Chip E2000 is the new chip launched jointly by Intel and Google Cloud to promote security and effectiveness in data centers. This chip, codenamed Mount Evans, is an infrastructure processing unit, or IPU. So that takes over network processing tasks while the CPU takes care of the computing part.
This new chip, which despite being jointly developed Intel you can sell it to other customers, too adds a layer of security between clients using a cloud service. For its part, Google has also integrated it into its virtual machines, such as the so-called C3 VM, which is powered by fourth-generation Xeon processors from Intel.
Both companies have presented Chip E2000 as a “significant milestone” in terms of a end-to-end programmable platform. In part, thanks to the fact that it has an architecture that offers the possibility that C3 machines work 20% faster than the previous ones, the C2.
Google Cloud’s first ASIC infrastructure processing unit
“We are proud to have co-engineered Google Cloud’s first ASIC infrastructure processing unit,” said Nick McKewon, Network Group Manager at Intel. The E2000 chip brings to the table the rise of infrastructure processing units (IPUs).), also called DPU. More advanced network chips than traditional smart NICs.
“The System on a Chip hardware architecture introduced in C3 virtual machines can enable better security, isolation, and performance,” Google Cloud also notes on its blog about C3 machines. In addition, this architecture is expected to enable a more comprehensive product portfolio in the future, such as support for bare metal native instances.
It should be noted that this new architecture enables block storage by decoupling compute instance size from storage performance. The IPU provides C3 virtual machines with 200 Gbps low latency networking, encrypted at line speed by the open source PSP protocol.
Looking to the future, Google ensures that the IPU and TPU will offer the option for the infrastructure to become more autonomousa, as they highlight in their blog: “With the exponential increase in the complexity of cloud infrastructure (…) we must turn to automation to manage these platforms efficiently at scale (…) along with Infrastructure as Code, custom chips like Titan, TPU, and IPU pave the way for a not-too-distant future where we will automate more than half of all infrastructure decisions, dynamically configuring systems in response to usage patterns.”
