[sf-lug] Fwd: [PenLUG] Donation for hosting
rick at linuxmafia.com
Mon Oct 23 02:22:57 PDT 2006
Quoting ron (rondosxx at yahoo.com):
> I am totally unaware of how even our own LUG covers its operating
> costs. I suspect they are minimal, though perhaps a couple of people
> are supporting it under the radar; Jim and Rick come to mind.
Jim Stockford picks up the only direct cost I'm aware of, which is the
approximately $12-15 per year for domain registration (used for the Web
site). The SF-LUG mailing list shares my cheap ol' Linux server on my
house aDSL line. That doesn't per-se cost me a penny, because I keep
that server up and running for other reasons, the mailing list adds
negligible extra load to what the host already does, and y'all are
perfectly welcome guests.
I cannot comment on PenLUG matters, but the general topic reminds me of
something I've been pondering for a while:
Typical Web/e-mail sites (Internet servers) for LUGs or similar small
non-profits or small businesses can run perfectly on throwaway-hardware
Linux servers drawing ridiculously small amounts of bandwidth. Here's a
quick'n'dirty way of guesstimating bandwidth requirements.
1. Run "/sbin/ifconfig [interface]" to get the transmitted (TX) and
received (RX) byte counts. Note down the time/date.
$ /sbin/ifconfig eth1
eth1 Link encap:Ethernet HWaddr 00:D0:B7:93:31:0E
inet addr:22.214.171.124 Bcast:126.96.36.199 Mask:255.255.255.248
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:66321724 errors:0 dropped:0 overruns:0 frame:0
TX packets:77788011 errors:0 dropped:0 overruns:0 carrier:0
RX bytes:1500480318 (1.3 GiB) TX bytes:903698173 (861.8 MiB)
Interrupt:11 Base address:0x1080 Memory:fa202000-fa202038
Mon Oct 23 00:13:41 PDT 2006
2. Wait a while -- preferably at least a day. Do step #1 again (to get
a second data point at a later date/time). Subtract figures between the
dates, to get bytes received and transmitted during the period in
question. Add the TX and RX byte differences together. (They're both
traffic.) Divide by elapsed time, to get average network traffic per
unit of time.
Traffic levels may have spikes, but typically not significant ones for
small groups' Internet sites' Web and mailing list traffic. Over at
SVLUG, one of the largest LUGs in the world, we did this for our
"www.svlug.org" AKA "lists.svlug.org" AKA "svlug.org" host, and found
that average traffic levels were so low that we could have comfortably
put the machine on a dial-up modem line, and not suffered any noticeable
So, what does such a site really need, if not non-trivial bandwidth?
o IP service that works, at one or a series of locations.
o one or more cheap throwaway machine (PII, 128MB RAM, etc.)
o ability to re-point DNS records fairly quickly (1/2 day, say)
o system administration at the level of an interested and careful amateur
A friend of mine runs a set of interrelated domains hosting Web sites
and mailing lists for some large, annual volunteer events -- which were
for a long time hosted on one of the volunteers' Solaris machines.
(For any who don't know, Solaris is a Unix, recently open-sourced.)
That machine lived at the volunteer's house, which was on PacBell/SBC
home aDSL line. Recently, in a development that will be familiar to
frequent SBC broadband customers, the volunteer has been suddenly
subjected to weeks of mysterious service outage, with no end in sight.
My friend is livid, but unfortunately for him had no Plan B, and so is
currently stewing over the situation.
What went wrong? a) There was no standby spare machine or even standby
spare hard drive. b) There were no offsite regular backups. c) Nobody
in the group had the expertise to make use of those things, had they
Despite the fact that building a replacement server (with Linux, BSD, or
Solaris) and deploying it any-old-where is dead simple, my friend is
still suffering extended downtime while looking for a "hosting service"
to magically take care of all backup, maintenance, and system
administration for him, for some serious bunch of money per month.
The way I'd do it is: Run the Internet services on a cheap machine
running just about anywhere that provides a static IP. (Avoiding
PacBell/SBC broadband would be common sense, though.) Have daily or
better backups. Have at least one standby machine available, preferably
OS-loaded and ready to go. (Making the backup be directly onto one of
the spare machines would be a nice touch.) Be able to re-point the DNS
in a hurry.
Not a lot of expertise is needed, frankly, and rudiments of Unix system
administration aren't exactly brain surgery. (Dealing effectively with
the spam problem's about the only challenging part.) Yet, people spend
a lot of money and time on paid hosting arrangements, colos, etc., when
most of the time they don't have to.
More information about the sf-lug