RE: Killing caches (was RE: Pay-Per-Click Question)

From: Justin W. Hitt <Justin.Hitt_at_hittpansophism.com>
Date: Thu 14 Jun 2001 11:15:32 -0500

ALEX CHUDNOVSKY <alexc_at_jungle.com> WROTE:

>No way, at least not for us (jungle). More and more
>sites are dynamic, and this trend will continue.
>There will be more personalised content, which isn't

Alex, I agree with your points on measurements of site
success as well as the differences between the value of
local log files and aggregated statistics. I do feel it
necessary to make the point that even though a sites pages
are generated dynamically, static content exists and a site
may benefit from caching at the expense of log data. For
examples, you update your home page every 6 hours, it
averages 8 Kbytes, it changes 80% of the page, yet your
logo and several page elements stay the same (20%) --
you receive 20,000 visitors that day.

Remember, some caches can cache page elements -- the things
between the HTML.

In my example raw without any caching, 160MB is transferred
(8K * 20,000), according to the formula above, 32MB of which
is cachable. Utilizing caches would change that pages
transfers from 160MB to 128MB saving you 970MB in traffic
a month. Imagine your site has 500 pages with the average
page view of 20,000 per day, caching saves the Internet
485,000MB per month! Of course, I am purposefully being
conservative looking at visitors and not impressions,
either way caching plays a big role on everyones network.

Unless you forbid caching (which some systems will do it
anyway) some caching happens which leaves little spots in
your logs, but big savings for your connection. Sites like
mine with mostly static content has bigger holes (only 1
in 10 pages are logged) while more dynamic sites will have
a clearer picture. I can't do much pathing, however, my
site is designed to be only 3 clicks deep, so that isn't
very necessary.

>cut it. Personally I prefer to have _ALL_ log files
>in data warehouse which enables me to run customer
>queries which would link use actions to actual sales.

That is certainly the best way of doing it! I actually
measure everything on my site by leads, sales, and derived
conversion rations. Log files are only used to verify
(check) ads impression reports, and derive some entry
points from various search engines. Logs are also used to
check for errors and odd traffic trends. Most of my tracking
revolves around a back-end client system and the merchant
server -- it's the money that matters most.

>Whilst Alexa's data can be used for some useful things,
>they are of questionable quality and can't be taken as

Talks of Alexa on this list and a few others have convinced
me Alexas okay for a reference tool if you are interested in
very general information, but look for Media Metrixs, your
own stats, and sales reports if you really want to know how
you compare. I have uninstalled Alexa from my main system
and my browsers run faster -- go figure.

It has been a great chat. I hope you consider ways to use
caching to improve your sites performance, there are many
things you can do to make the site faster even with
dynamically created content. From my ten years as a Unix
Systems Administrator for several high traffic websites and
software development environments, I have found you can always
make a site faster if you look at it's smallest most
frequented elements.

Warmest regards,

Justin

PS. Email me and I will tell you how I made my Apache server
241% faster for Internet Explorer and Netscape 4.0+ browsers!
It is so simple you will not believe it!
--
Save time with the Business Success Newsletter,
http://www.hittpansophism.com/newsletter/
(v) 301-904-3429 (f) 520-438-2941
(e) Justin.Hitt_at_hittpansophism.com



Received on Thu Jun 14 2001 - 11:15:32 CDT