How can Metamako devices work for you?
View the use cases below to find out how Metamako can be applied to your business.
MetaConnect makes for a great front-end to WAN links. It provides visibility into the signal integrity properties of the link. It also allows for media conversion from long-range SFP+ optical modules to direct attach copper (at the same line rate) offering potentially substantial savings by reducing the dependency on long-range optics from switch and router vendors who require that you use their transceivers. The management interface provides packet statistics, transceiver properties, signal quality monitoring (eye diagrams) and rudimentary packet header capture to improve troubleshooting and management of your links.
Reconfigurable Patch Panel
MetaConnect functions as a remotely configurable patch panel allowing a "wire-once" approach to the data center. Simply connect your devices and links up to any of the physical ports and from a comfortable distance you can reconfigure the interconnections between the ports as well as enabling and disabling them. Control is provided via a standard command line interface, web console or in future, directly by software on the device. It avoids trips the data center or test lab and there is no need to use 'remote hands'. MetaConnects does this with virtually no latency or jitter and no impact on bandwidth.
In this example, the MetaConnect can interconnect each of the machines separately to one of the WAN connections or the switch/router.
The demand for network media conversion has grown with the increased use of long haul optical and wireless (RF) links. It is not uncommon to have specific devices in the data center rack to convert from one type to another. One of the many benefits of MetaConnect is that it also functions as a media converter with next to no latency or jitter and no impact on bandwidth. For example, MetaConnect converts 10GBASE-LR to 10G direct attached copper, 10G-BASE-LR to 10GBASE-SR, 10G direct attached copper to 10GBASE-SR and so on (at the same line rate). It also allows you to interconnect devices from different manufacturers who require that you use their branded SFP/SFP+ transceivers.
In this example, the MetaConnect allows less expensive direct attach copper SFP/SFP+ to be interconnected to third party long haul SFP/SFP+ or optical SFP/SFP+ transceivers.
MetaConnect can be used to as part of a network failover strategy via the management interface, and in a future software release, MetaConnect may be programmed to automatically reconfigure the interconnects upon detection of link failure or high error rates.
In this example, two WAN connections are connected to a MetaConnect device, which allows for failover between the two WANs. WAN A can be directly connected to the router and if a failure is detected, WAN A can be disconnected from the router and WAN B substituted in its place. There is neglible latency impact to this and there are gains in the form of signal integrity circuitry, diagnostics and signal monitoring abilities.
Another example of a failover use case is for an internally provided service such as a risk checking gateway.
Prior to a failover event, the MetaConnect might be connecting between a trader and the primary gateway on the client side, and between the primary gateway and an exchange on the exchange side. If a failure is detected, the MetaConnect can be reconfigured to connect the backup gateway in its place, avoiding the need for four hops through a traditional switch and therefore substantially reducing the round-trip latency.
Circuit Breaker / Kill Switch
There is growing requirement for circuit breakers to be installed between traders and the market to handle out-of-control trading software. However, any latency sensitive trader will be concerned that intermediate devices and software will disadvantage their trading. The benefit of MetaConnect is that it can function as a circuit breaker without you noticing that it is there, while it performs all its other functions.
Orders can be tapped by MetaConnect on the way to the market and routed to the broker's margin platform. Likewise, bookings, fills and executions can be tapped on the return path and also routed to the broker. If a limit is exceeded the broker platform can reconfigure the MetaConnect to shutdown the trader's uplink access while leaving the downlink open for acknowledgements from the market (depending on the type of network protocol).
The traditional approach to broadcasting data is to use a multicast switch or regenerative in-line network taps. Neither approach scales very well with increased port counts with respect to performance and manageability. In contrast, MetaConnect can replicate an input to all of its ports in an extremely low 4ns. Each signal is regenerated to avoid degradation in signal quality. With two layers of cascading 48 port MetaConnects it is possible to scale out to over 2000 feed copies within 16 nanoseconds (including fibers).
