Some questions about FIXT1.1

, ,
1

Imported from previous forum

2

Hi all,

Got into some doubts while studying FIX :slight_smile:

  1. Do “Sequence Numbers” apply only to FIX application layer messages, or also to session administrative messages (as login, logout, heartbeat, …etc.)?

  2. Is it good idea / conventional design that, FIX Engine would be running on separate hardware, than the application of the FIX Engine, and these two would communicate by TCP/IP? This of course, is in addition to providing direct Application APIs.

  3. Would it be possible to obtain sample “FIX session” and few “FIX application” messages? I would like to test my own small fix engine :slight_smile:

  4. Would it be possible to obtain latency & max. number of simultaneous connections, and other stress criteria for some of best FIX engines in market?

Thanks :slight_smile:

Nick

3

Your questions and their answers apply to all versions of FIX. See my answers below each of your questions

  1. Do “Sequence Numbers” apply only to FIX application layer messages,
    or also to session administrative messages (as login, logout,
    heartbeat, …etc.)?

All messages have Sequence numbers, FIX engines use sequence numbers to find out if they have missed any inbound messages.

  1. Is it good idea / conventional design that, FIX Engine would be
    running on separate hardware, than the application of the FIX Engine,

In multiprocessor machines, FIX engine and trading application would run on the same machine on different processors. When a high hardware capacity machine is not available, FIX engine and trading application could run on seperate machines, but they must be close to each other in network terms to avoid network hops which would degrade performance.

and these two would communicate by TCP/IP? This of course, is in
addition to providing direct Application APIs.

There is no requirement that trading application and FIX engine communicate using TCP/IP though TCP/IP could be used.

In FIX.4.0 to FIX.4.4, TCP/IP was the only protocol used for communication between FIX engines (BuySide ↔ SellSide), but with FIXT.1.1 that restriction is removed and any reliable transport protocol could be used.

  1. Would it be possible to obtain sample “FIX session” and few
    “FIX application” messages? I would like to test my own small
    fix engine :slight_smile:

Scott Atwell has posted “Example FIX Messages” at

http://fixprotocol.org/documents/611/ExampleFixMessages.txt

  1. Would it be possible to obtain latency & max. number of simultaneous
    connections, and other stress criteria for some of best FIX engines
    in market?

You would have to contact each vendor to obtain their FIX engine’s performance numbers along with details of how the test was conducted to help comparision.