Correct server settings feedback thread

[Hostingspeeds]

Thanks to Mike of http://www.64bitshosting.nl/ the page generation time error has been solved.

This error occured because of all different settings of php.ini file related to output buffering.

The pageload.php script now takes care of this itself.

The updated version can be downloaded here:
http://www.hostingspeeds.com/pageload.zip

[Hostingspeeds]

Thanks for your feedback. The cheat system should already detect this.

jp

[wasith]

Sory for very long reply. I need little experiment to make sure about this.
My url address : www.hostingdw.net

Sorry that I was made too fast conclusion about server setting, because after restore my old server setting with gzip + turck mmcache enabled, the script still works with following .htacess

php_flag register_globals Off <-- not sure this will affect correct setting
mod_gzip_on no <-- absolutely make correct setting in my server

However, there is little 'cheat' that we can do to get 'better' local page load.
Consider this scenario :

with "mod_gzip_on yes" in .htaccess, we won't get result for local page load test. However we only need one result in 3 days. So I only need very high speed connection, make  "mod_gzip_on no", and then do the test. We will get 'better' result. After that make  "mod_gzip_on yes" (or remove it).

With this scenario we can get 'best' result that not affected by other user/connection test.

To avoid this 'cheat', you should give a penalty to registered server that don't have correct server setting. For example max_time + 1 second.

Just my 2 cents. Sorry for my english.

[Hostingspeeds]

That's GREAT news, wasith!!!

Where is th URL of the test script so I can verify?

[wasith]

After struggling with correct server setting, finally, we can make our server compliant with hostingspeed requirement, I hope so because right now the 'red text' doesn't show anymore.

After disable mod_gzip and turck_mmcache with out luck, finally it works with adding .htaccess file

php_flag output_buffering Off
php_flag implicit_flush Off
php_flag register_globals Off

And thanks to all for this great tool. 

[Bas]

Heres a little information about installing curl.

http://us2.php.net/manual/en/ref.curl.php

Can you explain the problem your haveing in more detail.

Got working, it's too simple to do: apt-get install php4-curl
After the install just a restart of Apache and voila, it's working

[Hostingspeeds]

Host domain: unixroot.org
IP: unixroot.org
Server location: Unknown
NS Lookup Results:
;; connection timed out; no servers could be reached

http://hostingspeeds.com/domainipcountry.php?host=unixroot.org
(working on this script to check server locations)

[RobM]

first you need to bring your server back online -> Sorry, I could not determine the IP for unixroot.org

[NEND]

Heres a little information about installing curl.

http://us2.php.net/manual/en/ref.curl.php

Can you explain the problem your haveing in more detail.

[Bas]

Hi SiServer,

We have a problem.....
JP and me tried to test your script on my server....
I did an apt-get install curl, but that didn't do the trick....
Also a safemode off didn't do it.....
Can you tell what needs to be there?
Help!

[RobM]

There's no way to stop a host from joining..    If it’s a non-bias site, then all should be allowed to join.

[Hostingspeeds]

The 18 server locations have nothing to do with the pageload.php test.

The 18 servers are used for the server response time tool, testing from all 18 locations world wide.

Out of these 18 web hosts here is a list of web hosts that might have some advantage of the test script being on their servers, because they are listed at hostingspeeds.com

These web hosts are:

http://www.bethehost.com/
http://www.plaingoodhosting.com/
http://www.xeonex.com.mx/
http://www.mywww.nl/
http://www.oneavenue.net/
http://www.fast2host.com/
http://www.mainswitch.com/
http://www.crusehosting.com/
http://www.westcoasthosting.net/
http://www.factorysp.com/

We pay about 10 of them ourselves (basic hosting accounts at different hosts world wide) and 8 were donated by web hosts. Some of them joined our project after we signed up for a hosting account there.

So 10 out of 18 are listed. I have looked up their rankings and none of them seem to profit from it substantially (we have 18 locations, and they represent just 1/18th) of it.

If you know a way to get other checking locations by forbidding them to ever sign up at hostingspeeds.com .... then please let me know how, 'cos I don't know how to do that!

[NEND]

If I reading this correct your telling me it does not work on the same host, so that means 75% of the host who donate the space / bandwidth will be killed by the script as it will report wrong...

I am right on this?

Im sort of lost here.  What are the donated servers currently used for now.  If there used to test other servers then they probally are not able to test themselves also.  Its not possible for a server to test its own speed.  Also if they are set up to test the pageload.php file on other host then this script will waiste the same amount of bandwidth as before.  This script just loads pageload.php on other host and gives allot more information about the transfer that the old script couldnt get in the past.  The only thing I would worry about is memory and cpu usage.  If the script is set to slowly test every site I believe it shouldnt use up much memory or cpu.  There hasnt been enough test on this script to verify it even uses allot of cpu or memory. 

[Bas]

I know....

I donate a lot more then just server space also

Like helping to fill the forum with letters?

I wish I could do more, I donated my server as one of the 18 too.....
Also brought JP in contact for a 19th server in Belgium....
Helping find bugg's....what else can I do? Bribe you all to make me no:1 in the list?

[RobM]

Rob,

The donated servers are the testing servers, they are not dedicated servers, but just a bit of space on them.....
One (me) could consider that space a donated server, as it's part of the 18 server response time check....
Ergo, I call them donated servers....we digg?

I know what they are I am 1 of the first to donate here 

I donate a lot more then just server space also 

[Bas]

Rob,

The donated servers are the testing servers, they are not dedicated servers, but just a bit of space on them.....
One (me) could consider that space a donated server, as it's part of the 18 server response time check....
Ergo, I call them donated servers....we digg?

