Overview

FIXEdge routes messages (e.g. orders, cancels) from source systems to destination systems and it is a common case that a destination system accepts messages at the time interval when a session is established. But in certain business configurations destination system may work as a broker that owns a gateway to exchanges, that are ultimate destinations for messages. Ultimate destination may be closed (work on schedule base) while a broker is running. In such cases, messages will be rejected by the broker. Thus there is a need to store them in a queue and release them at a specific time, according to an exchange schedule.   

To meet these needs Managed Queue solution provides a way: 

• To store messages in the queue during off-hours (the market is closed)
• To release messages stored in the queue at the time market is open
• To pass through messages during “market is open” hours
• To define multiple queues for different order types
• To distribute incoming orders among the queues relating to order type - to route only messages that can be accepted by a certain destination system.
• To apply smart logic for orders stored in the queue (cancel and cancel/replace)

Manage Queue feature description

Design

Logically FIXEdge server is a router that routes messages from a source system to a destination system. The following diagram demonstrates data flow for the case managed queue is used. Messages from a source system are parsed and put into the business layer. BL rule is executed and messages that fit the rule are moved to a certain virtual session. Each virtual session is actually a queue. If a virtual session is open (connected to a target session) the message is moved to the target session immediately. If a virtual session is closed (disconnected from a target session) then the message is stored in the queue. The smart logic is applied. The virtual session connects to a target session or disconnects from a target session according to a schedule configured for a virtual session. The target session may also have a schedule. If the target session is not established then messages stay in the queue and are released when the target session is established.

All messages are persistently stored in SQLite database.

SQLite database schema

SQLite database is used to store queued messages. Default DB name is fe.db. The filename and the path to it can be configured with property FixLayer.<managed_queue_name>.Storage.SQLite.FileName

Messages of all queues, sent to all sessions, are stored within a single table – “QueueStorage”. Messages that are delivered or removed are marked in corresponding columns.

Factual removal of messages from the table can be done manually or by a script using SQLite utilities.

Table “QueueStorage” is described below:

Database schema:

CREATE TABLE QueueStorage (GlobalMessageID INTEGER PRIMARY KEY, QueueID TEXT, MsgType TEXT, ClientID TEXT, ClOrdID TEXT, RawMessageText TEXT, ParserName TEXT, ApplVer INTEGER, TimestampEnqueued INTEGER, IsDone INTEGER, TimeStampDone INTEGER, IsRemoved INTEGER, TimestampRemoved INTEGER);

CREATE INDEX idx_QueueStorage_QueueID ON QueueStorage (QueueID, IsDone, IsRemoved);

CREATE INDEX idx_QueueStorage_ClOrdID ON QueueStorage (QueueID, ClientID, ClOrdID, IsDone, IsRemoved);

Queue releasing process

Queued messages are released when a virtual session is connected to a target session and a target session is established. Messages are released according to FIFO (First In First Out) logic. If some non-queued messages from the source system are passed to a target session simultaneously with queue release first go the queued messages and then other messages according to the FIFO principle.

Configuration

To use Managed Queue functionality user must configure the BL rule in BL_Config.xml and manage queue and session properties in FIXEdge.properties files.

Configuring BL

From the routing perspective, the BL rule for routing to a virtual session is similar to an ordinary BL routing configuration.

The example of BL rule for routing from RestAcceptorClient_Queueing source to MQDEST-XMIC-PreOpen virtual session:

<Rule Description="Route XMIC Limit orders to XMIC preopen (MQDEST) ">
      <Source>
            <Client Name="RestAcceptorClient_Queueing"></Client>
      </Source>
      <Condition>
            <MatchField Field="35" Value="D|F|G"></MatchField>
            <MatchField Field="207" Value="XMIC" ></MatchField>
            <MatchField Field="40" Value="2" ></MatchField>
       </Condition>
       <Action>
            <Send Name="MQDEST-XMIC-PreOpen"></Send>
       </Action>
  </Rule>

Configuring managed queue

Manages queue properties description:

Example

Managed Queue properties example:

FixLayer.ManagedQueue.Storage.Type = SQLite 
FixLayer.ManagedQueue.Storage.SQLite.FileName = fe.db

Configuring virtual session

The message is routed from the source connected to the virtual session.

Virtual session properties description:

Virtual session schedule properties description:

Virtual session properties example (source= RestAcceptorClient_Queueing, virtual session= MQDEST-XMIC-PreOpen):

FixLayer.ManagedQueue.Sessions = MQDEST-XMIC-PreOpen
...
FixLayer.ManagedQueue.Session.MQDEST-XMIC-PreOpen.TargetSession = MQTARGET
FixLayer.ManagedQueue.Session.MQDEST-XMIC-PreOpen.Schedule = MQTEST-XMIC-PRE

Virtual session schedule properties example:

FixLayer.ManagedQueue.Schedules.MQTEST-XMIC-PRE.TimeZone = America/New_York
FixLayer.ManagedQueue.Schedules.MQTEST-XMIC-PRE.ConnectTime = 0 20 18 * * *
FixLayer.ManagedQueue.Schedules.MQTEST-XMIC-PRE.DisconnectTime = 0 25 18 * * *

When the ConnectTime trigger is on a message is routed to the target session  MQTARGET.

