Post by IP_CAM » Wed Jul 23, 2014 3:57 am

I did not manipulate my Source yet, I just disabled the Category Counting by the Admin Switch Function. I wanted to keep everything as DEFAULT as long as possible, exept for upmost important (minor) details, such as, getting that really outdated IE7-Carbage out of the Page Header. VqModding is implemented in the VeloMech Site, but empty again after putting the FEW Routine Lines from the (HUGE) GOOGLE-JS-VqMod directly into the Page Header. Therefore, I have NO VqModding active on any of the NEW 3 Test Sites, the Sites represent just the default Shops, ipcam.li uses the default MyISAM DB, but I indexed it by use of my MySQL Admin indexing-option. velomech.ch / bigmax.ch use the indexed Atomix InnoDB's, velomech.ch the external Google JS-Linking. But this does not 'always' seem to be the best Connection...

For some (yet) UNKNOWN reason, after my first Test Site-Comparance-Night, the DEFAULT OC DB driven Shop
suddenly GAINED on speed very much. It's possible, that the Indexing Routine was, kind of' delayed, but
anyway, I was rather surprised to see what happen. I may have to ask my Hoster, if he gave me more memory,
because I am stressing my Server so much...

I really have to add another really DEFAULT/DEFAULT Shop Site on Air, to find out on how an UNINDEXED MyISAM DB SHop will do..., I still have plenty of url's to go!

Another Detail, I found a remarkable Difference, at least for those, looking for TOP NUMBERS, either using

Code: Select all

http://velomech.ch/shop/
http://www.velomech.ch/shop/
on One Site, www works faster, on the other Sites, www works slower!
Or, depending on how it's linked, like ipcam.li / ip-cam.li , both URL's beeing routed to the same 'Place'...

Image

Image

Must have something to do, similar with getting images from 'another Line',
I have not rechecked my Configuration yet, but it's worth looking into it ...

will be back...

Ernie

My Github OC Site: https://github.com/IP-CAM
5'200 + FREE OC Extensions, on the World's largest private Github OC Repository Archive Site.


User avatar
Legendary Member

Posts

Joined
Tue Mar 04, 2014 1:37 am
Location - Switzerland

Post by victorj » Wed Jul 23, 2014 4:28 am

looks like images are deliverd through the orher domain, wich allows for double conections.

i did atach a vqmod that disables product count completely if switched of in admin, reducing databse queries drastic, give it a try.

nice to see that others are also working on improving speed, from bare install with some minor tweaks there a a great performange gain to achieve

well opencart does not let me upload files, will send it through PM

Koeltechnische deurrubbers eenvoudig online op maat bestellen.
Alle niet stekplichtige onderdelen zoals scharnieren, sloten, randverwarming en verlichting voor alle typen koelingen en vriezers.
https://koelcel-onderdelen.com


User avatar
Expert Member

Posts

Joined
Sat Jun 25, 2011 4:09 am
Location - Alkmaar Holland

Post by IP_CAM » Wed Jul 23, 2014 8:07 am

ok, here it is, OpenCart v.1.5.6.4 Default, equipped with 5'019 Products and a total of 1'300 different Product Images.
It is unmodified, and it uses no VqMod Installements. The MyISAM DB has not been indexed. Counting is switched off by Admin Selection.

Depending on how it's called, (with or without www.) it will produce different test results as well.
I had it installed under http://www...

Ernie

UPDATE:
See latest Test Result here, note the huge Difference in Page Size with enabled (modified) .htaccess!
http://www.ipc.li/os/oc1564default.jpg

The only real difference are the MISSING DB-Indexes in the hitline.info-Version. I still ask myself, why
'they' removed this important (should be default) Option, when upgrading the OC-Version?? It has been, according to my limited knowledge, part of older Versions...
Image

http://whichloadsfaster.com/

http://hitline.info/shop/index.php
Last edited by IP_CAM on Wed Jul 23, 2014 9:41 am, edited 10 times in total.