Here is a trading example: a router is connecting through MetaConnect to a WAN from which it can receive routing information (BGP), multicast subcription (PIM) and other negotiations. Any downstream packets from the WAN to the router will be replicated and sent to each of the trader machines, with a latency between the WAN and the Trader machines of less than 4ns. Orders can be placed via a second network port on each trading machine.
MetaConnect does away with the need to have separate network tap devices in your rack to monitor or broadcast communications. The problems with separate tapping devices are that they are disruptive to install and manage and they do not scale. Non-regenerative taps can result in signal attenuation and both types may introduce a point of failure. MetaConnect's fabric uses a crosspoint switch to dynamically configure the interconnects between inputs and outputs. You can go from tapping a single input for monitoring purpose through to tapping to all ports to fully replicate and broadcast your data. You get full signal regeneration on input and output at just 4ns latency combined with the reliability of redundant power supplies and fans. More details are provided in our explanations of scalable broadcast, circuit breakers and timestamping.
If you do a lot of tapping, make sure you check out our blog article regarding the cost savings and performance improvement that is possible using MetaConnect.
MetaConnect has the ability to precisely timestamp packets using an internal clock synchronized to a PPS signal via a connector located on the front panel. The timestamper replaces the packet's frame-check-sequence with a 32-bit value representing the least significant bits of a timestamp in nanoseconds.
Timestamping adds around 300ns to the latency, however, it is possible to configure MetaConnect to have a 4ns 'fast path' and a separately tapped timestamped path. The timestamped path will be slower but the timestamps themselves are allocated up front based on when the packet is received. This two-path approach internally uses an additional port which means that 3 ports are needed (1 fast, 1 slow, 1 for internal use).
In the diagram below, the trader is bi-directionally connected directly to the WAN. A tap is taken off each direction inside MetaConnect and fed to the internal timestampers. The timestamps are synchronised to a PPS signal via the connector on the front panel.
Meticulous testing is a task which is often overlooked because it is too tedious and requires too much discipline. Ideally, latency tests should be run on every software change to immediately detect any regressions before deployment.
A group of test machines and test equipment might need to be connected in arbitrary ways for a number of different tests or types of tests - a kind of time-share for the lab. Monitoring and logging hardware might also need to be connected at different points in the test harness in order to run these. Different network cards might need to be connected in order to test the different models. Changing the configuration for different tests by hand is so tedious that it tends not to happen.
MetaConnect solves this problem by acting as a reconfigurable patch panel.
All of the different ports in a suite of test equipment can be connected to a MetaConnect. Tests can be triggered through third-party continuous integration packages such as Atlassian's Bamboo or Jetbrains' TeamCity. When a test is to be run, the MetaConnect can be configured in a pre-defined way, in order to connect the network for the test. The machines and other test equipment are set up remotely by the script. The test will be run, and any data logging will occur. The test might be run multiple times with different taps of different points in the network logged. Once the data has been gathered it can be post processed in order to obtain latency statistics, and the data logged in a database, or further post-processing done in order to flag errors in the continuous integration package.
Multiplexing Connections (Order entry)
Multiple connections can be funneled into a single connection in around 69 ns using MetaMux, assuming there is no congestion. Otherwise, packets will be queued. Packet ordering is strictly FIFO (first-in first-out) with no starvation if a link is saturated.
In this trading example, multiple trading clients are multiplexed into a single link. The downstream direction can be configured to use MetaMux's layer 1 matrix switching in order to deliver 5 ns in that direction. Therefore, the combined latency through the switch is a little over 74 ns compared to a few hundred nanoseconds in each direction for a conventional switch.
Network monitoring forms a core part of the network infrastructure for many financial services firms. It's used for latency monitoring, audit, risk analytics, and more. MetaMux can provide both tapping and aggregation in a single device, with all of the same tapping advantages discussed in the MetaConnect tapping use case. MetaMux can both tap and aggregate up to 16 bi-directional connections in a single device, with an overhead of only 4 ns, while simultaneously significantly improving the quality of the signals, providing a re-patching capability, providing network insights, and all at a lower cost than traditional passive optical taps.