From: Blake B. <Bla...@De...> - 2001-08-03 19:40:15
|
An alternative free caching engine which compares with Zend is APC http://apc.communityconnect.com/ When Rasmus was last here he recommended it.. so I assume it's quite good. :) * Blake -----Original Message----- From: Andreas Aderhold [mailto:a.a...@th...] Sent: Friday, August 03, 2001 3:37 AM To: bin...@li... Subject: [binarycloud-dev] Zend Cache and binarycloud = more power Hi, I just benched binarycloud with/without ZendCache in comparison with other scripts. And the results are surprising. Actually with Zend Cache enabled I gain huge performance boosts for sites without database access. Read on. --------------------------------------------------------------------------- Setup --------------------------------------------------------------------------- I ran the benchmarks on my devel box at home (crappy old box, with lots of stuff running (sshd, samba......)): * Intel Pentium 100 * 128MB PC100 RAM * 15GIG IBM HDD (IDE) * Linux Mandrake 8 (Kernel 2.4.3) * Apache 1.3.19 * MySQL 3.23.36 * PHP4.0.4pl1 (Zend Launch Pad Edition) * ZendOptimizer 1.1.0 * ZendCache 1.1.0 PHP compile options (launchpad - out of the box) ./configure --prefix=/Zend --with-apxs=/usr/local/apache/bin/apxs --disable-debug --with-config-file-path=/usr/local/Zend/etc --without-db2 --disable-sysvsem --disable-sysvshm --enable-xml --enable-wddx --with-mysql --with-gd=shared,/usr/local/gd --enable-gd-imgstrttf --with-jpeg-dir=/usr/local/jpeg --with-ttf=/usr/local/freetype --with-t1lib=/usr/local/t1lib --with-pgsql=shared,/usr/local/postgres --with-imap=shared,/usr/local/imap --with-ldap=shared,/usr/local/openldap --with-ibm-db2=shared,/usr/local/db2' --with-oci8=shared,/usr/local/oracle Zend Cache: - validate timestamps is enabled (so chache checks everytime if script changed, will perform even better if this is disabled in production) - reserved cache memory: 32Megs --------------------------------------------------------------------------- Benched scripts --------------------------------------------------------------------------- [1] Thyrell homepage www.thyrell.de/home/index.php A very simple script with sessions (language-detect) and smarty as template engine (smarty cache disabled). It is just one template with basic replacing work to do for the tpl engine (no section loops etc). No Database access. [2] Database script (huge) Manage script for category administation (unlimited depth for cats). Script is currently not r2 based [but r2 friendly and after the entitity/database stuff is out, it will be ported:)]. MySQL Database, about 20 entries for categories with a depth of 4. This script is part of a more complex community application administrtion I'm currently working on. So much things are not optimized and It's overloaded with includes that will be removed/optimized. Escpecially the layout for the admin interface is very grapic intensive. So currently the loading time is very long. Basically: - Request class, session class of r2 - Metabase - Smarty - PEAR for error handling (exceptions and formwarnings) - The "manager" responsible for db-access uses recrusion to look up the categories (in this example depth = 4). It also currently executes 2 queries per query (COUNT(*))(Few template chunks with descent template logic (looping over arrays). - Advanced browser scripting (mostly IE5 behaviours). [3] binarycloud: index.php The current (2001-03-08, 10am GMT) cvs r2/binarycloud/build/en/index.php. No changes. --------------------------------------------------------------------------- The results --------------------------------------------------------------------------- I ran the benches doubletwice and the results differ slightly. [1] Thyrell homepage ------------------------- With cache Starting benchmark for http://dev.thyrell.corp/home/index.php (5 tries)... Try 1: 6.68 req/s Try 2: 6.90 req/s Try 3: 6.82 req/s Try 4: 6.79 req/s Try 5: 6.77 req/s 6.792 requests/second from 5 tries Without cache Starting benchmark for http://dev.thyrell.corp/home/index.php (5 tries)... Try 1: 0.98 req/s Try 2: 0.60 req/s Try 3: 0.94 req/s Try 4: 0.86 req/s Try 5: 0.60 req/s 0.796 requests/second from 5 tries -> Speedup: 753.27% [2] Database script (huge) ------------------------------- With cache Starting benchmark for http://pubtour.thyrell.corp/manage/common/categories/index.php (5 tries)... Try 1: 0.48 req/s Try 2: 0.87 req/s Try 3: 0.36 req/s Try 4: 0.28 req/s Try 5: 0.28 req/s 0.454 requests/second from 5 tries Without cache Starting benchmark for http://pubtour.thyrell.corp/manage/common/categories/index.php (5 tries)... Try 1: 0.29 req/s Try 2: 0.28 req/s Try 3: 0.27 req/s Try 4: 0.27 req/s Try 5: 0.27 req/s 0.276 requests/second from 5 tries -> Speedup: 64.49% [3] binarycloud ------------------ With cache Starting benchmark for http://bc.thyrell.corp/index.php (5 tries)... Try 1: 5.54 req/s Try 2: 4.58 