MultiAccess is a network application that combines low-latency packet multiplexing with the ability to segregate traffic between individual clients, or groups of clients, via configurable filters.
Arista MultiAccess enables the creation of a shared, low-latency infrastructure ideal for managed service providers or brokers providing direct access to clients.
The MultiAccess application optimizes the paths for many-to-one environments where it is critical that clients maintain segregation of traffic between each endpoint. These paths are further enhanced to allow the processing to occur at the fraction of the latency of a Layer 2/3 switch.
Filtering is accomplished via configurable access control lists (ACL). Packets may be streamed through the filter using an aggressive cut-through technique, allowing for a very low latency of
Features | Benefits |
---|---|
Low latency, multi-tenant multiplexing | Aggregate client-to-server streams in as low as 53 ns and filter server-to-client streams in 89 ns*. |
Connection sharing | Take advantage of a cost-effective alternative to dedicated exchange links, WAN links or other shared services. |
Flexible ACLs | ACLs support permit/deny rules based upon source/destination IP address/port number. IP addresses may be extended using wildcards with CIDR style notation. |
VLAN features | Supports 802.1Q Tunneling (Q-in-Q) by performing VLAN insertion toward the host port, and stripping toward the client ports. |
Ingress filtering | Ingress filtering is available in 2x23_if and 1x47_if modes that allows small ACLs to be applied to traffic from clients to the server. |
Storm Control | Monitor per-port network bandwidth or packet rate, and drop packets to maintain rates lower than pre-defined limits. |
Multifunctional | Tapping – allow access port traffic to be replicated and sent out for monitoring/capture to Arista's MetaWatch application |
Extensive packet statistics | Record extensive telemetry information for MultiAccess and L1+ switch operation e.g. per-port counters and per-queue depths; for monitoring, diagnosis and troubleshooting. Statistics are available via CLI, InfluxDB and other telemetry mechanisms. |
Deterministic | Know and rely upon your system's latency for fairness, or implement an ideal execution environment for your orders. Without contention, the MultiAccess aggregation latency varies by only +/- 6 ns. |
Easy to monitor and manage | Arista provides a complete range of additional features including:
|
* based on lowest latency configuration
Latency Tables - Low Latency Modes
Low latency modes are highly optimised for low latency. Features that increase the latency are disabled in these modes, including VLAN support, ingress filtering, and large ACL scale:
Mode | Board Standard | Min (ns) | Average (ns) | Max (ns) |
---|---|---|---|---|
1x47_ll | E | 56.9ns | 61.9ns | 68.4ns |
1x47_ll | L | 47.8ns | 52.7ns | 58.5ns |
1x47_ll | LB | 48.3ns | 52.9ns | 58.9ns |
2x23_ll | E | 49.3ns | 54.3ns | 60.6ns |
2x23_ll | L | 47.6ns | 52.6ns | 58.3ns |
Table 1.1 Client to Server path – Low Latency modes
Server to Client latency in Low Latency modes is similar to regular MultiAccess modes, however only ACLs up to 8 rules deep are supported:
Mode | Board Standard | Min (ns) | Average (ns) | Max (ns) |
---|---|---|---|---|
1x47_ll | E | 109.1ns | 116.5ns | 126.8ns |
1x47_ll | L | 101.3ns | 108.9ns | 118.8ns |
1x47_ll | LB | 103.3ns | 111.0ns | 120.8ns |
2x23_ll | E | 103.2ns | 110.1ns | 129.1ns |
2x23_ll | L | 101.3ns | 109.0ns | 119.6ns |
2x23_ll | LB | 101.3ns | 108.6ns | 118.9ns |
Table 1.2 Server to Client path modes
Latency Tables - Full Featured Modes
Latency for the Client to Server path for modes supporting VLANs, ingress filtering, and larger ACLs on the return path. Using 7130-LB with 10G client and 10G server port configuration as a representative sample:
Mode | Min (ns) | Average (ns) | Max (ns) |
---|---|---|---|
1x47 | 93.2ns | 99.4ns | 106.1ns |
2x23 | 83.5ns | 89.8ns | 96.5ns |
4x11 | 83.6ns | 89.8ns | 96.7ns |
6x7 | 80.4ns | 86.8ns | 93.4ns |
Table 1.3 Client to Server path modes – larger ACLs
Latency for the Server to Client return path for modes supporting larger ACLs on the return path. Using 7130-LB with 10G client and 10G server port configuration as a representative sample:
Mode | Min (ns) | Average (ns) | Max (ns) |
---|---|---|---|
1x47 | 114.1ns | 120.7ns | 127.9ns |
2x23 | 111.4ns | 117.6ns | 124.5ns |
4x11 | 111.4ns | 117.6ns | 124.4ns |
6x7 | 111.4ns | 117.6ns | 124.8ns |
Table 1.4 Server to Client return path modes – larger ACLs
Arista MultiAccess for Managed Service Providers & Brokers
WAN Link Multiplexing
Arista MultiAccess can be used to aggregate several sources of network traffic into a WAN link and, optionally, de-multiplex from a WAN link. Often these links can be bandwidth limited (e.g. 1 GbE) with MultiAccess being able to provide a low-latency translation to and from the bandwidth-limited link. The layer 1 functionality in the Arista 7130 can be used to tap those links.
To reduce the traffic volume transmitted, Arista MultiAccess can filter market data down a single, bandwidth-limited (1GbE) line.
VLAN tagging and stripping can be used to multiplex multiple independent layer 2 links into a single VLAN trunk to be transmitted and then de-multiplexed. The link partner may be a traditional VLAN-aware switch, or another instance of MultiAccess.
Lower-latency Exchange Access
Managed Service Providers and Brokers offer their clients access to financial exchanges and can use MultiAccess to provide significantly lower latency than typical network switches. Arista MultiAccess allows these firms to share the connection to the exchange with more than one client, while maintaining isolation between the clients. MultiAccess hence provides an ideal solution combining multiplexing (aggregation) functionality with a high-performance return path; filtering the return-path individually per client. The latter benefit from significantly lower latency in their trading connectivity as well as smaller variation in that latency. Arista MultiAccess implements a number of features which enhance this use case, for example implementing ACL filters on ingress before multiplexing, and storm control, which can avoid specific clients using an unfair amount of the network capacity.
Optimized for:
- Arista 7130L Series with embedded Xilinx FPGA.
Compatible with:
- Arista 7130E Series with embedded Xilinx FPGA(s).
Product Overview
Product Briefs
- .MetaWatch Product Brief
- .MetaMux Product Brief
- .MetaProtect Product Brief
- .ExchangeApp Product Brief
- .MultiAccess Product Brief
- .SwitchApp Product Brief
- .Developer Product Brief
White Papers
- .An Overview of Arista Ethernet Capture Timestamps
- .5 Ways to Optimise Exchange Connectivity Latency
- .5 Things to Consider When Choosing an FPGA Platform
- .Measuring the Absolute Accuracy of 10GbE Packet Timestamping
- .Four Key Trends in the Networked use of FPGAs
- .STAC-TS™ BENCHMARKS - Accuracy of Network Timestamping and Burst Capacity of Capture
Case Study
- .Case Study: Electronic trading firm gains ultra-low latency exchange connectivity with Arista
- .Case Study: Deutsche Börse Group monitors every trade with Arista
- .Case Study: Enyx chooses Arista to accelerate ultra-low latency solutions for major financial Service Providers