<No Subject Supplied>

>From naomi Mon Jul 5 19:37:53 MST 1993 remote from qip From: naomi@qip (Naomi Walker) Message-Id: <9307060237.AA25115@qip> Received: from qip by misty.anasazi.com; Mon, 5 Jul 93 19:37:55 MST Received: by qip (5.61/1.2) (from naomi (Naomi Walker) to informix-list@rmy.emory.edu) id AA25115; Mon, 5 Jul 93 19:37:54 -0700 Subject: LRU Writes To: informix-list@rmy.emory.edu (Informix List) Date: Mon, 5 Jul 93 19:37:53 MST X-Mailer: ELM [version 2.3 PL11] Content-Type: text Content-Length: 1355 > > Can anyone explain exactly what triggers an LRU write in Informix- > OnLine, or what the distinction is between LRU writes and Idle writes? I sometimes play with setting LRU_MIN_DIRTY and LRU_MAX_DIRTY in tbconfig. The default values cause the queues to start cleaning when they are 60% dirty, and keep cleaning until they are 50% dirty. Since cleaning LRU's during a checkpoint locks shared memory, I set LRU_MIN_DIRTY to 5 and LRU_MAX_DIRTY to 10. This way the page cleaners are running pretty well flat out, and I do not see hesitation during intermittant times of the day. I used to let the checkpoint do all the work, thinking sorted writes would be much faster, but in real live life, I have had great success with this kind of tweaking. As usual, I wish I knew more. I especially like to play with these parameters on our multi-cpu Pyramid SMP platforms. It doesn make as much sense on our 486 platforms. -- Naomi Walker (aka N7FSA) naomi@anasazi.com Phoenix, Arizona To get the full value of joy, you must have someone to divide it with. ---Mark Twain -- Naomi Walker (aka N7FSA) naomi@anasazi.com Phoenix, Arizona To get the full value of joy, you must have someone to divide it with. ---Mark Twain