eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
High performance network vendor Arista Networks has launched two new products aimed at bringing low latency switching to the financial services industry. In this vertical, latency is everything and small differences in performance can have a big impact on revenue and competitive differentiation.
The company’s new products add to its 7130 family. The 7130LBR, based on Broadcom’s Jericho 2 silicon, is a compact, one-rack unit (RU) switch. The second is the 7130B, which is a 2 RU switch and uses the Intel Tofino chip.
Both run Arista’s EOS operating system for software and integrate full-featured layer two (L2) and layer three (L3) switching and open programmability with layer one (L1) connectivity. As mentioned previously, the switches are designed specifically for financial services firms to help reduce the complexity and latency of trading networks.
Also see: Top Edge Companies
Software Enables Silicon Diversity
One of the interesting aspects of Arista’s architecture is that since most of the differentiation is delivered in software, the company can use the silicon that best meets the needs of the target audience. Typically switching vendors that use Broadcom can only use Broadcom and must rewrite the operating system for other silicon. This flexibility is highlighted here as the company has one operating system but can build an Intel and Broadcom version, giving customers choice but not adding management complexity.
Low latency is an area Arista has excelled in since its inception. Eight years ago, Arista added to its leadership in this area when it introduced L1 switching—the equivalent of a telephone exchange—into trading networks, creating direct circuits between different points in the network. This year, Arista is moving to full-featured switching with the introduction of the two new products. Both switches integrate with CloudVision, Arisa’s network management.
Massive Scale and Low Latency
The 7130LBR is a 96-port system combining L1 with two high-performance, 48-port FPGAs. When adding different features into the 7130LBR switch, Arista took a massive amount of infrastructure and condensed it down. The switch is comparable to “having two of Arista’s current devices with a very large L1 switch, all in one rack unit,” Martin Hull, VP of Cloud and Platform Product Management at Arista, told ZK Research in an interview.
The 7130B Series provides a high density L1 fabric, which enables hundreds of devices to be interconnected with thousands of edge ports and hundreds of FPGAs—all while having a minimal impact on latency. For trading firms, latency is a major challenge due to the number of cables involved. The switch facilitates L1 forwarding, eliminating multiple network hops and cables.
“The trading market used to compete for the closest rack to the exchange. Now it competes in a different way because financial services firms have normalized link fibers to exchanges. But there are only so many racks they can have. The reason why L1 switches have a real place here is because firms can’t directly connect everyone, maybe just the most latency-sensitive algorithms,” Hull explained.
Also see: NTT Addresses the Why and When of Private 5G
Composable Networking
Together, the 7130LBR and 7130B enable composability, meaning the L1 switches can “daisy chain functions within the same devices,” said David Snowdon, VP of Engineering at Arista. FPGA, networking, and software applications that financial services firms use are dynamically composed within a single tier for better performance.
The two switches combined will help Arista scale up L1 networks. Financial services firms work with brokers that have their own trading algorithms. The trading market moves quickly. Firms need to scale up as they add people to trading teams. Scaling up allows firms to better compete in the trading market, and the network is core to that.
“In the past, firms would have to use several L1 and L2 switches with cabling. Even today, some firms are adding thousands of cables per month to interconnect different devices. Scaling up isn’t possible or practical due to the complexity involved,” said Snowdon. “We’re consolidating the number of devices and dramatically reducing cables. We’re bringing together agility, innovation, and simplification by putting these technologies into a single system.”
Reduction in Infrastructure Lowers Power Consumption
Deploying the switches also greatly reduces power consumption. The trading market can see a reduction in power usage of up to 90 percent, according to Hull. On top of that, there is a 100 percent reduction in the number of cables and a significant decrease in latency.
“The convergence of these two devices will allow financial services firms to deploy much larger networks and get increased functionality with fewer cables, therefore greatly reducing complexity,” Hull added. “Since FPGAs are already embedded in the network, firms can enable connections directly between FPGAs and network switches, endpoints, or anything else they may be using.”
7130LBR and 7130B are in advanced trials now. They will be available to order in Q3 with EOS and FPGA apps, and generally available after that in Q4.