Using SNA chaining to segment large messages (TYPETERM RECEIVESIZE, BUILDCHAIN, and SENDSIZE)

Systems Network Architecture (SNA) allows terminal messages to be chained, and lets large messages be split into smaller parts while still logically treating the multiple message as a single message.

Input chain size and characteristics are normally dictated by the hardware requirements of the terminal in question, and so the CEDA BUILDCHAIN and RECEIVESIZE attributes have default values which depend on device attributes. The size of an output chain is specified by the CEDA SENDSIZE attribute.

Effects

Because the network control program (NCP) also segments messages into 256-byte blocks for normal LU Type 0, 1, 2, and 3 devices, a SENDSIZE value of zero eliminates the overhead of output chaining. A value of 0 or 1536 is required for local devices of this type.

If you specify the CEDA SENDSIZE attribute for intersystem communication (ISC) sessions, this must match the CEDA RECEIVESIZE attribute in the other system. The CEDA SENDSIZE attribute or TCT BUFFER operand controls the size of the SNA element that is to be sent, and the CEDA RECEIVESIZEs need to match so that there is a corresponding buffer of the same size able to receive the element.

If you specify BUILDCHAIN(YES), CICS® assembles a complete chain of elements before passing them to an application. If you do not specify BUILDCHAIN(YES), each individual RU is passed to an individual receive-any in the application. With SNA/3270 BMS does not work correctly if you do not specify BUILDCHAIN(YES).

If you are dealing with very large inbound elements that exceed a maximum of 32KB, you cannot use the BUILDCHAIN attribute or CHNASSY operand. You must use multiple individual RUs, and this extends the transaction life in the system.

Where useful

Chaining can be used in systems that use VTAM® and SNA terminals of types that tolerate chaining.

Limitations

If you specify a low CEDA SENDSIZE value, this causes additional processing and real and virtual storage to be used to break the single logical message into multiple parts.

Chaining may be required for some terminal devices. Output chaining can cause flickering on display screens, which can annoy users. Chaining also causes additional I/O overhead between VTAM and the NCP by requiring additional VTAM subtasks and STARTIO operations. This additional overhead is eliminated with applicable ACF/VTAM releases by making use of the large message performance enhancement option (LMPEO).

Recommendations

The CEDA RECEIVESIZE value for IBM® 3274-connected display terminals should be 1024; for IBM 3276-connected display terminals it should be 2048. These values give the best line characteristics while keeping processor usage to a minimum.

How implemented

Chaining characteristics are specified in the CEDA DEFINE TYPETERM statement with the SENDSIZE, BUILDCHAIN, and RECEIVESIZE attributes.

How monitored

Use of chaining and chain size can be determined by examining a VTAM trace. You can also use the CICS internal and auxiliary trace facilities, in which the VIO ZCP trace shows the chain elements. Some of the network monitor tools such as NetView® Performance Monitor (NPM) give this data.

Related tasks
Networking and VTAM: improving performance
Setting the size of the terminal input/output area (TYPETERM IOAREALEN or TCT TIOAL)
Setting the size of receive-any input areas (RAMAX)
Setting the size of the receive-any pool (RAPOOL)
Using the MVS high performance option (HPO) with VTAM
Adjusting the number of transmissions in SNA transaction flows (MSGINTEG, and ONEWTE)
Limiting the number of concurrent logon/logoff requests (OPNDLIM)
Adjusting the terminal scan delay (ICVTSD)
Compressing output terminal data streams
Tuning automatic installation of terminals
[[ Contents Previous Page | Next Page Index ]]