Hi there,
first of all sorry for the long append but I need to explain my problem in
detail.
I was trying to optimize our backup performance for a particular node, so I got
a look at
INSTR_CLIENT_DETAIL trace output.
Server is ADSM 3.1.2.50 on OS/390 V2R6
Client is USS 3.1.0.5
Commethod is TCPIP using a CTC (Channel To Channel) connection
Client and Server are on two separate LPAR on the same 2064 machine
Here is our starting point:
Total number of objects inspected: 52,041
Total number of objects backed up: 1,554
Total number of objects updated: 0
Total number of objects rebound: 0
Total number of objects deleted: 72
Total number of objects failed: 0
Total number of bytes transferred: 8.10 GB
Data transfer time: 256.23 sec
Network data transfer rate: 33,168.89 KB/sec
Aggregate data transfer rate: 1,396.01 KB/sec
Objects compressed by: 55%
Elapsed processing time: 01:41:28
and trace output:
Elapsed time: 6088.402 sec
Section Total Time(sec) Average Time(msec) Frequency used
-----------------------------------------------------------------
Client Setup 0.063 62.9 1
Process Dirs 108.029 47.1 2292
Solve Tree 0.000 0.0 0
Compute 2.003 0.0 768476
Transaction 46.226 0.0 2312991
BeginTxn Verb 0.007 0.0 1181
File I/O 2659.223 4.6 580646
Compression 2964.974 4.3 692721
Data Verb 255.722 1.0 266454
Confirm Verb 0.202 4.9 41
EndTxn Verb 51.560 43.7 1181
Client Cleanup 0.392 392.2 1
So,according to common tuning consideration, I decide to turn client compression
off.
The results weer not as expected:
Total number of objects inspected: 57,067
Total number of objects backed up: 1,450
Total number of objects updated: 0
Total number of objects rebound: 0
Total number of objects deleted: 507
Total number of objects failed: 1
Total number of bytes transferred: 18.53 GB
Data transfer time: 6,941.87 sec
Network data transfer rate: 2,799.28 KB/sec
Aggregate data transfer rate: 1,858.10 KB/sec
Objects compressed by: 0%
Elapsed processing time: 02:54:18
and trace output:
Elapsed time: 10458.494 sec
Section Total Time(sec) Average Time(msec) Frequency used
------------------------------------------------------------------
Client Setup 0.766 765.9 1
Process Dirs 123.954 47.2 2625
Solve Tree 0.000 0.0 0
Compute 1.975 0.0 608333
Transaction 53.118 0.0 1831471
BeginTxn Verb 0.009 0.0 1179
File I/O 3239.495 5.3 609776
Compression 0.000 0.0 0
Data Verb 6940.654 11.4 608333
Confirm Verb 38.818 825.9 47
EndTxn Verb 59.321 50.3 1179
Client Cleanup 0.385 384.6 1
My question is; why Data Verb is so increased while data to be transferred are
only doubled?
>From dsmsched.log analisys, I found this time gap in sending data from client:
02/22/2001 02:28:57 Retry £ 1 Normal File--> 204,832,768
/ciadata/Mail_Trace.nsf Sent
02/22/2001 03:52:59 Normal File--> 338,083,840 /ciadata/StatisticheUOW.nsf
Sent
No errors detected, neither on client nor on server side.
The only thing I found is that, at 02:24, two migrationprocesses started moving
data from the
dasd stg where the client is sending its data; at 03:50 the last process
terminated its migration.
Maybe this was only a coincidence but ... I decide to add some volumes to the
stg
trying to accomodate the new data and to avoid the dasd_to_tape migration
processes.
Again, these are the results:
Total number of objects inspected: 56,890
Total number of objects backed up: 1,446
Total number of objects updated: 2
Total number of objects rebound: 0
Total number of objects deleted: 22
Total number of objects failed: 1
Total number of bytes transferred: 18.08 GB
Data transfer time: 2,091.27 sec
Network data transfer rate: 9,068.03 KB/sec
Aggregate data transfer rate: 3,619.54 KB/sec
Objects compressed by: 0%
Elapsed processing time: 01:27:19
and trace output:
Elapsed time: 5239.620 sec
Section Total Time(sec) Average Time(msec) Frequency used
------------------------------------------------------------------
Client Setup 0.680 679.9 1
Process Dirs 122.582 46.7 2627
Solve Tree 0.000 0.0 0
Compute 1.913 0.0 593671
Transaction 44.557 0.0 1787459
BeginTxn Verb 0.009 0.0 1175
File I/O 2914.685 4.9 595100
Compression 0.000 0.0 0
Data Verb 2090.089 3.5 593671
Confirm Verb 0.653 15.9 41
EndTxn Verb 64.053 54.5 1175
Client Cleanup 0.398 397.8 1
My questions:
1) why is Data Verb 8 times bigger?
2) can (I don't thing so) dasd stg migration stop client operation?
Maybe it can degrade performance but not stop completely ...
Your help will be very appreciated.
Thanks in advance.
Best regards.
Valeriano Bassi [EMAIL PROTECTED]