[rrd-users] Re: averaged history dampens spikes
oetiker at ee.ethz.ch
Wed Aug 16 22:02:45 MEST 2000
Today you sent me mail regarding [rrd-users] Re: averaged history dampens...:
*> Yep, I have seen the same thing... and spikes are important to me too. I am
*> not actually using cricket, but RRDtool and my own scripts.
*> The problem with a default data config with RRA's progressively decreasing
*> the frequency of your data points into the past is that, logically enough,
*> AVERAGEing brings down your spikes. Since spikes are often valid data that
*> I need to know about, I just use MAX as a consolidation function, and keep a
*> year's worth of data. Yeah, at 300 second steps thats a 1.8 MB file, but
*> disk space is cheap, and it is still better than storing my data in a MySQL
*> database. I also keep decreasing RRAs as in the examples, since they don't
*> take up much space, but unfortunately for my purposes, they lie. ;-)
*> So I pay more attention to MAX data. Comments? I guess it depends on how
*> much you care about valid spikes.
when you are interested in MAX consolidation, you should setup a
MAX RRA but not with a 5 minute resulution but with one which is
bound to be about the resolution you will be graphing at ... the
RRA will then contain the max 5 minute values for lets say each day
of the last year .... that is what max is about .... if you keep
all the values anyways there is no need not to use a max cf ...
if you sample every 1 minute, and then consolidate to 5 minute
intervals with max you will have the max 1 minute value stored
every 5 minutes ... which can not be compared to some other data
you sample every 5 minutes and then store away ... the 1 minute
peeks are bound to be higher ...
hope this helps
*> Yes, yes, I know for consolidation on 1 data point it doesn't matter what CF
*> you pick, its just the exact data, but I sample some ports every minute for
*> two days and consolidate them to every five minutes.
*> I am also curious what Tobi thinks of the performance hit of graphing
*> against every data point for a year rather than consolidated data. Since I
*> prefer MAX to AVERAGE, should I be consolidating my data down with MAX
*> still? Most of my graphing is of the last two days anyway.
*> ----- Original Message -----
*> From: <cricket-users-admin at lists.sourceforge.net>
*> To: <cricket-users at lists.sourceforge.net>
*> Sent: Thursday, August 17, 2000 3:09 AM
*> Subject: cricket-users digest, Vol 1 #214 - 6 msgs
*> > Message: 1
*> > Date: Wed, 16 Aug 2000 13:06:50 -0400 (EDT)
*> > From: Keith Makuck <makuck at email.unc.edu>
*> > To: cricket-users at lists.sourceforge.net
*> > Subject: [cricket-users] Graphs aren't precise for history
*> > I combed the archives, so if I've missed the answer to this, I apologize.
*> > I'm graphing ifInOctets and ifOutOctets of some switches and
*> > routers. I'm only graphing one target per graph -- nothing fancy. Only
*> > thing I've changed from defaults is the sampling, which is every 120
*> > seconds.
*> > If you look at the "Average bits in" and "Cur", the value is 25.0
*> > Mbits/sec. (for example) Let's also say that it says the max is 50.0
*> > Mb/sec. Well, when you look at the Hourly graph, the line denoting the
*> > current value is correct. But, looking back at the history of the hourly
*> > graph, the general ups and downs are right, but the peak never actually
*> > gets to 50.0 Mb/sec. Same is true for "Average bits out" as well.
*> > This happens to all of my graphs and is quite perplexing. Ideas?
*> 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://www.ee.ethz.ch/~slist/rrd-users
*> WebAdmin http://www.ee.ethz.ch/~slist/lsg2.cgi
______ __ _
/_ __/_ / / (_) Oetiker, Timelord & SysMgr @ EE-Dept ETH-Zurich
/ // _ \/ _ \/ / TEL: +41(0)1-6325286 FAX:...1517 ICQ: 10419518
/_/ \.__/_.__/_/ oetiker at ee.ethz.ch http://ee-staff.ethz.ch/~oetiker
Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
Help mailto:rrd-users-request at list.ee.ethz.ch?subject=help
More information about the rrd-users