Re: Level 0 Archive Sloooooooow

Ken Miles (kenm@vcd.hp.com) wrote: > I'm looking for ideas on where to look for the source of my performance > problem with tbtape. I have two HP9000/700's with OnLine 5.03 UC1 and > am running a level 0 archive on both of them today. The slower model 750 > with 3 Gbytes of data and a heavy system load finished after 4 hrs, 50 mins. > The "faster" 735 with 2 Gbytes of data and a lighter system load is still > going after 8+ hours and according to the output of tbtape is only "12 > percent done." > The two systems are pretty equivalent with HP Raids (set for "0,1"), > and DAT tape drives. Informix tape block size from tbstat -c on both is: > TAPEBLK 64 # Archive tape block size (Kbytes) > I'm running GLANCE which shows tbtape CPU util at an average of 1% with > about 10 - 15 disc I/O's per second. The CPU is usually only 30% busy. > Surely, it could plod along a little faster than this. Any ideas? > -Ken Miles > Vancouver Division > kenm@vcd.hp.com First take a look of what your 12% means: is the percent of backup respect all of the data space defined, not matter if it used or not. Usualy this percent jumps from your used space to 100%. There is a lot of factor that influence the performance: 1- checkpoint activitie ( the "change rate" activity) 2- size of shm 3- size of phisical log file 4- Hardware config ( are the tape and the disk on the same controler ) 5- The block size: If you have a lot of memory in you system, try using more. Use "dd" to test which is the optimal block size . Hope it helps -- Renato Santis Applications Technical Support Oracle Chile e-mail:rsantis@cl.oracle.com voice:(562) 203-5353 ext 237 fax:(562) 203-5379 Av. El Bosque Norte 0440 Piso 15/16 Las Condes Santiago Chile