| I DON'T THINK OF BIG AS THE NUMBER OF FOCUS FILES OR FEX'S BUT AS
THE TOTAL AMOUNT OF DISK SPACE IN BLOCKS USED BY A SYSTEM, AND THE
NUMBER OF SEGMENT INSTANCES IN FOCUS FILES. WHEN FOCUS FILES GET
BIG (50,000 TO 100'S OF THOUSANDS OF BLOCKS) THEN DESIGN ISSUES
BECOME THE CRITICAL FACTORS, WHETHER THERE'S 10 FEX'S OR 100 FEX'S
1 .FOC OR 20 . I'VE NEVER REALLY COUNTED THE FEX'S OR DISK BLOCKS
IN THE APPLICATIONS I'VE WORKED ON BUT I'LL GUESSTIMATE A COUPLE:
GENERAL LEDGER - 25-30 .FOC'S 250,000 BLOCKS 150 FEX'S
TRACE - 15 .FOC'S 200,0000 BLKS 50 FEX'S
MAYBE THIS SHOULD LEAD INTO A DISCUSSION OF DESIGN ISSUES .
WILL ANYONE TAKE IT FROM HERE??
RICHARD
MADDEN
|
| I agree with the fact that the number of FOCUS databases and Focexecs
do not indicate how large a system, but the size of those databases
and focexecs do.
We currently have a few large FOCUS databases. We have 20 - 25
in total and five a relatively large. Two are 200,000 blocks, one
is 325,000 and two are 400,000.
I am also aware of a group in Westford with FOCUS databases spanning
disk packs.
Of course, with the use of these large files come the FOCUS problems
never before encountered. A problem we ran into had to do with
indexing. For your information, with release 1.3 or lower, if your
index physically is located on a FOCUS page beyond 32,000 (use FDT
to find out), then when using the indexing for a value on those
pages will put you in a loop. Release, 1.31A fixed that problem
and we discovered a new one. If referencing an index value from
a page beyond 32,000 and it is the first value on the page, the
message, 'invalid index pointer' will be generated from FOCUS.
This problem is now fixed and will be in release 5.2. The way around
these problems is with location files. Just make sure that the
index page is not beyond 32,000.
Has anyone run into any other problems or defficiencies with using
large FOCUS databases? Has anyone converted any of these large
files to RDB?
Regards, Bob Girard.
|
| I agree with Richard and Bob - 'big' means database size - not appication
programs. When FOCUS databases become large (i.e. >100k blocks) then
design is everything. The VAX environment just makes you a lot more
aware of design issues: breaking data into several physical files, more
segments, unique segments, moving physical files onto more than one
disk to avoid contention problems, changing the disk cluster size,
watching your disk's fragmentation) - all radically effect performance.
There are plenty of large applications running in FOCUS. The Area Info
Center has databases per year which when concatenated with a USE
statement, produce a logical 600k block database that is suprisingly
efficient..
-rpr-
|