Hardware and performance

 
 
Picture of shrikant gakhare
No performance Improvement after migration to new hardware
 

We are running MOODLE 1.9.4 on old hardware configuration given below. The server has been running since last 7 years. We witnessed performance degradation in the last six months and decided to replace old server with new (configuration given below) doubting hardware issue. Site has been migrated to new server (under testing) but performance is same like old one even when only one user is accessing it (it is still under testing).  Static pages load faster but pages with query take time to load.

Number of registered users in  current set-up in production- 100000. (expected to grow upwards at 20% every year). Can MOODLE and MySQL support these many users without affecting the performance?? Concurrent users are not more than 300. Do you suggest going for performance benchmarking? Pls suggest some tool for same?

We doubt the performance of MySQL (5.1.37 from latest AMP package). 

OS- Solaris 11, PHP- 5.2.17, Apache- 2.2.22

Existing Configuration: (currently LIVE)

Sys-em Configuration:  Sun Microsystems   
System Model : Sun Fire 480R 
System clock frequency: 150 MHz 
Serial No:0539AM0057 
Warranty Expiry:31/07/2010 
Release: 5.10 
Kernel architecture: sun4u 
Application architecture: sparc 
Hardware provider: Sun_Microsystems 
Kernel version: SunOS 5.10 Generic_139555-08 
Memory size: 8192 Megabytes 
System :SunOS 10v 
Harddisk Size :73.40GB*2 
OEM# = 0 
Origin# = 1 
NumCPU = 2

Procured Configuration (NEW)

LINE # Part # / Line Type Part Description Qty
1
7102570
Netra SPARC T4-1 server: model family
1
1.2
7103495
Netra SPARC T4-1 server: 2 RU AC base chassis with 1 SPARC T4 4-core processor 2.85GHz
1
1.4
333A-25-10-IN Power cord: India, 2.5 meters, IS1293 plug, C13 connector, 10 A (for factory installation)
2
1.6
7100648
Two 4 GB DDR3-1066 registered DIMMs (for factory installation)
4
1.8
RA-SS2NF-300G10K2 One 300 GB 10000 rpm 2.5-inch SAS-2 HDD with mounting bracket (for factory installation)
4
1.1
SG-SAS6-R-INT-Z Sun Storage 6 Gb SAS PCIe RAID HBA, Internal: 8 port and 512 MB cache (for factory installation)
1
1.12
8224A-N Netra X4250 SAS Cable Kit
1
1.14
7101664
Oracle Solaris 11, Oracle VM Server for SPARC, and Oracle Electronic Prognostics preinstall (for factory installation)
1
1.16
7901A-4 19-inch 2-post rackmount kit (for factory installation)
1

 

 

 
Average of ratings: -
Picture of Andrew Nicols
Re: No performance Improvement after migration to new hardware
Group DevelopersGroup Moodle HQGroup Particularly helpful MoodlersGroup Testers

Hi Shrikant Gakhare,

In general, I'd really recommend reading the advice at the top of this forum first and providing as much information as possible. You don't say what you've done so far to tune the software you have installed for example, and whether you are using a PHP accelerator. It would be difficult to offer any specific advice without knowing this information first.

You also really need to identify your bottlenecks before going any further - both on the old hardware and the new hardware if this is at all possible.

You do need to do some benchmarking - both client-side and also server side. My memory is a little hazy on the relevant software for the server side. On linux, I frequently use dstat but I'm not sure what the Solaris equivelant is - perhaps prstat? If you're using ZFS, there are tools to check your ZFS performance. Similarly, there are NFS tools available too.

For MySQL, there are some reasonable performance tuning scripts to give you advice. For example https://github.com/rackerhacker/MySQLTuner-perl. Make sure that you understand the suggestions, and consider the interactions with Apache too!

On the client side, start looking at things like jmeter. There is a Moodle 2 jmeter script, but I don't think it's available for Moodle 1.9. I wouldn't bother with ab - it doesn't offer any real value.

Typically though, replacing hardware is the last port of call after all other avenues have been exhausted, or it has been conclusively proven that the hardware just doesn't cut it and more and newer hardware really is required.

Looking at your new hardware spec, I assume that your moodledata is on a NAS mount of some kind given that you only have a single 300GB internal disk, and no external RAID controller. I'd investigate this to see whether you're hitting any bottlenecks here. We found that our NFS server was one of our biggest bottlenecks to concurrency with Moodle 1.9.

We used to run a set of six SunFire T1000s for Moodle and found their performance to be adequate at best. We initially ran Solaris, but during our testing phase we switched to Debian, largely due to availability of packages, and ease of management (we're a Debian shop). Since you are running Solaris, you're probably pretty familiar with its quirks and pitfalls though.