req/s Try 3: 4.55 req/s Try 4: 4.54 req/s Try 5: 4.12 req/s 4.666 requests/second from 5 tries Without cache Starting benchmark for http://bc.thyrell.corp/index.php (5 tries)... Try 1: 1.21 req/s Try 2: 2.10 req/s Try 3: 1.98 req/s Try 4: 1.04 req/s Try 5: 1.04 req/s 1.474 requests/second from 5 tries -> Speedup: 216.55% Resumee: --------------------------------------------------------------------------- IMHO, the ZendCache is VERY groovy :-) But it's also a expensive pice of software. So for performance critical and larger business applications it is a must, I think. The performance boosts for non-database sites is great. And with the binarycloud cache the pages will become real speed monsters, I guess :-) More detailed information is published in Zend and Techmetrix whitepapers. You can download them from Zend.com or with some screenshots of the above tested sites from: http://www.thyrell.de/info/downloads/cache_bench.zip. Especially the behviour of database accesss with zend cache is interesting. According to the paper is error free with cache (with their setup), but there are errors without. Andi _______________________________________________ binarycloud-dev mailing list bin...@li... http://lists.sourceforge.net/lists/listinfo/binarycloud-dev |
From: Alex B. <en...@tu...> - 2001-08-03 20:10:40
|
> An alternative free caching engine which compares with Zend is APC > http://apc.communityconnect.com/ > > When Rasmus was last here he recommended it.. so I assume it's quite good. > :) > > * Blake yeah, I need to have a look at it.... I've heard it's quite good, though not as polished as Zend Cache. _a > -----Original Message----- > From: Andreas Aderhold [mailto:a.a...@th...] > Sent: Friday, August 03, 2001 3:37 AM > To: bin...@li... > Subject: [binarycloud-dev] Zend Cache and binarycloud = more power > > > Hi, > > I just benched binarycloud with/without ZendCache in comparison > with other scripts. And the results are surprising. Actually > with Zend Cache enabled I gain huge performance boosts for > sites without database access. Read on. > > --------------------------------------------------------------------------- > Setup > --------------------------------------------------------------------------- > I ran the benchmarks on my devel box at home (crappy old box, with > lots of stuff running (sshd, samba......)): > > * Intel Pentium 100 > * 128MB PC100 RAM > * 15GIG IBM HDD (IDE) > * Linux Mandrake 8 (Kernel 2.4.3) > * Apache 1.3.19 > * MySQL 3.23.36 > * PHP4.0.4pl1 (Zend Launch Pad Edition) > * ZendOptimizer 1.1.0 > * ZendCache 1.1.0 > > PHP compile options (launchpad - out of the box) > ./configure > --prefix=/Zend > --with-apxs=/usr/local/apache/bin/apxs > --disable-debug > --with-config-file-path=/usr/local/Zend/etc > --without-db2 > --disable-sysvsem > --disable-sysvshm > --enable-xml > --enable-wddx > --with-mysql > --with-gd=shared,/usr/local/gd > --enable-gd-imgstrttf > --with-jpeg-dir=/usr/local/jpeg > --with-ttf=/usr/local/freetype > --with-t1lib=/usr/local/t1lib > --with-pgsql=shared,/usr/local/postgres > --with-imap=shared,/usr/local/imap > --with-ldap=shared,/usr/local/openldap > --with-ibm-db2=shared,/usr/local/db2' > --with-oci8=shared,/usr/local/oracle > > Zend Cache: > - validate timestamps is enabled (so chache checks everytime if > script changed, will perform even better if this is disabled in > production) > - reserved cache memory: 32Megs > > --------------------------------------------------------------------------- > Benched scripts > --------------------------------------------------------------------------- > [1] Thyrell homepage www.thyrell.de/home/index.php > A very simple script with sessions (language-detect) and smarty as template > engine (smarty cache disabled). It is just one template with basic > replacing work to do for the tpl engine (no section loops etc). No Database > access. > > [2] Database script (huge) > Manage script for category administation (unlimited depth for cats). > Script is currently not r2 based [but r2 friendly and after the > entitity/database stuff is out, it will be ported:)]. MySQL Database, > about 20 entries for categories with a depth of 4. > > This script is part of a more complex community application administrtion > I'm currently working on. So much things are not optimized and It's > overloaded with includes that will be removed/optimized. Escpecially > the layout for the admin interface is very grapic intensive. So currently > the loading time is very long. > > Basically: > - Request class, session class of r2 > - Metabase > - Smarty > - PEAR for error handling (exceptions and formwarnings) > > - The "manager" responsible for db-access uses recrusion to look up > the categories (in this example depth = 4). It also currently > executes 2 queries per query (COUNT(*))(Few template chunks with > descent template logic (looping over arrays). > > - Advanced