[prev in list] [next in list] [prev in thread] [next in thread] 

List:       mandrake-newbie
Subject:    Re: [newbie] Re: FireFox speedup
From:       David Reynolds <naeblis () kc ! rr ! com>
Date:       2005-01-09 18:19:12
Message-ID: 200501091219.12643.naeblis () kc ! rr ! com
[Download RAW message or body]

On Friday 07 January 2005 04:53 pm, Miark wrote:
> On Thu, 6 Jan 2005 18:44:12 -0600, David wrote:
> > > I've tried it for about a week now, and honestly, I can't tell any
> > > difference.  Besides, my conscience forbids me to send 30 request
> > > to a server in order to get one page.  There's enough clogging the
> > > net already.  But that's just me...
> >
> > This 'scheme' was proposed elsewhere last week to almost universal
> > condemnation. There are pieces in it that are worthwhile (like #3), but
> > as Kaj mentioned, there's a lot that's unethical as well:
> >
> > http://www.livejournal.com/community/linux/1065384.html
>
> I read the whole page (without following each thread) and quite
> honestly, I didn't read any compelling arguments against it.
> If 10 people hit a page with 10 items on it, that's 100 items
> that have to be served. Whether those items are served in
> big bursts (as with pipelining) or little bursts (serving
> 10 items 10 times for 10 users without pipelining) the CPU
> and storage drives are going to work just as hard. The only
> difference will be the distribution of the used resources.

IANASA (I am not a sysadmin), but I believe the problem is not SO much with 
the CPU and HD, as with the outward bound pipes and access points for the 
website. You're familiar with the Slashdot effect, yes? By setting up your 
computer to access the same site with 9x, 19x, 49x as many pipes, you're 
simulating the effect of 9x, 19x, 49x as many users accessing the site at the 
same moment. It's for a briefer period of time (theoretically, although my 
experience with broadband vs. dialup is that if you've got more bandwidth, 
you're more likely to visit more pages on any given site, and more sites), 
but you're crashing the processes that serve pages out to the public. So 
while *your* experience is improved, everyone else may be getting "Server 
unavailable" messages, and the webmaster of that site may end up blacklisting 
you for it (DDOS is subjective, after all), and then your experience of the 
site will be vastly diminished.

> Perhaps someone can do a better job of explaining the problem.
> But in the meantime, I've changed mine to 9. It's a huge
> difference in performance, and it's low enough not to be
> construed as a DOS attack.

I dunno. A lot of it flew over my head, but here was the bit that decided for 
me that it (everything but the page refresh piece) was a bad idea:
"The problem is that the settings are using a lot of resource in general, it's 
an abuse of common courtesy by maximizing the amount of resource used by you 
at expense of others. A lot of the internet is dependant upon good behavior 
by the majority. 
A common example is the general TCP protocol, where you are supposed to 
practically stop transmitting if you detect congestion on the link. 
(detection is done by seeing a lost packet due to overburdened router.) 
Theoretically you can "improve" your performance by continuing to send at 
maximal speed and thus ignoring the congestion. As everyone else will 
back-off to let the routers work through the congestion and will try again 
from low speeds, if you do not follow the convention your speeds would be 
maximized. However, such behavior is considered impolite and all major 
implementations of TCP follow similar standard behavior."

> Miark

Regards,
David


["message.footer" (text/plain)]

____________________________________________________
Want to buy your Pack or Services from MandrakeSoft? 
Go to http://www.mandrakestore.com
Join the Club : http://www.mandrakeclub.com
____________________________________________________


[prev in list] [next in list] [prev in thread] [next in thread]