[rrd-users] Performance problem updating many RRD files

[Alfredo Diehl]

Hi,
   I'm using RRDTool to store data series produces by flowscan/JKFlow and I hit a performance problem updating many files. With 256 Mb RAM caching works OK up to about 2000 RRDs, but after that # performance fall suddenly. About 180Mb of RAM are dedicated to file caching, and I imagen that at certain amount of rrd files the caching system start to drop based on some LRU algorithm and the windows of caching just run useless behind the secuential update done on rrd files.
  Initially I was using RRDTool 1.0.49 and now I changed to 1.2.11 hoping "MMAP IO" would help with this. I made sure I compiled it with "--enable-mmap" but there was no diference in the # of files where performance degrades. I did not find any detail in how does MMAP works in the web, docs, neither have enough programming knoledge to understand the code, so I would apreciate if someone can give me information on:
   
  *The process of updating rrd files: I thought it was just needed to read/write a few bytes in every file, but it seems it is necesary to load a considerable part of the file in the cache in order to update it.
  *How MMAP IO works and how I use it: I imagen writing the updates of many files in memory and then sending all the request together to the disk, letting it to be more efficient. I didn't see any reference in rrdupdate doc on how to indicate how to use mmap.
   
   One quick solutions seem to be adding RAM, but I just want to do this as efficient as possible.
   
   Thank in advance,
   
   Alfredo
		
---------------------------------
 Yahoo! FareChase - Search multiple travel sites in one click.  

--
Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:rrd-users-request at list.ee.ethz.ch?subject=help
Archive     http://lists.ee.ethz.ch/rrd-users
WebAdmin    http://lists.ee.ethz.ch/lsg2.cgi