On Thu, 4 Dec 2003, Mark A. Lewis wrote:
> Except that with ISA you can use a SMP box. The 2 products have
> vastly different hardware requirements, if you buy/build a box for
> squid it will be a 1 CPU box, which would not be a good choice for
> ISA.
Huh? Squid can take advantage of a multi-processor machine just fine.
Squid may not take advantage of it directly, I don't know whether it
can or not given the right OS and threads library. That is a question
for someone like Henrik. However, if you use the diskd storage type,
each diskd process certainly will use every CPU it can get, and if you
use external redirectors, each redirector process will as well, and
that is what can really eat up the CPU. I'm running Squid on a
dual-processor box for exactly that reason, and the more CPUs the
merrier if you have the need for it.
> 2) ISA is more scalable in the sense that you can do more with one
> big box. With a 4 CPU 900mhz 2GB ProLiant we get over 100gig of
> throughput a day and tens of thousands of users and millions of
> requests. I have not seen any recent stats on what Squid can do with
> a large single CPU box, but I don't think it can approach that.
Hmm, we're using a 2-CPU 800Mhz Proliant with only 1GB RAM. It has
seen about 4GB and at least a million requests per day recently,
peaking at over 140 requests per second with over 1000 clients. This
is all while also doing content filtering based on a list of almost 1
million domains, URLs, and some regular expressions, and it also has
to deal with delay pools for each client. It currently has to manage
over 2.1 million cached objects totalling about 23GB, spread over only
four 7200RPM SCSI disks.
It doesn't do 100GB/day, but I'm fairly sure thats because we only
have a T1 (1.544Mbit/sec) to feed it with and there isn't enough hours
in the day to transfer that much. The server is barely taxed even at
peak times. Given the amount of disk I/O and CPU I have left over, it
could easily do several million requests per day (and I'm talking
about approximately 8 hours of usage per day, not 24 hours!) and many,
many GB given enough bandwidth. If this server had a 24-hour load and
a much fatter pipe like an ISP would have, the numbers would likely be
in the tens of millions of requests per day and a whole-lotta GB.
Now imagine what it could do if it didn't have to deal with that pesky
content filtering.
> 4) There are no promises that Squid will be supported in the future.
> As much as some may not like it, MS will be around for a while.
Squid is open-source. There is no such thing as "no support in the
future" as long as you have the source code and someone with the
knowledge to work with it. Even if for some odd reason most Squid
users did decide to stop using Squid in the future, if just a handful
of people use Squid it will continue to be improved. A handful of
people is how it started, and look where it is now. When Microsoft or
any vendor of closed-source software discontinues support for any of
their products, especially in the form of security and feature
updates, there isn't a single thing you can do about it because they
hold all the cards.
-- Chris Dillon - cdillon(at)wolves.k12.mo.us FreeBSD: The fastest, most open, and most stable OS on the planet - Available for IA32, IA64, PC98, Alpha, and UltraSPARC architectures - x86-64, PowerPC, ARM, MIPS, and S/390 under development - http://www.freebsd.org Q: Because it reverses the logical flow of conversation. A: Why is putting a reply at the top of the message frowned upon?Received on Fri Dec 05 2003 - 00:01:33 MST
This archive was generated by hypermail pre-2.1.9 : Thu Jan 01 2004 - 12:00:05 MST