Best performance with cache and vBSEO 3.0.1

Im interested in share experiences about diferents cache combinations.

I have a big board (3.500.000 posts and 44.000 users, 1500 -2000 concurrent users) with vBulletin + vBGallery + vBSEO.
I have two servers one for database and one for front end.

I make different combinations for example memcached + eAccelerator-xCache, only memcached, and only eAccelerator-xCache.

The best result that i obtain is with all cached with xCache.
For example with memcached i save 2 queries but i think that the process of serialize-unserialize over the net is slow.

Please post your experiences


Excuse my broken English

Memcached + xCache here.

xcache for all here also.

I have similar setup. Two identical servers one for mysql and one for PHP.

My msyql CPU avg about 2% - 10% and PHP server between 40% - 100% spike during peaks. But with vbseo disabled less than 10%.

What have you guys done to reduce web server CPU loads?

Attachment with arrow is running 3.0.0. But so far 3.0.1 has been around the same. Will know for sure tomorrow during peak hours. But already see 70% spikes :(

Other attachment is with vbseo 3.01 enabled then disabled.

What version of eaccelerator is compatible with 3.0.1?

I would use xcache if I were you. Esp if you on IIS but if not EA should be fine.

Ok, vbseocp says that eaccelerator is supported (it's the only thing my sysadmin would install, he said he didn't like the other caching systems) but when I go to save the option, it will never save it...it keeps taking the radio button and putting it back to "none".

cmod or set permissions to write for the config_vbseo.php (located in forums/includes/)

Then save again

I have it as writable...

I've tried chmod 666 and chmod 777, neither of which work.]


I'm opening a support ticket for this :( I have no clue why this is not working right.

Originally Posted by jbartle

I have it as writable...

I've tried chmod 666 and chmod 777, neither of which work.]


I'm opening a support ticket for this :( I have no clue why this is not working right.

You dont need a support ticker for this is only that the function to put and get from cache are disabled of eAccelerator, please follow the steps related in my entry blog.

Enable caching functions get and put in eAccelerator - vBulletin SEO Forums

Thanks for the guide, I'll pass it on to my sysadmin. But my problem is also lying somewhere else. I can't save any option...so that's why I opened the ticket up.

Odd to see xCache with a higher rating in the poll. I've yet to see xCache outperform eAccelerator in any benchs that I have done although there have never been a huge difference.

Has anyone actually seen xCache outperform eAccelerator with vBulletin? I can understand it slightly outperforming with other things even flat files but with an actual application I suspect eAccelerator might just edge it.

With the benchs I performed this morning with standard source installs of mySQL/PHP/Apache and using the default options for xCache and eAccelerator I clocked eAccelerator at 46.15 requests/second and xCache at 44.52 requests/second, not a huge difference.

Take that with the same install slightly optimized (optimized mysql, optimized apache configuration [includes dropping modules that are not needed], reconfigured accelerators) and eAccelerator went to 652.99 requests/second , with xCache clocking at 648.49 requests/second.

Those were testing the index of a vBulletin forum with the following versions and compile options,

httpd-2.2.6 - ./configure -–prefix=/usr/local/apache –-enable-mods-shared=most
mysql-5.0.45 - ./configure --prefix=/usr/local/mysql --enable-assembler --with-mysqld-ldflags=-all-static
php-5.2.5 - ./configure -–prefix=/usr/local/php -–with-apxs2=/usr/local/apache/bin/apxs -–with-mysql=/usr/local/mysql

Anyone else have any real world examples that show xCache actually clocking faster than eAccelerator on vBulletin?

Originally Posted by Scott-Mc

Anyone else have any real world examples that show xCache actually clocking faster than eAccelerator on vBulletin?

hmmm...
Was on eAcc for 6+ month
and moved to xCache(along with datastore cache).Now using xcache for more than 1 yrs.
well....it really works far more better than eAcclereator.
performance gain- faster page loading/reduce bandwidth/reduce sql query

My site is Golpo Community Forum - Most Alive Bangladeshi Online Community [vbseo3.0.1]
xcache's dev. is continiously working on it...thus eAcc dont got that speed at all..

check by urselves
XCache demo
future relase for xcache

Originally Posted by Nneel

hmmm...
Was on eAcc for 6+ month
and moved to xCache(along with datastore cache).Now using xcache for more than 1 yrs.
well....it really works far more better than eAcclereator.
performance gain- faster page loading/reduce bandwidth/reduce sql query

My site is Golpo Community Forum - Most Alive Bangladeshi Online Community [vbseo3.0.1]
xcache's dev. is continiously working on it...thus eAcc dont got that speed at all..

check by urselves
XCache demo
future relase for xcache

Do you have any real world benchmarks to confirm those results, I have yet to this day to see any real world benchmarks that show xCache as faster than eAccelerator on vBulletin.

With the results you said however I highly doubt that it's the php accelerator alone which would produce any real huge differences explicitly because you will struggle for them to gain more than an extra 10 requests/second using all of the most popular ones. Which again would vary depending on the application itself and in this case the mods you have for vBulletin.

dont ask me to google it for u....
as i said my own experience

Originally Posted by Nneel

dont ask me to google it for u....
as i said my own experience

Excuse me?

So what you are saying is you have no benchmarks so in turn you are taking snap guesses?

I am not entirely sure why you are being hostile, I asked if anyone has any real world benchmarks to prove otherwise which you have not done. I fully appreciate you may have had slightly different experiences but my own would indicate otherwise and I have yet to find anyone to prove otherwise. Now if you have benchmarks that can prove otherwise then please by all means post them.