Browse Prior Art Database

Communication BUFFER Reduction Technique

IP.com Disclosure Number: IPCOM000038612D
Original Publication Date: 1987-Feb-01
Included in the Prior Art Database: 2005-Jan-31
Document File: 3 page(s) / 16K

Publishing Venue

IBM

Related People

Jeffries, LM: AUTHOR [+4]

Abstract

A method is described which allocates fewer buffers to a communication buffer pool and while maintaining existing performance levels. The IBM PC 3270 Emulation Program is a software package which allows IBM Personal Computers to emulate IBM 3270 family of devices. Functions emulated include a subset of 3274 51C control unit, 3278 display station, and 3287 printer. Four configurations are possible. Except for the stand-alone configuration, the IBM PC Network hardware or Token Ring is also required for data distribution. In these configurations, one of the PCs on the network is designated as the "Gateway." A "Gateway" acts as a concentrator/distributor. On the one hand, "Gateway" communicates to the host either through a SDLC (Synchronous Data Link Control) or a TCA (Telecommunications Access) link.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 48% of the total text.

Page 1 of 3

Communication BUFFER Reduction Technique

A method is described which allocates fewer buffers to a communication buffer pool and while maintaining existing performance levels. The IBM PC 3270 Emulation Program is a software package which allows IBM Personal Computers to emulate IBM 3270 family of devices. Functions emulated include a subset of 3274 51C control unit, 3278 display station, and 3287 printer. Four configurations are possible.

Except for the stand-alone configuration, the IBM PC Network hardware or Token Ring is also required for data distribution. In these configurations, one of the PCs on the network is designated as the "Gateway." A "Gateway" acts as a concentrator/distributor. On the one hand, "Gateway" communicates to the host either through a SDLC (Synchronous Data Link Control) or a TCA (Telecommunications Access) link. On the other hand, the "Gateway" distributes data to and concentrates data from workstations on the IBM PC Network or Token Ring network. Thus, the "Gateway" performs the store and forward functions. And as such, it requires buffers as temporary storage. A "Gateway" can be configured to support, in increments of 8, from a minimum of 8 to a maximum of 32 SNA (System Network Archotecture) sessions. In the Release 1.0 of the IBM PC Network SNA 3270 Emulation, a "Gateway" is allocated 14 buffers for each session supported. The size of each buffer is 272 bytes or 4K bytes per session. To support 32 sessions, over 120K bytes of memory (448 communication buffers) are required for the communication buffer pool. Since the IBM PC 3270 Emulation runs in a multiple application environment, it is necessary to reduce its memory requirement to make room for other coexisting applications. In the IBM PC 3270 Emulation Program Version 2, there are three components in the "Gateway" that can make requests to get a buffer from the communication buffer pool: 1. The HOST INTERFACE code requests a buffer when data is re

ceived from the host computer (via either the SDLC or the TCA

link).

2. The LAN INTERFACE code requests a buffer when data is received

from the Network Station (via the IBM PC Network LAN).

3. The SNA code requests a buffer when it generates an SNA re

quest or response. In the Release 1.0 of the IBM PC Network SNA 3270 Emulation Program, the number of buffers allocated to the communication buffer pool is equivalent to the combined maximum number of buffers that could possibly be requested by these three components at one time. If the buffer pool is reduced by any amount, a problem arises as to how to handle the depleted communication buffer pool situation. Several solutions exist, but not all solutions maintain the performance level of a system with the maximum number of buffers allocated. The simple solution of trying again later when a buffer is not available does not suffice, since one component which uses buffers faster than another component could effectively starve out the slower component from gett...