Intel has launched an IPU (Infrastructure Processing Unit), which, like SmartNIC, aims to improve processing, networking and storage by freeing up the computer’s CPU.
In this blog you will learn what an IPU is, what it does and why it is important. We will also introduce you to the IPU we are currently working on for a network platform.
What is an IPU?
Exempt from Intel in 2021, an IPU or infrastructure processor is a programmable network device designed to enable cloud and communications service providers to reduce costs and free up processor performance.
Like SmartNIC, IPUs aim to address the complexity and inefficiency of data centers that result from information overload.
The main goal of the IPU is to enable customers to make better use of resources with a secure, programmable and stable solution that allows them to balance processing and storage.
How does the IPU work?
An IPU is a programmable network device that intelligently manages infrastructure resources at the system level by securely accelerating these functions in a data center.
It allows cloud operators to move to a fully virtualized storage and networking architecture while maintaining a high degree of performance, predictability and control.
IPU has special functionality to accelerate modern applications that are built using microservice-based architecture in the data center.
As a result, the cloud provider can securely manage infrastructure functions while allowing its client to fully control CPU and system memory functions.
What are the main features of an IPU?
There are four main features of an IPU:
- Highly intelligent infrastructure acceleration
- Security, control and isolation at the system level
- Common software frameworks
- Programmable hardware and software built according to customer needs
Why are IPUs important?
The IPU has the ability to:
- Accelerate infrastructure features, including storage virtualization, network virtualization, and security with dedicated protocol accelerators.
- Free the CPU by transferring the network storage and virtualization features that were previously done in the CPU software to the IPU.
- Improve data center utilization by allowing flexible workload placement.
- Allow cloud service providers to customize the deployment of infrastructure features at the speed of software.
On a larger scale, evolving data centers will require a new smart architecture in which large-scale distributed computing systems work together seamlessly as a single platform.
This will help address today’s challenges related to blocked resources, congested data flow and incompatible platform security.
Within this new architecture, there will be three calculation categories: CPU for general purpose calculations, XPU (cross-platform module) for application-specific or load-specific acceleration, and IPU for infrastructure acceleration.
All three categories will be connected through programmable networks for efficient use of data center resources.
Emerging data centers will require a new intelligent architecture in which large-scale distributed computing systems work together seamlessly as a single platform.
Let’s look at one IPU we’re working with right now: the Intel Mount Evans IPU.
Intel Mount Evans IPU
Trenton Systems is currently being switched on Intel Mount Evans IPU in a network platform that we create for government and military applications.
IPU Mount Evans has three main features:
- Ready for hyper-scale: Designed in collaboration with the best cloud provider with integrated training from many generations Intel FPGA (Programmable input matrix) SmartNIC to ensure high performance under real loads with security and insulation from the start.
- technology Innovation: Has a highly programmable packet handling mechanism, NVMe storage interface enhanced by Intel Optane technologyreliable next-generation transport, advanced crypto and compression acceleration.
- software: It has joint design of software / hardware / axel, use of P4 boss, leverage and advanced Data plane development kit (DPDK) and Storage performance development kit (SPDK).
IPU is a critical component of the computing architecture that seeks to unload tasks from the host computer’s processor to improve networking, processing, and storage.
With infrastructure accelerations, system-level security, and adaptive hardware and software, IPUs can help deliver secure, accelerated AI / ML / DL workloads and improve big data analysis to provide useful real-time information.
Through fully virtualized critical capabilities, IPUs allow data center users and administrators to avoid congestion through improved predictability, performance, and control.