Calum, your explanation, though lengthy, was really worth of reading. I am glad to see that even others had the concerns same as me about accuracy of data being stored. When I was going throgh the tutorials for the first time, I had the same doubt in mind - if a rrd file is created with a step of 5 minutes, does that mean the external updating script (perl/shell) has to run and provide the data exactly at 5 mins? What if I run updating script every 2.5 mins (twice in 5 minutes)? What if the script doesn't provide the data in exact 5 mins? Interpolation gave some vague idea about it, but didn't make clear the data storing mechanism. I hope Tobi can shed a light on this :) After your point, now I have two concerns - validity of data polled from targets and accuracy of data while storing into the database. As you have mentioned, second concern can be waived based on the assumption that there is not much of the loss of accuracy because of interpolation. I still have to convince managers that first point is not too much to bother about if we rely on credibility of snmp. After my monitoring system would start working successfully, I am planning to come up with a document with some more detailed description. Tutorial by Alex is excellent. All I am planning to do is just to add some more information, that I feel, will fill in the missing links. Thanks Calum cheers k2 Calum Anderstrem <[EMAIL PROTECTED]> wrote: Although you have been concerned with validating the data you wish to store within RRDtool, there is potentially another issue which will affect the accuracy of the results displayed. This issue centers around the interpolation function which accepts that it would be almost impossible for a data collection process to pass information to RRDTool at the exact moment the sample is required. This interpolation function attempts to calculate what the reading would be at these times and stores it in the respective RRA. Sometimes this calculation gives results which appear to be incorrect, and a number of email queries have been generated ( please see http://groups.google.com/groups?dq=&hl=is&lr=&ie=UTF-8&oe=UTF-8&group=mailing.unix.rrd-users&selm=abcf4a18.0302241420.30a30467%40posting.google.com
(a couple of the items here I don't think are problems) I think it only fair to point out that although I am concerned about potential inaccuracies in the recorded information, our company uses RRDTool extensively for the logging and graphing of all our network elements, and it has been extremely useful. We don't use it for billing, so these potential inaccuracies are not a problem. I apologise for the lengthy mail, but hopefully you will find it of interest. Regards, Calum Anderstrem ----- Original Message ----- From: "Ketan Patel" To: Sent: Tuesday, February 25, 2003 5:28 AM Subject: [rrd-users] Validation of data collected > > In a meeting while discussing use of RRDTool, a question was raised about verifying the data being collected from target using SNMP queries. We want to make sure what we are looking at, is actually a correct data and represents the true health of the target. > > I understand SNMP counters represents actual values. However, I am not sure if there is any hole that gives wrong replies to queries. > > So, does anyone know how to validate the data before storing it? At the initial stage we are monitoring load on a server by a running process and a router carrying out the traffic. > > cheers > > Ketan > > > > --------------------------------- > Do you Yahoo!? > Yahoo! Tax Center - forms, calculators, tips, and more > > -- > 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 > --------------------------------- Do you Yahoo!? Yahoo! Tax Center - forms, calculators, tips, and more -- 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
