EZ-Megabuf/Sequential

Sequential PreFetch Option

• Full-Track or Cylinder Reads - Data is read into Sequential PreFetch buffer using fulltrack or cylinder EXCPs. This can result in enormous reductions in I/O and elapsed run-times for sequential jobs.
  
• Combined Random and Sequential Processing - By using separate buffers for the sequential PreFetch,
  DB-Megabuf avoids flooding the DMCL buffers with data, thereby minimizing the impact on normal random
  access processes. DB-Megabuf PreFetch option is the complete solution to users processing data in random and sequential patterns.
• Dynamic PreFetch buffer storage management - For each AREA or FILE in a given DMCL, the user
  may specify whether Sequential PreFetch is active, or may be automatically activated. If automatic activation is requested, the PreFetch buffers are only allocated when DB-Megabuf detects that the AREA/FILE is being processed sequentially.
• Multiple PreFetch Buffers - Any number of FILES or AREAS may have PreFetch buffers assigned. The user specifies the number and size of the extents to be used in each buffer. An optional parameter allows the user to control the maximum amount of storage to be used for PreFetch buffers.
• Database in Memory - The user may specify that an entire AREA or FILE is to be read into a PreFetch buffer. DB-Megabuf will allocate enough space to hold the entire AREA or FILE in memory.
• Full-Track Writes - The user may specify that full-track writes are to be used for jobs updating files in sequential order. Note: This option is only available when running in local mode.
• Backwards Sequential Processing - Sequential PreFetch supports sequential processing for files being processed in reverse sequential mode, thus providing performance benefit to database loads, IDMSDBL2 and any other process that reads backwards through a database file.
• Reduced I/O and Elapsed times - By using Sequential PreFetch and Full Track/Cylinder Writes, the elapsed times for sequential processing runs can be reduced by up to 90%. This benefit can be
  achieved by user programs or CA-IDMS utilities.

MVS/ESA Data space Support

Assign Entire Files to data spaces. DB-Megabuf users may request that particular CA-IDMS files are to be loaded into ESA data spaces. A data space is a "data-only" address space which may contain up to 2-gigabytes of data. Support for up to 256 separate data spaces is provided.

• Update Areas in Data spaces - Data spaces may be allocated to both retrieval and update areas. When used for update areas, any modified pages are also updated in the data space.
• Shared Data spaces - When used in conjunction with DB-Share®, the data spaces may be shared by up to 32 concurrently executing CVs and local batch jobs. For example, if a particular file has been loaded to a data space by one CV, then other CVs or local mode batch jobs may reference that file in the data space with ZERO READ I/O. When used in conjunction with DB-Synchro©, updates to a file allocated to a data space are reflected in the data space, even where the update is performed by another CV or local mode batch job.

DB-Megabuf Standard (CA-IDMS 10.2 only)

CA-IDMS already has a highly effective and efficient buffer management technique resulting from many years of experience
and development. Its only short-coming (in release 10.2) is that its buffers must be located below-the-line, consuming valuable
space and restricting the maximum buffer sizes which may be assigned.

DB-Megabuf takes advantage of this technology without the short-comings. It allows CA-IDMS to perform its own buffer
management with buffers above-the-line.

• Virtual Storage Relief - One of the largest consumers of below-the-line storage is database buffers. With DB-Megabuf, users may allocate database buffers above--the-line for BDAM access, for both Central Version and local mode batch processing.
  DB-Megabuf provides maximum virtual storage relief by assigning all buffer related control blocks above-the-line. eg. BMEs, BASH storage, not just buffers.
• Large Buffer Pools - By allocating buffers above-the-line users may exploit the full XA addressing range (up to 2 gigabytes) for buffer storage. This can greatly increase buffer hits and hence reduce I/O.
• EXCP I/O - DB-Megabuf intercepts the IDMS BDAM I/O requests and instead issues highly efficient EXCP requests DIRECTLY INTO THE XA BUFFERS.
• Central Version and Local Mode - DB-Megabuf also allows allocation of XA buffers for local mode batch processing. This may significantly reduce elapsed job times in some circumstances. Where local mode batch processing is random, rather than
  sequential in nature, DB-Megabuf has been shown to vastly outperform other caching or Read-Ahead type processes