HTTP compression

julien_saadeh

Hello all,

I was searching today about HTTP compression and if it can be enabled locally. I found out that it is done on the server's side and all browsers support it. The thing is that the Internet in Lebanon sucks and people should work on improving things in every possible way - I believe.

HTTP compression reduces bandwidth usage (quota) and speeds up browsing typically by 50% for HTML because it is highly compressible. And CPUs nowadays are mostly "doing nothing" when it comes to Internet browsing, so a little bit of usage will not affect it at all.

I found this tool from Microsoft to test whether a website is HTTP-compression enabled or not:
http://www.microsoft.com/search/tools/default.aspx

Finally, I am trying to shed some light on this subject so any website owner will take it into consideration.

Cheers.

kareem_nasser

URL: http://www.lebgeeks.com/forums/

HTTP status code: 200 OK

HTTP conditional GET: not enabled for the date selected
HTTP compression enabled using gzip

HTTP headers:

Transfer-Encoding: chunked
Connection: keep-alive
Pragma: no-cache
Content-Encoding: gzip
Cache-Control: post-check=0, pre-check=0
Content-Type: text/html
Date: Sun, 02 May 2010 12:08:21 GMT
Expires: Thu, 21 Jul 1977 07:30:00 GMT
Last-Modified: Sun, 02 May 2010 12:08:21 GMT
Server: nginx

rtp

i doubt the actual code is the one that uses bandwidth and stuff
i believe it's the images and flash thingies and stuff like that codes take very few bytes

julien_saadeh

@bluewolf: It's not about the code only but the whole page (excluding photos and videos).
Most websites contain a lot of text and some is repeated amongst article pages like tomshardware.com or any newspaper. I personally open these types of sites daily. Reducing the size of web content will help you not exceed your quota when you have a limited connection like 3GB download/month.

rtp

i see :)

BashLogic

for a long time, IIS did not support this, luckily things have changed a bit.
yes the compression is a must for static and reasonable sized content.
the issue is the more you attempt to compress dynamic content, the slower the server response would become.
in addition, if the size of the content is too small and/or of the wrong file type, compressing it might enlarge the file size! CACHE and PROXY can have issues with it so having it enabled does not necessarily always optimize the service end to end.

Joe

I might be wrong but compressing every generated page would add more computation to the server thus considerably increasing load and decreasing speed. Sure the transfer itself would be faster, and it could help with quotas (as BashLogic pointed out, it is ot always the case).
I believe it should be used sparingly in cases where benefits are clear :-)

for a long time, IIS did not support this, luckily things have changed a bit.

**Random troll about Apache being way better than M$ products. And iPhones suck!**

rolf

You can implement in through a PHP module (I think gz_ or something like that).
I think there were some issues, maybe with older browsers, or CPU usage maybe... I dont know.

julien_saadeh

I think this technology is similar to NTFS compression for hard disk. A Pentium 2 does it instantly, you barely realize a 10% increase in usage when you are installing a program.

Servers have at least 16 logical cores. Will compression really affect the performance?

Joe

Compressing a pages is relatively considerable to generating the page (in PHP, Python, Ruby, ...). In itself it will not take a long time, but what if you have really heavy load on your server? You need every bit of resource you can get.

We are actually running a website with 15 000 000+ page views/month on simply two servers. During peak load time, believe you me, we cannot afford even the slightest operations :)

arithma

On projects of any scale this should be compared based on cost:
- Will it cost more to add a new proxy server that compresses and dispatches the page (as in save bandwidth versus setup and running costs of the server).
- If applied on a single server, will it cross over some constraints of performance (will it break the server under high loads)

There's also the target audience: For audience with low bandwidth and high latency how will it affect the experience (it will benefit with the bandwidth, and the increase in latency would be marginal). In high bandwidth, low latency situation, the increase in latency will become considerable and the repercussion for bandwidth is minimal (compression would be more suited for videos and large images or galleries.. the text and markup could easily be less than 1/1000 of the bandwidth used for such audience so the effort is mostly wasted (premature optimizing and lack of insight into proper effort/gain).

BashLogic

a smart setup is to have a web platform. such a platform is a combination of servers for the portal (talking about heavy duty site here). the platform wifht be distributed as the following:
- http servers that host static files such as images, documents, etc
- http servers that provide stream content
- proxy http server which polls the static components and pages from the actual http servers that render the page. this would ofload the task of transacting the "pages" and generating the pages
- https servers separate for secure content
- http servsers for cgi-bin executions

each of the above is separetly clustered for load balance and redundancy. all of these are so called front end servers. application, midleware and backend DB each require their own equivelent optimization.

etc etc

that was to give you a picture of what some do in the "big world" ....

amkahal

Try www.fastun.com to speed-up ur internet connection by compressing images and some other compressions, it's free.