My Github OC Site: https://github.com/IP-CAM
5'200 + FREE OC Extensions, on the World's largest private Github OC Repository Archive Site.


User avatar
Legendary Member

Posts

Joined
Tue Mar 04, 2014 1:37 am
Location - Switzerland

Post by JAY6390 » Wed Jul 23, 2014 8:10 am

Using just the OC turbo mod on a site with 5,000 products, I get it to load a category in 0.9 seconds down from 15! This really should be part of the core. You can see this live at http://ocdemo.com/pagecache/desktops - Note that the time on the left in the blue bar at the bottom is the actual generation time of the page, and the value on the right is with my page cache as well

Image


User avatar
Guru Member

Posts

Joined
Wed May 26, 2010 11:47 pm
Location - United Kingdom

Post by melbagnato » Wed Jul 23, 2014 9:10 am

If you are willing to use a forked version of OC, take a look at this.

Definitely faster and worth a look.

- Mel

http://online.enterpriseconsulting.com.au

Site with OpenCart extensions & code downloads, many new extensions coming soon!
Follow us on twitter for more updates

Image


User avatar
Active Member

Posts

Joined
Wed Jan 13, 2010 1:39 pm
Location - Melbourne

Post by IP_CAM » Wed Jul 23, 2014 9:26 am

melbagnato wrote:If you are willing to use a forked version of OC, take a look at this.
- Mel
For whatever Reason, I could not get one single Mod to work , offered by villagedefrance, so, I gave up trying any longer. But this was at the beginning opf my 'OC carrer', I should look again asap...

Ernie

@victorj: So far, they let me do, what I do here, but, after all, it's an OpenSource Site. But, after all, it still takes an 'engaged' person to 'follow' this Theme, I imagine, we don't put everybody else out of business because of this. Mmmhhh

My Github OC Site: https://github.com/IP-CAM
5'200 + FREE OC Extensions, on the World's largest private Github OC Repository Archive Site.


User avatar
Legendary Member

Posts

Joined
Tue Mar 04, 2014 1:37 am
Location - Switzerland

Post by victorj » Wed Jul 23, 2014 10:27 pm

tryed to replicate with images server from non www on a www. site, by changing http adres in config files, even moved images to a domain with no trafic on same server and changing files in config to match, did not make a difference at oll here.

dont think we will put anyone out of business, unless everything found is combied into a mod and offered for free :-)

Koeltechnische deurrubbers eenvoudig online op maat bestellen.
Alle niet stekplichtige onderdelen zoals scharnieren, sloten, randverwarming en verlichting voor alle typen koelingen en vriezers.
https://koelcel-onderdelen.com


User avatar
Expert Member

Posts

Joined
Sat Jun 25, 2011 4:09 am
Location - Alkmaar Holland

Post by IP_CAM » Thu Jul 24, 2014 3:35 am

Hallo all,

I now have 10'200 Products, using 1'500 Product images, online.
I tested a Cat-Page with 569 Category-Products listed.
After a Test-Serie, it ended up, beeing just a little better
than a (5'019 Products) Stock OC Shop Cat-Page, loading it's 120 Product Listing.
If you click on the Mainpage Image, you'll get the Results.

http://www.velomech.ch/shop/

to let you know...,

queries: just about 4:1
overall: just about 5:1

Ernie

PS. External JS-Scripting by Google not activated at present.

and here is, what the velomech site does, when loading 120 Products:
http://www.webpagetest.org/video/view.p ... c6e&data=1

My Github OC Site: https://github.com/IP-CAM
5'200 + FREE OC Extensions, on the World's largest private Github OC Repository Archive Site.


User avatar
Legendary Member

Posts

Joined
Tue Mar 04, 2014 1:37 am
Location - Switzerland

Post by IP_CAM » Thu Jul 24, 2014 8:28 am

I got it all together now. This Place here:
http://hitline.info/shop/
is the slowest, but the only thing, making it SOOO slow, compared to this:
http://ip-cam.li/shop/
are the missing the DB-Indexes. The Rest is the same...

