The Adabas Fastpath command analysis sampler (AFPLOOK) can be used to determine where the best results may be expected from Adabas Fastpath by reporting on the command constructs that qualify for Adabas Fastpath.
The sampler can be controlled and viewed online using SYSAFP and, in addition, prints the results at normal database termination.
Note:
For those Adabas customers who do not have an Adabas Fastpath license,
the AFPLOOK sampler along with a demo version of SYSAFP is distributed as part
of the Adabas release tape. In this form, by default, AFPLOOK will sample a
full database session.
AFPLOOK is enabled using the following ADARUN command:
ADARUN FASTPATH=YESAFPLOOK is set up with certain operational defaults that control the amount of memory used during command analysis by restricting
the maximum number of files sampled; and
the number of concurrent users.
If any parameter is exceeded, AFPLOOK tries to ignore the excess while still reporting maximum information. In this way, AFPLOOK audits a general sampling of the database command workload to determine Fastpath optimization parameters. The operational defaults can be modified for site requirements as described in the section APLOOK Parameters.
This section describes the AFPLOOK parameters, which are used to define the boundaries of the sample and limit the amount of memory required.
The maximum number of files to be sampled.
Once the maximum number of files is put in the analysis table, no additional files are sampled; however, additional files show in a command count so that it can be determined whether or not this parameter should be increased for subsequent executions.
Default: 64
The maximum number of command/descriptor entries per file.
In conjunction with the Maximum Files parameter,
this parameter restricts the amount of memory used. If the maximum entries is
reached for a file, the last entry is converted into a general accumulator.
Note that only one entry is required for the Adabas command types
L1,L2,S8, and S9.
Default: 32
The maximum size for the table of concurrent users.
If all the user areas are being used at one time, a new request is satisfied by releasing the 'oldest usage' user area. The number of times this reusage occurs is noted and printed in the summary. When a sample contains a high percentage of reusage, this parameter should be adjusted.
Default: 100
The maximum concurrent Adabas Command IDs (CIDs) sampled for each user processed.
In conjunction with the Maximum Concurrent Users parameter, this parameter restricts the amount of memory used. Commands for Command IDs that exceed this maximum are ignored and reported as rejected. When a sample contains a high percentage of rejections, this parameter should be adjusted.
Default: 10
The maximum number of commands to be sampled.
Default: No limit
Used to restrict sampling to a particular job name. One or more asterisks (*) can be used in the job name as a wild card character so that the sample can select all jobs that match the name ignoring the character positions occupied by an asterisk (*).
Default: None
Used to restrict sampling to specific files.
This option may be useful where the maximum files overflowed, or file activity is known and detailed analysis is required.
Default: All files
This section describes the types of information available on the AFPLOOK report:
This section of the report provides a summary of the file commands.
--- -- ---- ---------- -- ---------- ---------
FNR CC DESC DIRECT ACC RC SEQUENTIAL SEQUENCES
--- -- ---- ---------- -- ---------- ---------
20 L1 -- 1
L2 -- 4 4
L3 CC 1
L9 AA 1
L9 BB 2
L9 CC 2 1 1
S1 AA 3 1
TOTALS 7 8 21(18%)
EXCLUDED COMMANDS: 2
ALREADY PREFETCHED: 3
(UPDATES 2,INSERTS 1,DELETES 1) (MAX.RBL DA 0,SEQ 32)
------------------------------------------------------------------- |
| Column | Explanation |
|---|---|
| FNR | Adabas file number. |
| CC | Adabas command code. |
| DESC | Primary descriptor. |
| DIRECT ACC | Maximum number of direct access commands that can be optimized. |
| RC | Maximum number of RC commands that
can be optimized.
|
| SEQUENTIAL | Maximum number of sequence commands that can be optimized. |
| SEQUENCES | Number of sequences that caused the number of sequential commands. The sequence factor for optimization may be calculated from these two numbers. |
The rightmost number shows the total sampled commands for the file together with the percentage relative to all the sampled file commands. On a large report, this number can be used to determine quickly which files should be considered for optimization.
Commands that have been ignored for the file are also listed along with the reason for exclusion.
The final line shows the update commands as well as the maximum record buffer lengths found for direct access or sequential commands that can be optimized.
This section of the report summarizes the total commands sampled for all files and expresses this as a percentage of all commands seen. Excluded commands are similarly reported.
------------------------ POTENTIAL OPTIMIZATION SUMMARY -------------------
SAMPLED COMMANDS MAXIMUM OPTIMIZATION
SAMPLED FILE COMMANDS 116 ( 77%) <------ SEQUENTIAL: 55 (47%)
DIRECT ACCESS: 32 (27%)
RCS: 4 ( 3%)
EXCLUDED COMMANDS 33 ( 22%)
TOTALS 149 (100%) 91 (61%)
------------------------------------------------------------------------------- |
The maximum optimization numbers are an estimation of potential
optimization. The sequential commands, direct access, and
RC totals are expressed as a percentage of the total
sampled file commands. The total is expressed as a percentage of all
commands.
These numbers indicate the estimated potential optimization using Adabas Fastpath. The actual optimization will depend on various factors unique to each user site. Contact Software AG for assistance when interpreting samples.
This section of the report provides command analysis information.
---------------------------------------------------------------------
COMMAND ANALYSIS
REJECTED COMMANDS
MAX. USERS EXCEEDED: 0
MAX. CIDS EXCEEDED: 0
MAX. FILES EXCEEDED: 0 0 ( 0%)
EXCLUDED COMMANDS
BAD COMMANDS: 4
NON-FILE COMMANDS: 7
NON-FILE RCS: 2
EXCLUDED FILE COMMANDS: 8
UPDATE COMMANDS: 4
ALREADY PREFETCHED: 8 33 ( 22%)
SAMPLED FILE COMMANDS 116 ( 77%)
ALL COMMANDS SEEN 149 (100%)
---------------------------------------------------------------------- |
The numbers shown:
illustrate the type of commands processed, and
put the previous section into perspective.
Rejected commands are categorized by users, CIDs, and files exceeded. If the total percentage is high, estimates reported elsewhere may not give an accurate assessment.
Excluded commands are split into the following categories:
| Category | Explanation |
|---|---|
| Bad commands | Unexpected Adabas response codes. |
| Non-file commands | Commands that cannot be attributed to a file; for example,
OP,CL,ET,C1,RE. Plus file commands
HI,LF,RI.
|
| Non-file RCs | All RC commands plus any
RC for which the CID is not stored by AFPLOOK.
|
| Excluded file commands | L4,L5,L6,S4,S5.
|
| Update commands | A1,A4,E1,E4,N1.
|
| Already prefetched | Any command that could qualify for sequential optimization that has prefetch or multifetch already set. |
This section of the report
shows the important parameters used to produce the report; and
gives an indication of the parameters needed.
------------------------------------------------------------------
PARAMETERS USED
MAX. FILES: 64 FILES NEEDED: 5
..MAX. DE: 32 OVERFLOWS: 0
MAX. USERS: 100 HIGH USERS: 15
..MAX.CID: 10 HIGH CIDS: 4
* REUSED USER AREA OCCURRENCES: 0
MAX.RECORDS: NO LIMIT
------------------------------------------------------------------ |
