Date: Fri, 13 Oct 2000 00:37:16 -0700 From: Don Lewis <Don.Lewis@tsc.tdk.com> To: Bill Fumerola <billf@chimesnet.com>, Doug Barton <DougB@gorean.org> Cc: Matt Dillon <dillon@earth.backplane.com>, cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/etc inetd.conf Message-ID: <200010130737.AAA05256@salsa.gv.tsc.tdk.com> In-Reply-To: <20001013031655.K37870@jade.chc-chimes.com> References: <200010121857.e9CIvAi30686@earth.backplane.com> <Pine.BSF.4.21.0010121438170.17795-100000@dt051n37.san.rr.com> <20001013031655.K37870@jade.chc-chimes.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Oct 13, 3:16am, Bill Fumerola wrote: } Subject: Re: cvs commit: src/etc inetd.conf } On Thu, Oct 12, 2000 at 02:53:03PM -0700, Doug Barton wrote: } } > Finally, I (and I suspect Bill F. too) DO realize that cron'ing } > sendmail -q can be more expensive than keeping it running. But, I would } > almost be willing to put money on running it once a day from cron using up } > less system resources in a 24 hour period than keeping it running } > does. But even if I'm wrong, it still gives ME control over when I run it, } > which means I can schedule it for a period that suits my needs without } > having to worry about it being running all the time when I know it's not } > needed. } } Exactly. -q30m is ambigious, 0,30 * * * * is exact. While being exact can sometimes be important, it would not be wise to implement something like this on a wide scale because of the potential for large spikes in network load. In the past there have been widespread deployments of network software with example cron entries. Most users would just configure it to run at the default time, and the poor server at the other end would melt down. Just imagine what would happen if FreeBSD were distributed with a commented out crontab entry for cvsup and a comment that said that the start time should be randomized. Most users would just enable cvsup by removing the comment character and the cvsup servers would get hammered at hour intervals, with the size of the load peaks dependent on how many FreeBSD users lived in each time zone. Even if users could be counted on to randomize the start time, there would still be network load spikes at 60 second intervals because thats the minimum granularity that can be configured in the crontab. The spikes get sharper as time synchronization using NTP becomes more popular. NTP actually does a fair amount of work to randomize the time at which it sends its packets to avoid this sort of problem. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200010130737.AAA05256>