What grid of distributed computing platforms currently uses the services of trading systems

Question

There seem to be many grid computing infrastructures, but which of them are heavily used by investment banks for distribution purposes with low distribution latency? I would be interested to hear answers regarding both Windows, Linux, and cross-platform. Also, which RPC mechanisms seem to be the most preferred?

I heard that due to the low latency and speed, the calculations themselves are often written in C ++ / C, since the calculations performed on virtual machines are several orders of magnitude slower than the native code. Does this look like a normal scenario in practice? for example, a distributed .NET grid Framework that performs calculations written in C ++ / c?

Answer 1

Some areas (actually used in some corporate investment banks):

• PC Home Solutions
  farms (traders queue up requests for calculations)
• GPU

since computationally intensive intensive operations (for example, pricing in Monte Carlo) are usually highly parallelized.

Answer 2

low latency distribution calculation

"Low latency" and "distributed" are mutually exclusive :)

But, saying this, it depends on how low you mean low latency. If you are talking about high-frequency trading (HFT), then any implementation will use the fastest network code that they can access - most likely, custom TCP / IP stacks (for example, OpenOnload, native infiniband, etc.). The network will always be the slowest part of your code, so you need to minimize the network.

If you say fast, but not HFT-fast (for example, exotic price options / structured products), you can pretty much use whatever you like. I worked on systems that used any of .Net / RPC, JMS (ActiveMQ), TCP / IP sockets, etc. It's more about the flexibility and simplicity with which you can define and send data, rather than the raw speed of the networks.

Answer 3

The G-WAN server begins to understand this area. It mixes the speed (and area) of ANSI C with (fully compatible ANSI C) scripts (allows you to edit / update in real time without stopping the server).

Full JSON RPC is expected by the end of this year in the same spirit of efficiency. With it, you can implement distributed computing.

It is crazy to understand that they managed to install in 200 KB (server + script engine + many functions such as diagrams, compression, cryptography, etc.).

Answer 4

Check out www.zircomp.com. The zNet C ++ structure is used in several electronic trading systems. It is a cross-platform platform with many and distributed cores, based on a data-driven architecture, specially tuned for high performance using its own OS, with an intuitive API that uniformly supports data and the parallelism task.