[RobM]

Rob,

I think you are on another planet at the moment

I always on the other planet 

In any matter 3 gigs a month is not much for a donated testing server.
The web host will be waisting the same amount of bandwidth though as you would if somebody started a bulk test from the hosting speeds site or tested the web host from the hosting speeds banner on the web site.

I not sure I following this... There not donated servers...

See people give space off there server, that have clients on them. Do you really think somebody going to give a full server away 

If I reading this correct your telling me it does not work on the same host, so that means 75% of the host who donate the space / bandwidth will be killed by the script as it will report wrong...

I am right on this?

[Bas]

Rob,

I think you are on another planet at the moment
The thing is, SiServer is bussy to work out a script that maybe can kill the anomalies on some servers when a 18-server test is being done.
No harm in that, he's trying to figure out some way to make the 18-server test more resistant to the strange results comming from some servers.
Never hurts to see if it can be improved, updated or done better.....

[NEND]

Well this script was never intended to be run on the web host being tested but on the donated servers testing the web host.  It will not give accurate results running on the same server being tested.

I havent tested the load it will put on a server.  From what I hear curl is the best way to go, when testing with this method.  I guess it would be easier to test in portions, instead of the whole list all at once or add a wait period in the script.  That will cut down server loads.  In any matter 3 gigs a month is not much for a donated testing server.

The web host will be waisting the same amount of bandwidth though as you would if somebody started a bulk test from the hosting speeds site or tested the web host from the hosting speeds banner on the web site.

I would suggest that the script be used to call the pageload.php file that is already there on the web host.  What I sent to the admin is not a entirely ready script.  Its his call on how he wants to set it up and if he sets it up.

[RobM]

I not sure I am following this subject fully, but here a few things I see.

• 99.5 - 99.7%  of the host our here are using resellers packages, so your going to kill there bandwidth as HOST who own there servers have more bandwidth. I also know a few host who don't have large amounts of bandwidth as there servers are more highend.

[li]what the load going to be, have you study that yet? see running cron places a load on a server the more test the more load.[/li]
[/list]

[NEND]

Yeah, I would say its the admins call.

I didnt know that curl gave that much information though.  I just looked at it because I was using fopen in php and some of my online buddies kept telling me you can open 100 more files with curl than you can with fopen with much less resource usage or something like that.  Allot of them are using it for streaming or download purposes and say it works allot better than fopen.

Once I figured out I can get information about the transfer with curl I decided to tell the admin here.  Just suggesting things.  Dont matter if they get used or not. 

[Bas]

If it ran on a cron ever day and the amount of web host that have to be tested it will be about 103 mb a day and 3.1 gb a month.

Still not big numbers when it comes to web hosts.  On most web hosts it will only account for about 1% of their allowed bandwidth usage a month.

That isn't much, but, this project is growing all the time.....so I don't think running it via a cron job is the right solution.

As for those missing bytes....you will find them....did you look under your desk?

[NEND]

If it ran on a cron ever day and the amount of web host that have to be tested it will be about 103 mb a day and 3.1 gb a month.

Still not big numbers when it comes to web hosts.  On most web hosts it will only account for about 1% of their allowed bandwidth usage a month.

[NEND]

Will those are siservers speeds which is in the same hosting facility as the server running the script.  I notice it because for some reason when accessing my pageload.php it sends out 2 extra bytes than the one on your server does.  Dont know why .  It could be the server headers or other information that usually goes on during network traffic that doesnt get reported to the screen of the browser like headers, pings, server information, etc.

Even though your browser will not show it theres always more bytes sent to and from servers. 

*edit let me clear some things up.

The first one is self explanitary.  Its just the error code if any.  200 means OK.
The second test is not really working.
Total transaction time is the total connection time it took to retrieve the document.
Name resolveing is how long it took to get your information from a nameserver.

The next three are debateable.  Which justifies a response time.  So I went ahead and put all three test in the results. 
You got time to establish a connection.  This is the time it took for the server to open a connection for your request.
 You got just before file transfer begins.  This is where the server verifies it has the file your requesting and is getting ready to send it. 
Then you got first byte is about to be transfered.  This is when the first byte of the file requested is about to be transfered.

The rest is pretty simple.  Some is just extra information about sizes of other stuff that got sent during the connection.

[Bas]

Here are my values:

FF

200 - Last received HTTP code
-1 - Remote time of the retrieved document, if -1 is returned the time of the document is unknown
0.562 - Total transaction time in seconds
0.005 - Time in seconds until name resolving was complete
0.005 - Time in seconds it took to establish the connection
0.005 - Time in seconds from start until just before file transfer begins
0.009 - Time in seconds until the first byte is about to be transferred
103252 - Total number of bytes downloaded
183722.41992883 - Average download speed
215 - Total size of all headers received
144 - Total size of issued requests, currently only for HTTP requests

IE6

200 - Last received HTTP code
-1 - Remote time of the retrieved document, if -1 is returned the time of the document is unknown
0.67 - Total transaction time in seconds
0 - Time in seconds until name resolving was complete
0 - Time in seconds it took to establish the connection
0.002 - Time in seconds from start until just before file transfer begins
0.007 - Time in seconds until the first byte is about to be transferred
103252 - Total number of bytes downloaded
154107.46268657 - Average download speed
215 - Total size of all headers received
144 - Total size of issued requests, currently only for HTTP requests

It seems to be OK, now I understand it's about the 0.67 - Total transaction time in seconds, I guess I was looking with my nose
If that's the value that matters, then your script looks very good to me.

So next question, what will the traffic be of the script when it's e.g. run once a day via a cron-job?
Don't forget it has to run a batch over all 1099 servers....I very much doubt if donators of such hosts would like that to happen.