Configuring target session

The target session is configured as an ordinary FIX session.

Target session properties example:

FixLayer.FixEngine.Sessions = MQTARGET
...
FixLayer.FixEngine.Session.MQTARGET.Version = FIX44
FixLayer.FixEngine.Session.MQTARGET.SenderCompID = FIXEDGE
FixLayer.FixEngine.Session.MQTARGET.TargetCompID = MQDEST
FixLayer.FixEngine.Session.MQTARGET.Role = Initiator
FixLayer.FixEngine.Session.MQTARGET.StorageType = persistent
FixLayer.FixEngine.Session.MQTARGET.IntradayLogoutTolerance = true
FixLayer.FixEngine.Session.MQTARGET.RecreateOnLogout = true
FixLayer.FixEngine.Session.MQTARGET.TerminateOnLogout = false
FixLayer.FixEngine.Session.MQTARGET.IgnoreSeqNumTooLowAtLogon = true
FixLayer.FixEngine.Session.MQTARGET.ForceSeqNumReset = 0
FixLayer.FixEngine.Session.MQTARGET.Schedule = MQTEST
FixLayer.FixEngine.Session.MQTARGET.HBI = 60
FixLayer.FixEngine.Session.MQTARGET.Host = 10.11.11.111
FixLayer.FixEngine.Session.MQTARGET.Port = 8901

Target session schedule example:

Schedules.MQTEST.StartTime = 0 0 18 * * *
Schedules.MQTEST.ConnectTime = 0 05 18 * * *
Schedules.MQTEST.DisconnectTime = 0 55 19 * * *
Schedules.MQTEST.TerminateTime = 0 59 19 * * *

Order flow scenarios

Order is queued in trade-off hours, released when the market is open

Order goes through without queueing

Cancel eliminates order in the queue

Cancel is queued (no corresponding order in the queue)

Cancel/replace is received, Cancel Reject generated, Order stays in the queue

Smart logic features

Smart Logic is a handler, that can play specialized logic for incoming messages, like canceling orders, sending notifications, and so on.

The FIXEdge package has an example of the smart logic plugin (located in plugins/libb2b_fixserver_plugin_queue_observer.so or plugins/b2b_fixserver_plugin_queue_observer.dll ).

The implemented features in this plugin:

• Matching incoming Cancel with queued Order, deleting both, and sending confirmation to BL.
• Cancel/Replace queued Order – send Cancel Reject message, the order stays in the queue

For details see order flow scenarios.

Logic in the module is hardcoded, but it can be replaced via the FIXEdge plugin system. Smart logic can be turned on/off for all sessions simultaneously.

Since the FIXEdge release 6.11.5, it is possible to enable/disable smart logic per FIX session.

A new parameter with the default value "false" is introduced

FixLayer.ManagedQueue.Session.<session name>.SmartOrderProcessing = true|false.

Troubleshooting

To access SQLite storage, use the sqlite3_analyzer program or other command-line tools for managing SQLite database files from the SQLite Download Page.

The SQLite database file location is configured with the parameter FixLayer.<managed_queue_name>.Storage.SQLite.FileName. 

How to list names of tables matching LIKE pattern TABLE

Inside SQLite, the Command-Line Shell runs the below command.

sqlite> .tables QueueStorage

How to show the structure of a table is to use the following PRAGMA command.

Inside SQLite, the Command-Line Shell runs the below command.

sqlite> .header on

sqlite> .mode column

sqlite> pragma table_info('QueueStorage');

Example of structure output of QueueStorage table

sqlite> pragma table_info('QueueStorage');
cid  name               type     notnull  dflt_value  pk
---  -----------------  -------  -------  ----------  --
0    GlobalMessageID    INTEGER  0                    1
1    QueueID            TEXT     0                    0
2    MsgType            TEXT     0                    0
3    ClientID           TEXT     0                    0
4    ClOrdID            TEXT     0                    0
5    RawMessageText     TEXT     0                    0  
6    ParserName         TEXT     0                    0
7    ApplVer            INTEGER  0                    0
8    TimestampEnqueued  INTEGER  0                    0
9    IsDone             INTEGER  0                    0
10   TimeStampDone      INTEGER  0                    0
11   IsRemoved          INTEGER  0                    0
12   TimestampRemoved   INTEGER  0                    0