Maximizing vCMTS Data Plane Performance with 3rd Gen Intel® Xeon® Scalable Processor Architecture

Intel platform technologies boost virtualized cable modem termination system (vCMTS) data plane performance by 70%.




The standardization of distributed access architecture (DAA) for DOCSIS and further advancements in the flexible MAC architecture (FMA) standard have enabled the transition to a software-centric cable network infrastructure. This paper will explore how Intel technologies can be utilized to increase performance for vCMTS for the various deployment scenarios.In each scenario, the same DOCSIS MAC software may be deployed, whether as:

  • a virtual MAC core (vCore), also known as a virtualized cable modem termination system (vCMTS) on a server in a multiple-system operator (MSO) headend,
  • part of a remote-MAC-core (RMC) and remote-PHY deployment on an edge compute node,
  • or part of a remote-MAC-PHY device (RMD) deployment on an edge compute node.

Independent software vendors (ISVs) can significantly improve vCMTS performance on 3rd Gen Intel® Xeon® Scalable processor architecture (codename “Ice Lake”) by taking advantage of the gen-on-gen CPU architecture enhancements which include bigger cache-sizes at each level, higher core-count, more memory channels at higher speed and greater I/O bandwidth. This paper focusses specifically on how to take advantage of advanced features such as:

  • Enhanced Intel® Advanced Vector Extensions 512 (Intel® AVX-512)
  • Dual AES encryption engines
  • Intel® Vector AES New Instructions (AES-NI)
  • Intel® Vector PCLMULQDQ carry-less multiplication instruction
  • Intel® QuickAssist Technology

The paper also provides insights into implementation options and establishes an empirical performance data baseline that can be used to estimate the capability of a vCMTS platform running on industry-standard, high-volume servers based on 3rd Gen Intel Xeon Scalable processor architecture. Actual measurements were taken on a 3rd Gen Intel Xeon Scalable processor-based system (see Appendix D for configuration details).

The data demonstrates how a single 3rd Gen Intel Xeon Scalable processor core running at 2.2 GHz clock speed can support the downstream channel bandwidth for close to five orthogonal, frequency-division multiplexing (OFDM) channels in a pure DOCSIS 3.1 configuration for an Internet mix (IMIX) traffic blend. When Intel® QuickAssist Technology (Intel® QAT) acceleration is employed in the system, a single core can satisfy close to the maximum downstream bandwidth of a pure DOCSIS 3.1 configuration. Since vCMTS workloads exhibit good scalability, a typical server blade based on dual 3rd Gen Intel Xeon Scalable processors (e.g. with 64 processor cores total) can achieve compelling performance density.

It is worth noting that these performance benefits can also be delivered by deploying a general-purpose compute component in the RMC and RMD scenarios. In other words, the same performance benefits of Intel architecture and complementary technologies, such as Intel QAT and the Intel Multi-Buffer Crypto library, can be used to achieve similar performance on an edge compute node.

DOCSIS MAC functionality can be broken down into four categories: downstream data plane, upstream data plane, control plane, and system management. From a network performance perspective, the most compute-intensive workload is data plane processing, and consequently, is the focus of this paper.