Question:
hi,guys
I want to do a remote client copy from a AIX production client to a NT demo client. The production is not very large because of the short bis running time.
But now this action have spent about 30 hours, and only more than 8000 tables are copyed. The total table number is more than 17000.
After study the two system, I found the bottlenock is the source system with 4cpu/4G mem.
The io of the 7133 array(raid5) is very busy, alway greater than 90%.
why?
there is only read action, but there still many write action at the same time. Before client copy, I have locked all the users on the source system.
tks
Answer:
how did u plan your datafiles, redo log files and arch?
_________________
The heart has its reasons which reason knows not of
PASCAL
Answer:
Hi,
What's the Network Bandwith between your two systems..
Also on the the NT machine have you deleted the client first or are you doing an delete recreate.
How many parallel processes are you using..
Mark
Answer:
In addition to Mark Norman's query on network bandwith, you may check these:
During client copy, not only the production system be "unavailable" to all non-basis users, but batchjobs should also be disabled. Why is there a write operation going on on that arraY?
8000 tables copied and 17000 pending could be misleading. Many of these 17000 tables could be very small (some get copied in 10 seconds!).
How many (total) batchjobs have got configured in target (Windows) SAP system? I vaguely recall, 2 batchprocess per cpu on *target* system can be used (in client copy, you have option to use "number of batchjobs to be used in parallel" or somesuch).
Last I did remote client copy (90Gig production), it took about 48 hours.
Trust this helps.
Cheers,
BT
Answer:
The two machine are in the same LAN segment, and there is no network problem.
there is no btc job on both system, I have canceled all the btc job.
all datafiles are in one raid5 array
all logfile are in another raid5 array
From the output of aix command filemon, the IO of SAPDATA1-6 are very busy.
I am strange for why there is write action too.
I know the production is very big, but this production is not, because it have run for only several months.
there is a strange phenomena:
I use sm50 to see the processes, when the action is direct read, it is very slow, but when the action is sequential read, it is very fast.
and the action will be direct read for several hours and then be sequential read for one or two hours.
To my knowledge, the problem should on the source machine, but not the target system.
tks all
yours
lhill
Answer:
know the production is very big, but this production is not, because it have run for only several months
Assuming your source system is Oracle, did you update Oracle Optimizer statistics before starting Remote Client Copy. Your mention of "direct read, it is very slow, but when the action is sequential read, it is very fast" could be because Oracle is doing full tablescans.
Cheers,
BT
Answer:
also for aix, use nmon, it can even monitor your san disks.
just to check, but your network adapter should be fibre channel, full duplex and 1000mbps right?
_________________
The heart has its reasons which reason knows not of
PASCAL
Answer:
tks for all replys
I have solve the problem by restart the source system.
I think the problem is that there is something wrong with the database of the source system.
But I didn't find the reason.
Fortunately, the remote client copy still go on after the restart.
the database is DB2. statistics is updated