Arista's Latest Switches Pack AMD Xilinx FPGAs To Fuel High-frequency Traders

Arista Networks aims to eliminate every last ounce of excess latency from your network. And in pursuit of this goal, the company tapped an all-star cast of Broadcom, Intel, and AMD Xilinx silicon to power its latest switches.

Suffice to say, these aren’t your standard L3 switches, and you certainly aren’t going to find them in your average datacenter or network closet. The latest entries to Arista's 7130-series pack a CPU, FPGA, and switch ASIC into a single appliance that’s designed to accelerate latency-sensitive applications.

“When every meter of fiber counts for around five nanoseconds of latency, network architects have to choose carefully how to construct applications and manage traffic flows,” Martin Hull, VP of Arista’s cloud titans and platform product groups, wrote in a Wednesday blog post.

High-frequency trading is the obvious use case for these kinds of devices, where a few nanoseconds could mean the difference between making money and losing it. However, Arista claims the onboard FPGA and CPU resources also make it ideal for other applications, including high-performance network monitoring, multiplexing, and even as a low-latency firewall.

• Fully automated AI networks less than 5 years away, reckons Juniper CEO
• Arista Networks makes $4.3b bet on components amid supplies drought
• Was Nokia's Microsoft switch deal win a reward for SONiC contributions?
• Despite shortages, networking hardware market grew strongly in 2021

Arista expands 7130 switch line

“There is no perfect solution for all customers in this space that combines ultra-low latency, FPGA, L2/L3 features, performance, and density,” Hull wrote.

However, the networking vendor’s latest switches aim to change that. Arista's 7130LBR pairs a Broadcom Jericho2 switch ASIC with a pair of AMD Xilinx FPGAs and an eight-core CPU.

Broadcom’s Jericho2 provides the low-latency switching, while workloads are executed on the intergated FPGAs rather than on an external server.

While it's possible to achieve similar functionality using a standard L3 switch and an FPGA-equipped server, the round-trip latencies would make the combination untenable in many applications, according to Hull.

For this reason, the CPU is only intended to handle Arista's EOS network operating system and CloudVision intent-based networking stacks, as opposed to customer workloads.

By comparison, Arista's 7130B is all about density. It ditches the FPGAs and Broadcom silicon in favor of Intel's Tofino switch ASICs. The appliance provides up to 256 ports, which can be broken out using one of 32 QSFP-DD 10Gbit/sec interfaces.

Developed by Barefoot Networks prior to its acquisition by Intel in 2019, Tofino was designed with programmability in mind. In addition to sub-10 nanosecond latencies, the chip is fully P4 programmable, enabling customers to optimize packet processing for their specific workloads.

"The Arista 7130B is an ideal partner to the 7130LBR with up to 256 ports to scale out the network without adding incremental latency," Hull wrote.

Both switches are sampling to customers now, with general availability expected in the fourth quarter. ®