I even used 80x80px Thumbnail Images in hitline.info, in other Shops, I use 100x100px thumbs,
it would not help, despite of the 120 x 0.5-1.x kB load saved...

The other Sites in Tests all use ATOMIX InnoDB/Indexing Mod, my ipc.li Test Site
some other indexing/caching Goodies as well.

In Addition, HITLINE.INFO does ONLY LOAD 120 Products (of total 5'019 Products /1'300 Test Images), compared to the 300 Products (of total 10'200 Products /1'500 Test Images), loaded into all other SHOP Category-Pages tested.

CONCLUSION:

Regardless of using "MyISAM" or "InnoDB" Database-Engines, it's the Indexing, what makes
the HUGE PERFORMANCE-Difference, in the first Place. Everything else just add's to speed...

Ernie

I have indexed my IP-CAM.LI OC MyISAM DEFAULT DB by use of the MySql Admin, selecting ALL Tables and then click ONE TABLE INDEX-Button only. For some reason, it DID NOT IMMEDIATELY have an effect, therefore, clear all Temp File sections within the Software, clear yout the Browser Temp Section as well. It may have an effect, I really don't know. Give 'am some time, anyway, some time later, suddenly, the (indexed) shop started to act just about as it would have been indexed by Atomix superb Mod. I, honestly, rechecked, if I uploaded something into the wrong place...

Sometimes, it get's hard to keep oversight, if fooling around in 5 Shops at once!
Last edited by IP_CAM on Thu Jul 24, 2014 9:13 am, edited 7 times in total.

My Github OC Site: https://github.com/IP-CAM
5'200 + FREE OC Extensions, on the World's largest private Github OC Repository Archive Site.


User avatar
Legendary Member

Posts

Joined
Tue Mar 04, 2014 1:37 am
Location - Switzerland

Post by JAY6390 » Thu Jul 24, 2014 8:38 am

Out of interest, are they all on the same server Ernie?

Image


User avatar
Guru Member

Posts

Joined
Wed May 26, 2010 11:47 pm
Location - United Kingdom

Post by IP_CAM » Thu Jul 24, 2014 9:07 am

JAY6390 wrote:Out of interest, are they all on the same server Ernie?
I could not answer this, I only have ONE hoster, and I also have all my ~38 Domains under one Customer-Admin-Control and (assumingly )all running under one (shared with others) IP (Never checked on it). But my SITE-Directories may be placed 'internally' on different servers, who knows. HOSTPOINT is not a small Place, compared to Swiss Standards, they have a high reputation and very solid Base.

I just found out, that, by testing 4 Server at once, I noticed a massive general slowdown of all the Sites,
compared with a single Site tested. Possibly due to calling for so much data at the same time, trough the
same small wire...

http://www.webpagetest.org/video/view.p ... a776b7199e

Ernie

My Github OC Site: https://github.com/IP-CAM
5'200 + FREE OC Extensions, on the World's largest private Github OC Repository Archive Site.


User avatar
Legendary Member

Posts

Joined
Tue Mar 04, 2014 1:37 am
Location - Switzerland

Post by JAY6390 » Thu Jul 24, 2014 5:44 pm

Ah OK - Calling all 4 at once creating a slow down would suggest that at least some of them are on the same server

Image


User avatar
Guru Member

Posts

Joined
Wed May 26, 2010 11:47 pm
Location - United Kingdom


Post by melbagnato » Thu Jul 24, 2014 11:47 pm

I found that going through the SQL code looking for where statements and joins gave me an understanding of what indexes to add to the tables. One of my client's reports went from 90 minutes, to 3 seconds.

SQL can be very wasteful if the code isn't written right, or the database isn't structured or optimised well.

