Generally, you should have a separate RRD file when the source data do not arrive together, or if the number of data are likely to change. However, updating a single RRD with many DS is cheaper (in IO terms) than updating many RRD with one DS each.
EG1: Querying CPU usage on a single machine results in stats for User, System, Wait and Idle. These are all retrieved simultaneously, so can be held as separate DSs within a single RRD. EG2: An ISP is storing data usage by its clients. Clients may come and go, and the statistics for client data usage are obtained by separate queries from separate routing devices. In this case, we have a separate RRD for each client, each RRD holding just the DSs that relate to that one query (probably bytes in and bytes out). Of course every application is different and the optimal way to use RRDtool depends on your particular requirements. So think about where the data come from, if the number of DS will change, how often the data arrive, do you have control over when the data arrive, and so on. For example, our setup has 30,000+ RRD files, each about 1.5MB in size, usually holding 8 RRAs and 2 DSs. The number of RRD changes often, the data all arrive at different times. We use rrdcached to manage disk IO requirements. Steve _____ Steve Shipway ITS Unix Services Design Lead University of Auckland, New Zealand Floor 1, 58 Symonds Street, Auckland Phone: +64 (0)9 3737599 ext 86487 DDI: +64 (0)9 923 6487 Mobile: +64 (0)21 753 189 Email: <mailto:s.ship...@auckland.ac.nz> s.ship...@auckland.ac.nz P Please consider the environment before printing this e-mail : 打印本邮件， 将减少一棵树存活的机会 From: rrd-users-bounces+s.shipway=auckland.ac...@lists.oetiker.ch [mailto:rrd-users-bounces+s.shipway=auckland.ac...@lists.oetiker.ch] On Behalf Of S Ahmed Sent: Thursday, 24 October 2013 9:58 a.m. To: rrd-users Subject: Re: [rrd-users] large dataset considerations Hi, Is this tool used by any large scale usages? What is considered a large database size? Just looking to get an idea on how people use this tool at scale. Scenerio: Say you want to store time series time informaiton in a Saas application. I'm guessing there is some sort of threshold where it makes sense to partition your data e.g. by a group of customers in order to scale out the usage
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ rrd-users mailing list rrd-users@lists.oetiker.ch https://lists.oetiker.ch/cgi-bin/listinfo/rrd-users
