Product Documentation

Use Case 2: Configuring Rule Based Persistence Based on a Name-Value Pair in a TCP Byte Stream

Jun 08, 2015

Some protocols transmit name-value pairs in a TCP byte stream. The protocol in the TCP byte stream in this example is the Financial Information eXchange (FIX) protocol. In its traditional, non-XML implementation, the FIX protocol enables two hosts communicating over a network to exchange business or trade-related information as a list of name-value pairs (called “FIX fields”). The field format is <tag>=<value><delimiter>. This traditional tag-value format makes the FIX protocol ideal for the use case.

The tag in a FIX field is a numeric identifier that indicates the meaning of the field. For example, the tag 35 indicates the message type. The value after the equal sign holds a specific meaning for the given tag and is associated with a data type. For example, a value of A for the tag 35 indicates that the message is a logon message. The delimiter is the nonprinting “Start of Header” (SOH) ASCII character (0x01), which is the caret symbol (^). Each field is also assigned a name. For example, the field with tag 35 is the msgType field. Following is an example of a logon message.

8=FIX.4.1 9=61 35=A 49=INVMGR 56=BRKR 34=1 52=20000426-12:05:06 98=0 108=30 10=157

Your choice of persistence type for a tag-value list such as the one shown above is determined by the options that are available to you for extracting a particular string from the list. Token-based persistence methods require you to specify the offset and length of the token that you want to extract from the payload. The FIX protocol does not allow you to do that, because the offset of a given field and the length of its value can vary from one message to another (depending on the message type, the preceding fields, and the lengths of the preceding values) and from one implementation to another (depending on whether custom fields have been defined). Such variations make it impossible to predict the exact offset of a given field or to specify the length of the value that is to be extracted as the token. In this case, therefore, rule based persistence is the preferred persistence type.

Assume that a virtual server fixlb1 is load balancing TCP connections to a farm of servers hosting instances of a FIX-enabled application, and that you want to configure persistence for connections on the basis of the value of the SenderCompID field, which identifies the firm sending the message. The tag for this FIX field is 49 (shown in the earlier logon message example).

To configure rule based persistence for the load balancing virtual server, set the persistence type for the load balancing virtual server to RULE and configure the rule parameter with an expression. The expression must be one that extracts the portion of the TCP payload in which you expect to find the SenderCompID field, typecasts the resulting string to a name-value list based on the delimiters, and then extracts the value of the SenderCompID field (tag 49), as follows:

set lb vserver fixlb1 -persistenceType RULE -rule "CLIENT.TCP.PAYLOAD(300).TYPECAST_NVLIST_T('=','^').VALUE(\"49\")"

Note: Backslash characters have been used in the expression because this is a CLI command. If you are using the configuration utility, do not enter the backslash characters.

If the client sends a FIX message that contains the name-value list in the earlier logon message example, the expression extracts the value INVMGR, and the NetScaler appliance creates a persistence session based on this value.

The argument to the PAYLOAD() function can be as large as you deem is necessary to include the SenderCompID field in the string extracted by the function. Optionally, you can use the SET_TEXT_MODE(IGNORECASE) function if you want the appliance to ignore case when extracting the value of the field, and the HASH function to create a persistence session based on a hash of the extracted value. The following expression uses the SET_TEXT_MODE(IGNORECASE) and HASH functions:


Following are more examples of rules that you can use to configure persistence for FIX connections (replace <tag> with the tag of the field whose value you want to extract):
  • To extract the value of any FIX field in the first 300 bytes of the TCP payload, you can use the expression CLIENT.TCP.PAYLOAD(300).BEFORE_STR("^").AFTER_STR("<tag>=").
  • To extract a string that is 20 bytes long at offset 80, cast the string to a name-value list, and then extract the value of the field that you want, use the expression CLIENT.TCP.PAYLOAD(100).SUBSTR(80,20).TYPECAST_NVLIST_T('=','^').VALUE("<tag>").
  • To extract the first 100 bytes of the TCP payload, cast the string to a name-value list, and extract the value of the third occurrence of the field that you want, use the expression CLIENT.TCP.PAYLOAD(100).TYPECAST_NVLIST_T('=','^').VALUE("<tag>",2).
    Note: If the second argument that is passed to the VALUE() function is n, the appliance extracts the value of the (n+1)th instance of the field because the count starts from zero (0).
Following are more examples of rules that you can use to configure persistence. Only the payload-based expressions can evaluate data being transmitted through the FIX protocol. The other expressions are more general expressions for configuring persistence based on lower networking protocols.