On Fri, Nov 30, 2001 at 10:17:36AM -0800, Jason Fesler wrote: > > > Of course, there is a need having a scalable frontend and data collector > > (SNMP/custom scripts) without reeventing the wheel. And also code which > > will be > > a startingpoint with clear and functionable code on which i can rely and > > use to > > Ultimatley I wrote my own that was better suited to my environment than the > other front ends were. RRD's biggest expense is IO, not CPU, at least when > it comes to data storage. Graphing is CPU intensive, but we solved that by > only showing graphs when needed, and making the display system flexible (I > have a single iamge showing thing that takes all of its info by URL, and > people just make web pages refering to the url to pull the data they want > displayed..). > > I currently update 8700 files once a minute; each file has on average 10 DS's. > I'm doing this on an e450 with 2x 8 drive raid5 configurations. I probably > should have striped things for IO speed - I use every bit of IO I have but > hardly touch the space. On average I'm in about 60 %busy on IO, which is > definately taking its toll on the machine interactiveness.
A good idea is to spool the updates, then every X minutes sort them so you have multiple updates to the same RRD file right after each other, that helps a LOT on the IO load. /Jesper -- Jesper Skriver, jesper(at)skriver(dot)dk - CCIE #5456 Work: Network manager @ AS3292 (Tele Danmark DataNetworks) Private: FreeBSD committer @ AS2109 (A much smaller network ;-) One Unix to rule them all, One Resolver to find them, One IP to bring them all and in the zone to bind them. -- Unsubscribe mailto:[EMAIL PROTECTED] Help mailto:[EMAIL PROTECTED] Archive http://www.ee.ethz.ch/~slist/rrd-users WebAdmin http://www.ee.ethz.ch/~slist/lsg2.cgi
