In the coming weeks I am going to setup an ALPHA 4100.
This machine will be used as a query-machine.
Our DBA says there will be a lot full table-scans
We use ADVFS, LSM, HZ40 (RAID5)
I am thinking to implement the following configuration:
1 LSM stripeset 80GB ---> 4 * RAID5 sets ( 24 * RZ29B )
I know that ADVFS has a maximum physical I/O size off 64K (chvol)
In the HSZ40 configuration manual they say when there is much sequential
I/O
you must choose a chunk size smaller the maximum I/O size.
In that case there are more disks replying to 1 I/O request.
- HSZ40 chunk size RAID5 sets: 8K
- # volassist make <volume> 80GB layout=stripe stwidth=64K db-vol1
db-vol2 db-vol3 db-vol4
# mkfdmn <device> data1
Q1: should I use ADVFS striping instead off LSM striping ?
Q2: If LSM striping is ok, should I use a different stripe-width ?
Q3: should I use a different chunk-size ?
Our DBA thinks to set the following parameers in the init.ora:
+++++++++
db_block_size =16 kb
db_file_multiblock_read_count = 8
+++++++++
( Assuming a stripe size off 256 kb )
stripe size=2 x db_block_size x db_file_multiblock_read_count
Q4: Is the stripesize the stripe-width * number of members ??
Q5: Any more comment on my configuration ?
========================================================
"Creative minds always have been known to survive any kind of bad
training."
-- Anna Freud
=========================================================================
==
regards _at_vB
Received on Wed Jan 21 1998 - 11:23:37 NZDT