vision is completely realized, but the number of networking, security and storage platforms that will take advantage of eBPF in the months ahead is expected to expand considerably.
The challenge, of course, is finding a way to deliver those services at a high enough level of abstraction that mere IT mortals can manage them without having to rely nearly as much on specialists who all too often are hard to find and retain.