browser scripting (mostly IE5 behaviours). > > > [3] binarycloud: index.php > The current (2001-03-08, 10am GMT) cvs r2/binarycloud/build/en/index.php. > No changes. > > --------------------------------------------------------------------------- > The results > --------------------------------------------------------------------------- > I ran the benches doubletwice and the results differ slightly. > > > [1] Thyrell homepage > ------------------------- > With cache > Starting benchmark for http://dev.thyrell.corp/home/index.php (5 tries)... > Try 1: 6.68 req/s > Try 2: 6.90 req/s > Try 3: 6.82 req/s > Try 4: 6.79 req/s > Try 5: 6.77 req/s > 6.792 requests/second from 5 tries > > Without cache > Starting benchmark for http://dev.thyrell.corp/home/index.php (5 tries)... > Try 1: 0.98 req/s > Try 2: 0.60 req/s > Try 3: 0.94 req/s > Try 4: 0.86 req/s > Try 5: 0.60 req/s > 0.796 requests/second from 5 tries > > -> Speedup: 753.27% > > > [2] Database script (huge) > ------------------------------- > With cache > Starting benchmark for > http://pubtour.thyrell.corp/manage/common/categories/index.php (5 tries)... > Try 1: 0.48 req/s > Try 2: 0.87 req/s > Try 3: 0.36 req/s > Try 4: 0.28 req/s > Try 5: 0.28 req/s > 0.454 requests/second from 5 tries > > Without cache > Starting benchmark for > http://pubtour.thyrell.corp/manage/common/categories/index.php (5 tries)... > Try 1: 0.29 req/s > Try 2: 0.28 req/s > Try 3: 0.27 req/s > Try 4: 0.27 req/s > Try 5: 0.27 req/s > 0.276 requests/second from 5 tries > > -> Speedup: 64.49% > > > [3] binarycloud > ------------------ > With cache > Starting benchmark for http://bc.thyrell.corp/index.php (5 tries)... > Try 1: 5.54 req/s > Try 2: 4.58 req/s > Try 3: 4.55 req/s > Try 4: 4.54 req/s > Try 5: 4.12 req/s > 4.666 requests/second from 5 tries > > Without cache > Starting benchmark for http://bc.thyrell.corp/index.php (5 tries)... > Try 1: 1.21 req/s > Try 2: 2.10 req/s > Try 3: 1.98 req/s > Try 4: 1.04 req/s > Try 5: 1.04 req/s > 1.474 requests/second from 5 tries > > -> Speedup: 216.55% > > > > Resumee: > --------------------------------------------------------------------------- > IMHO, the ZendCache is VERY groovy :-) But it's also a expensive pice of > software. So for performance critical and larger business applications it > is a must, I think. The performance boosts for non-database sites is great. > And with the binarycloud cache the pages will become real speed monsters, > I guess :-) > > More detailed information is published in Zend and Techmetrix whitepapers. > You can download them from Zend.com or with some screenshots of the above > tested sites from: http://www.thyrell.de/info/downloads/cache_bench.zip. > Especially the behviour of database accesss with zend cache is interesting. > According to the paper is error free with cache (with their setup), but > there are errors without. > > Andi > > > _______________________________________________ > binarycloud-dev mailing list > bin...@li... > http://lists.sourceforge.net/lists/listinfo/binarycloud-dev > > _______________________________________________ > binarycloud-dev mailing list > bin...@li... > http://lists.sourceforge.net/lists/listinfo/binarycloud-dev > -- alex black, ceo en...@tu... the turing studio, inc. http://www.turingstudio.com vox+510.666.0074 fax+510.666.0093 |
From: Michael K. <mi...@ta...> - 2001-08-03 20:42:51
|
Alex Black wrote: >>An alternative free caching engine which compares with Zend is APC >>http://apc.communityconnect.com/ >> >>When Rasmus was last here he recommended it.. so I assume it's quite good. >>:) >> >>* Blake >> > > yeah, I need to have a look at it.... I've heard it's quite good, though not > as polished as Zend Cache. > > _a Our own testing/usage/demos (in production on a couple machines too) has shown APC to give, on average, about 50% throughput/increase. Some pages are more, some are less. I was planning on testing the zend cache as well, but haven't had time yet. Zend cache caches code after it's been through the zend optimizer, which accounts for at least some of its extra speed increase over APC and Bware cache. Very little of what we do is NOT through a database, so we've never bothered testing code that wasn't database connected. APC may be faster than our 50% for non-db sites. I just can't think of too many sites I'd bother using PHP on that weren't db-enabled. APC and Bware won't work with the zend optimizer (well, I KNOW APC won't - I don't remember if bware does or not). The optimizer code is not open source, so it's kinda hard for APC to compete with them on that level, but the APC people claim many millions of page views per day, and significant capacity boosts because of their product. It's kinda hard to beat the price, for even a 50% speed up, imo. -------------------------------- Michael Kimsal http://www.tapinternet.com/php/ PHP Training Courses 1-877-745-3660 734-480-9961 |