Mellanox and Arista May Announcements
New Switches & Routers Spell New Possibilities
The hyperscale, cloud and hyper converged industry — despite a pause in data center equipment growth in Q1, as we reported last week— saw two major Ethernet product announcements so far this month.
Both new platforms, announced by Mellanox and Arista respectively, are supercharged to take the data center to the next level. They play into all of the big trends of cloud hosted, mass virtualization and automation of trouble shooting through telemetry. They also seem to set some new performance records.
First there is Mellanox, unveiling this week its “Ethernet Cloud Fabric” (aka ECF) new switches underpinned by its Spectrum2 chipset. Mellanox is emphasizing automated troubleshooting with telemetry, support of virtualization, scale to 400GbE, and a fully shared buffer architecture for improved performance in hyper scale cloud environments.
Mellanox ECF Platform Highlights:
- Mellanox “What Just Happened” (WJH) Telemetry: Offers improved visibility as to the root cause of network/server problems while reducing the amount of telemetry data that needs to be stored.
- Spectrum2 chipset, developed in house, has a fully shared buffer architecture for better microburst absorption. The new switches fairly and predictably divide traffic between servers.
- Fairness is an issue for ESF and ECF use cases. It is found in Hadoop, CEPH, and SAP environments. This is where fair and predictable performance matters.
- For ECF in clouds. You have to stop one tenant from robbing resources of others, whereas neighboring tenants might get robbed of another performance.
- Virtualization: Mellanox makes better spine switches with increased visibility, performance and virtualization support of VXLANs.
- 200 Gigabit: The 200GbE switches will be useful for high speed applications like Machine Learning and NVMEoF.
- 200GbE will also be deployed in volume by hyperscaler / social media companies. Some large customers prefer 200G over 400G for the larger radix as it can eliminate a switching tier in a large data center
- 8.33 Billion Packets per second
- 42MB Monolithic and fully shared packet buffer to provide high bandwidth and low-latency cut-through performance
- Robust RoCE Datapath to enable hardware accelerated data movement for Ethernet Storage Fabric and Machine Learning applications
- Half a million flexible forwarding entries to support large Layer-2 and Layer-3 networks
Next is the announcement from stalwart cloud player Arista, and it has delivered its own new products, its’ R3 Series routing platforms. The emphasis is on massive throughput, density, power efficiency and price performance.
R3 Platform Highlights
- Throughput: Packet processing offers 8X the throughput of the previous generation R1 Series and 5X the throughput of the R2 Series.
- 400G Ethernet: high-density 400G ports across all platforms to support these next-generation networks.
- New Capabilities: New R3 packet processing silicon offers significant enhancements in route scale, security and telemetry. .
- Wide Range of Form Factors: The new R3 Series comes in a wide range of form factors, including four fixed form factors, an upgrade to our award winning 7500 family, and the brand new 7800 modular chassis with up to 460 Tbps capacity per chassis
- The Arista 7280R3 delivers unparalleled Performance with Dynamic Deep Buffers and VOQ for Lossless forwarding
- Flexible design choices for 10G/25G/40G/50G/100G and 400G
- FlexRoute™ and EOS NetDB™ for internet peering with high density and low power
- Embedded Accelerated sFlow and IPFIX for high capacity network security and application monitoring
- Flexible pipeline for programmable features including Algorithmic ACLs and Inband Network Telemetry for flow and packet forensic