Optimizing infrastructure and software for Intel processors
Ericsson’s cloud infrastructure platforms and applications have been optimized for Intel x86 Central Processing Units (CPUs) to capture the advantages of Intel architecture and boost performance. In this blog post, we explore the benefits of this collaboration further.
Ericsson and Intel have a long history of working together to combine their complementary technologies. Over the years, Ericsson and Intel have collaborated to meet industry requirements across numerous generations of Intel silicon. As a result, Intel technology has become a key component embedded in – or tested against – Ericsson products to ensure improved application performance.
In this post, I’d like to explore the collaboration between the two companies across several product lines in more detail and highlight why it’s important for Ericsson cloud infrastructure and packet core to take advantage of Intel x86 CPUs.
But to start off on a more general note: the strength of the Intel architecture is leveraged to trim the Ericsson virtualization layer software, orchestration software, virtualized network functions’ performance, or any other Ericsson software – or hardware product – destined to run on Intel technology. And as a result, we’re able to ship Ericsson products and applications that are optimized for Intel x86 processors with an associated performance tag attached.
Bipin Thomas, Product Manager at Ericsson Digital Services says: “We’ve worked persistently together with Intel to capture the advantages of our applications which have been developed to use Intel silicon architecture, and as a result we’ve been able to show our customers leading performance.”
The attractiveness of Intel’s large ecosystem of server vendors also comes into play, meaning that an Ericsson application, or any other Ericsson software optimized for Intel® Xeon® Scalable processors, automatically executes equally well irrespective of the server vendor. This is important for tapping into the ecosystem, but I’ll share more on that below.
A long history of collaboration
Ericsson-Intel collaborations span from technology exchange for radio base station development, to application optimization, and on to Ericsson’s platforms running native and virtual network functions. The latter include, for example, the NEBS3-compliant Ericsson blade server platform BSP 8100 and the Ericsson Hyperscale Datacenter System 8000, which both utilize the Intel® Xeon® Scalable processor and serve as hardware platforms in the Ericsson Software Defined Infrastructure (SDI) offering. And, as mentioned in a blog post earlier this year, Ericsson and Intel are collaborating to develop a software-defined infrastructure management platform.
With this kind of widespread and long-term partnership as base, constructive ways of working have been established whereby the development organizations leverage each other’s capabilities. Let’s take the packet core user-plane as an example: obviously, the packet core user plane is the function in a service providers’ mobile core deployment that carries the burden on account of the current or imminent 5G data explosion.
Hence, the rational for optimizing this particular application is especially strong, which is why Ericsson’s Evolved Packet Gateway and dual-mode 5G Core user plane are developed with a tight integration towards Intel x86 CPU, utilizing architecture-specific optimization for packet processing, encryption and cache optimization.
Bipin Thomas, Product Manager at Ericsson Digital Services says: “Since utilizing Intel’s new 2nd Generation Intel® Xeon® Scalable Processors for Ericsson packet core user plane, we’ve measured improvements of around 10 percent compared to the earlier generations. The difference may not be perceived as huge, but it represents a valid and important step as we continuously improve performance and shave off incremental costs in all steps.”
Intel architecture strengths
Intel unified server architecture provides deterministic and scalable high-throughput packet performance with zero packet loss particularly suited for high reliability service provider deployments. Intel DDIO (Direct Data IO) is one of the critical components, which enables optimal networking performance with higher throughput and lower packet latencies.
Purposefully built Intel® Xeon® Scalable processors with DDIO also provide optimal packet processing while memory bandwidth is leveraged for actual applications, and balanced Intel Xeon architecture with DDIO offers high throughput with efficient Core-Memory utilization.
In conclusion, Intel offers an end-to-end scalable solution highly appropriate for a great many Ericsson applications. It’s a balanced architecture with segment-specific accelerators, storage and connectivity, all with a unique software ecosystem enablement.
Intel ecosystem strengths
The Intel x86 ecosystem is a mature and well-established IT and Telco NFV server ecosystem and Intel® Xeon® Scalable Processors hold the largest market share globally. In an open multivendor environment, Intel silicon constitutes a critical component in all large vendors’ server offerings.
Given the high degree of application optimization on Intel x86 CPU, including the above mentioned Ericsson packet core, but also Ericsson’s OpenStack based IaaS offering (Cloud Execution Environment, CEE) and Kubernetes based CaaS offering (Cloud Container Distribution, CCD), compatibility with a large ecosystem is critical for an efficient go to market strategy.
The magnitude of the Intel x86 ecosystem is obviously also beneficial in that communication service providers get access to a wide variety of server vendors, while still experiencing optimized Ericsson application and software performance.