- Mel


Sent from my T100TA using Tapatalk

http://online.enterpriseconsulting.com.au

Site with OpenCart extensions & code downloads, many new extensions coming soon!
Follow us on twitter for more updates

Image


User avatar
Active Member

Posts

Joined
Wed Jan 13, 2010 1:39 pm
Location - Melbourne

Post by IP_CAM » Fri Jul 25, 2014 3:58 am

melbagnato wrote:If you are willing to use a forked version of OC, take a look at this.Definitely faster and worth a look.- Mel
... you made me curious anyway , so I tested the Thing:

http://www.evelo.li/shop/
click on Welcomepage images to see (single-Site tested) video-tests of both Sites, e-velo.li with enabled and with 'manually' removed Counting-Routines. They all contain 120 Products, displayed on a Cat-Page.

Looks very nice, and it contains some usefull things. I would advise Newbies to check on it, it would save 'em some work. But in Performance, as it comes, to me, it makes absolutely NO difference to my hitline.info default Test-Shop.

Before trying to further 'improve' the System, I manually removed all count-related stuff from the 3 or 4 Files involved, and tried again. Howwww, what a difference on a clean default nonindexed DB OC Shop, containing 5'019 Products.

SO much for the moment, I keep testing...

Ernie

Enclosed, you find a zip-file, containing some content. An image is enclosed to explain,
what happen. An extensionless File was put into the Sub, probably, during the installation, and I would not
even have checked on/for it, if I would not have tested an OLD VqMod Content (an old cache mod!), asking for a subdirectory, seemingly not existing. Trying to create one, without success, I found this extension-less File,
using the Name I wanted to create my SubDir with...
It's even full with (some cache file) code...

Any explanation, does ot eventually belong there?
If noone knows, I will ask the frenchmen...

Ernie

My Github OC Site: https://github.com/IP-CAM
5'200 + FREE OC Extensions, on the World's largest private Github OC Repository Archive Site.


User avatar
Legendary Member

Posts

Joined
Tue Mar 04, 2014 1:37 am
Location - Switzerland

Post by melbagnato » Fri Jul 25, 2014 6:44 am

Hi ernie,

I'll take a look at the .zip file later today and get back to you.

Also, with regards to the optimisations, I have a wide range of clients who have different product counts in their store. These range from 600,000+ (site selling ebooks), to just 50+ (new store to sell clothing). Some modules had behaved worse when there are high stock levels, others behave worse when the query was too complex (like some of the aggregated reports). Either way, OpenCart's generic database setup & interaction isn't a one-size fits all.

I find that the optimisation check list I go through provides different benefits to different database sizes. Sometimes I'll write a custom mod to change the query to be specific to the needs of the customer, other times I'll change their setup on Amazon Web Services (AWS) to take better advantage of the other networking elements (like CDN, cache, load balancing, elastic demand etc.).

- Mel

http://online.enterpriseconsulting.com.au

Site with OpenCart extensions & code downloads, many new extensions coming soon!
Follow us on twitter for more updates

Image


User avatar
Active Member

Posts

Joined
Wed Jan 13, 2010 1:39 pm
Location - Melbourne

Post by IP_CAM » Fri Jul 25, 2014 11:44 am

unfortunately, I was forced to write my first VqMod, but it had to be. Compared with what I felt
about on how a VqMod functions, really to disable Counting in early Stage, by VqMod,
could not work, it would just prevent the Numbers from beeing displayed, because the Counting
has been made much earlier in the Page-Building Process, VqMod- Interception comes last!!!
Early Count-Processes could therefore not be influenced by an Overwrite-Engine-type-Mod.

It seems that I was right. Despite of 3 more Requests required and therefore 6 KB (+3images)
more total Filesize, to build a similar Page, E-Velo.li seems to work faster, in most tests anyway.

Anyway, I assume that I was wrong as well, when claiming, that the bIggest Speed-Improvement would be
reached by Indexing the DB. The upmost biggest Gain is to just disable Counting. Exept in Places, where
it's used to define/display Pagebreak related Information.

It's early morning, I have to get some sleep...

