Fast/Cram

The Fast/Cram program product is a full function replacement for the MVS and MVS/ESA versions of the Cross Region Access Method (CRAM) SVC distributed with the Model 204 database management system. It provides dramatically improved support for communications between Model 204 and teleprocessing monitors, such as CICS or TSO. It also provides support for Advantage SQL and Host Language Interface (IFAM2) programs executed in batch, as well as the BATCH2 utility program.

Fast/Cram consumes considerably less system resources than CRAM. Each call is between 10 and 200 times faster when compared to CRAM. Total CPU savings range from 20% to 70% depending on the application. CPU savings of 50% are often reported for CICS users. A SAGE user experienced a 50% reduction in one four-hour daily SAGE job. Similar improvements might be achieved in heavy-hitting COBOL jobs using IFAM2 or Advantage SQL.

Fast/Cram is more secure and more reliable than the version of CRAM distributed with Model 204. It can work alongside that, but being plug-compatible it can replace it to provide maximum performance for your CRAM processing needs.

Technical Background

Some may wonder about the technical underpinnings of the product, which make it not only compatible but even superior in several ways to the Model 204-provided CRAM.

Fast/Cram doesn't require subtasks. This removes the single-threading of CRAM OPEN and CLOSE and dramatically reduces the overhead encountered by CICS when many CRAM connections with Model 204 are active.

Fast/Cram makes extensive use of SRB routines. These routines take full advantage of multiple-processor complexes such as the IBM 3090.

Fast/Cram is totally compatible with MVS and MVS/ESA. All storage is allocated "above the line". 31-bit application programs are fully supported.

Fast/Cram is reliable. Extensive use of MVS recovery facilities ensures that Fast/Cram has no "hanging states" or integrity exposures. Fast/Cram to the installation responsibilities for MVS integrity as outlined by IBM. All work areas are in key zero storage; data buffers are fetch protected; and all user-provided parameters are validated.

Storage Usage

Fast/Cram allocates storage in three distinct areas: CSA, SQA, and the private area. All of this storage is "above the line" in MVS/XA and MVS/ESA systems.

SQA

Fast/Cram allocates SQA space only when Model 204 is initialized. The amount of SQA storage allocated for a Model 204 online depends on parameters specified for the job. Each Model 204 online may establish up to three CRAM channels. One channel is for any host language threads (IODEV=23); another channel is for any full-screen remote threads (IODEV=11); and a third channel is for any non-full screen remote threads (IODEV=29). SQA storage for each CRAM channel is allocate din a contiguous block when the channel is opened, and freed when the channel is closed. Channel open generally corresponds to Model 204 initialization, while channel close generally corresponds to Model 204 termination. Essentially, the SQA usage is 560 bytes per possible user.

CSA

Fast/Cram allocates CSA space for data buffers. This space is fetch-protected, unlike the space allocated by CRAM. This provides increased data security by preventing unauthorized users from examining Model 204 data streams which may contain passwords and other sensitive data.

Fast/Cram data buffers are not allocated when a channel is opened. Instead, they are allocated the first time that a particular connection is used. Fast/Cram data buffers are only freed when the channel is closed, typically corresponding to Model 204 termination. Thus, the number of Fast/Cram data buffers allocated is determined by the high water mark for the number of concurrent users. CRAM allocates data buffers the same way, and they are the same size.

Private Area

Each Fast/Cram private work area is 136 bytes long, and is allocated from subpool 252 with key zero. One work area is allocated in the Model 204 region for each Fast/Cram channel open. This work area is freed when the Fast/Cram channel is closed. No other private area storage is allocated or freed by Fast/Cram.

One Fast/Cram work area is allocate din the application region each time a cram connection is established. This area is freed when the connection is broken. Individual Fast/Cram data transfer calls do not allocate or free storage.

Coexistence Considerations

Fast/Cram may be installed and used in parallel with CCA's CRAM. However, successful coexistence requires that the correct operational procedures be carefully followed. Simplicity agues fore just one version of CRAM. Using two versions of CRAM will require two separate SVC numbers and two separate subsystem names.

Installation Requirements

Fast/Cram executes within the IBM operating systems: MVS/SP Version 1 (MVS/370), MVS/SP Version 2 (MVS/XA), and MVS/ESA Version 3 (MVS/ESA). One set of modules supports all of these system types.

Fast/Cram comprises a single SVC load module and an interface module. The SVC load module may be installed as a Type 2, Type 3, or Type 4 SVC. The interface module may be installed in the Link Pack Area (LPA), or a private library, and need no be authorized.

Fast/Cram requires a subsystem name, reserved through an entry in an IEFSSNxx member of SYS1.PARMLIB. Both Fast/Cram modules are totally re-entrant and refreshable and do not need to be fixed in real storage. The SVC load module always executes in 31-bit addressing mode under MVS/XA and MVS/ESA. The interface load module executes in the addressing mode of its callers.

If your applications or software mix makes heavy use of the Model 204 cross region communications facilities you should seriously consider testing Fast/Cram.