You all have a great day

Ernie

http://www.webpagetest.org/video/view.p ... 5e19222909

My Github OC Site: https://github.com/IP-CAM
5'200 + FREE OC Extensions, on the World's largest private Github OC Repository Archive Site.


User avatar
Legendary Member

Posts

Joined
Tue Mar 04, 2014 1:37 am
Location - Switzerland

Post by melbagnato » Fri Jul 25, 2014 4:23 pm

Hi Ernie,

don't forget that counting isn't an issue if you have cache enabled on your database, and the fields are indexed. I do agree that it usually isn't worth the value of the query though.

- Mel

http://online.enterpriseconsulting.com.au

Site with OpenCart extensions & code downloads, many new extensions coming soon!
Follow us on twitter for more updates

Image


User avatar
Active Member

Posts

Joined
Wed Jan 13, 2010 1:39 pm
Location - Melbourne

Post by IP_CAM » Fri Jul 25, 2014 8:19 pm

melbagnato wrote:Hi Ernie, don't forget that counting isn't an issue if you have cache enabled on your database, and the fields are indexed. I do agree that it usually isn't worth the value of the query though.- Mel
I know, but I wanted to avoid playing around with caching or indexing, before getting my tests done, on my way find answers to all this. I will be back tonight for further testing.

Ernie

My Github OC Site: https://github.com/IP-CAM
5'200 + FREE OC Extensions, on the World's largest private Github OC Repository Archive Site.


User avatar
Legendary Member

Posts

Joined
Tue Mar 04, 2014 1:37 am
Location - Switzerland

Post by JAY6390 » Sat Jul 26, 2014 1:26 am

Hey Ernie

What speed hits are you seeing with vQmod? And why don't you like it? Me and Qphoria have put a lot of time and hard work into that :'(

Image


User avatar
Guru Member

Posts

Joined
Wed May 26, 2010 11:47 pm
Location - United Kingdom

Post by IP_CAM » Sat Jul 26, 2014 4:14 am

Just don't get me wrong! I admire your work very much, I am very sorry, we did not have it, 15 Years ago, because it would have changed the (Perl) world. At least for Boozobrains like me, not knowing nothing about anything. But I'm coming from a scene, where everybody tried to 'overpower' all others, so finally, it was a matter of microseconds, deciding on win or loose.

VqModding is, to me, somehow similar to 'Active-X', from the Scripters Side, one does not really fall in love with 'em, but evenybody is using 'em, since it makes Live just a little 'nicer' for all. But when it comes to speed, if one want's to really compete with the 'Big-Names', no compromises can be made. For those, eager to get a top Shop, I declare my 'point of view', not for the others. To prove, that I'm neutral, I even spend the past 15 hours, trying to get the best possible Mod created, without (possibly) turning off Selection- and Pagebreak-Functions as well (then it really fires off...). It's doing etaxtly the same as the built-in Mod's, done by the hardcoded Versions. So far, it works well, and it proves, to me, again, that my (similar built) hardcoded Site is, like in the test, mostly (a few microseconds!) faster, than the One, equipped with the one (remove counting) VqMod only.

Here is the latest 2-Site Video Test Report, on ipc.li/shop/ , more tech-details can be found on the Main Page.
The middle Site displayed has removed Product Counting from SOURCE , the left/right Site has enabled/disabled VQMOD Product Counting.

http://www.webpagetest.org/video/view.p ... f214d3e68d

will be online later...

Ernie

@melbagnato:
the frenchmen was wrong, his thing comes out , beeing slowest, despite of the same 'Configuration' as another one.
Still, it's very nice and with love created, I gues I'll stay with it for future projects...
Last edited by IP_CAM on Sat Jul 26, 2014 9:35 am, edited 5 times in total.

My Github OC Site: https://github.com/IP-CAM
5'200 + FREE OC Extensions, on the World's largest private Github OC Repository Archive Site.


User avatar
Legendary Member

Posts

Joined
Tue Mar 04, 2014 1:37 am
Location - Switzerland
Who is online

Users browsing this forum: JackSm, paulfeakins and 24 guests