One thing we specifically found with the SPARC architecture was that the MySQL connection time was far higher than a 5 year old Dell PE2850 on intel architecture and there was nothing we could do to improve matters there. I think it was in the region of ten times slower than the Intel architecture at just connecting to MySQL. We are only talking milliseconds really (3 vs 30 IIRC) but under high concurrency it can add up. We decided it wasn't really worth the worry for our loading and given the number of servers we were running.

If you really are tied into running sparc architecture, and running Solaris on it, then one thing you may not be aware of is the TX range of Solaris packages. It's been a while since I last looked at them, but when Sun ran things, the T series had a set of packages compiled for them which leveraged processor and system specific optimisations. One of these for example is the hardware SSL engine found on many of the T series motherboards. I'm not sure whether the Netras have these and my memory is a little hazy on the area, but it may be something to look into. Alternatively, look at compiling your own versions with the various patches. If I recall, the openCSW project was also looking at doing the same - it was discussed at some point when I was a maintainer on the project. 

Really though, we need much more information if we are to help. At the very minimum:

  • PHP accelerator in use
  • Apache tuning performed
  • MySQL tuning performed

Best wishes,

Andrew

 
Average of ratings:Useful (2)
Picture of shrikant gakhare
Re: No performance Improvement after migration to new hardware
 

Thanks Andrew!

In fact our current installation which is live and people are using, shows better performance that new one. Existing is live, and URL may be shared.

And new one, we testing with just one user, and its performance is damn slow.

Please see the hardware configuration below

Existing Configuration: (OLD- Presently live):

 

System Configuration:  Sun Microsystems   
System Model : Sun Fire 480R 
System clock frequency: 150 MHz 
Serial No:0539AM0057 
Warranty Expiry:31/07/2010 
Release: 5.10 
Kernel architecture: sun4u 
Application architecture: sparc 
Hardware provider: Sun_Microsystems 
Kernel version: SunOS 5.10 Generic_139555-08 
Memory size: 8192 Megabytes 
System :SunOS 10v 
Harddisk Size :73.40GB*2 
OEM# = 0 
Origin# = 1 
NumCPU = 2

 

Procured Configuration (NEW-under testing)

 
LINE # Part # / Line Type Part Description Qty
1
7102570
Netra SPARC T4-1 server: model family
1
1.2
7103495
Netra SPARC T4-1 server: 2 RU AC base chassis with 1 SPARC T4 4-core processor 2.85GHz
1
1.4
333A-25-10-IN Power cord: India, 2.5 meters, IS1293 plug, C13 connector, 10 A (for factory installation)
2
1.6
7100648
Two 4 GB DDR3-1066 registered DIMMs (for factory installation)
4
1.8
RA-SS2NF-300G10K2 One 300 GB 10000 rpm 2.5-inch SAS-2 HDD with mounting bracket (for factory installation)
4
1.1
SG-SAS6-R-INT-Z Sun Storage 6 Gb SAS PCIe RAID HBA, Internal: 8 port and 512 MB cache (for factory installation)
1
1.12
8224A-N Netra X4250 SAS Cable Kit
1
1.14
7101664
Oracle Solaris 11, Oracle VM Server for SPARC, and Oracle Electronic Prognostics preinstall (for factory installation)
1
1.16
7901A-4 19-inch 2-post rackmount kit (for factory installation)
1

 Apart from hardware, everything is same. Except that we installed latest versions of AMP. A- 2.2.22  M- 5.1.37 and P-5.2.17.

We ported application as it is. In fact in the live set-up, we have not used PHP acclerator and its performance for one user is best as compared to the new we are testing. This is something strange.

Please guide!

P.S- We can have a call and you may guide us. Other Moodlers on forum can join  or if they can guide us, please let me know thru this forum. 

 

 
Average of ratings: -
Picture of Andrew Nicols
Re: No performance Improvement after migration to new hardware
Group DevelopersGroup Moodle HQGroup Particularly helpful MoodlersGroup Testers

Hi Shrikant,

As I mentioned above, you need to identify your bottlenecks first.

You also really need to install a PHP accelerator.

We can't offer any proper advice on these forums without the basic information I described before.

Andrew

 
Average of ratings:Useful (1)
Picture of Visvanath Ratnaweera
Re: No performance Improvement after migration to new hardware
Group Particularly helpful Moodlers
@all

It looks like that the discussion continues here: "Same application,different hardware,No change in performance" https://moodle.org/mod/forum/discuss.php?d=218782.
 
Average of ratings: -
Picture of shrikant gakhare
Re: No performance Improvement after migration to new hardware
 

Yes Sir! Writing to Helen to delete those threads.

 
Average of ratings: -
Picture of Helen Foster
Re: deleting posts
Group DevelopersGroup Documentation writersGroup Moodle HQGroup Particularly helpful MoodlersGroup Testers

As I replied to Shrikant in an email, I am not going to delete posts which others have taken the time to provide helpful replies to.

Shrikant, I hope you have learnt to keep to one discussion thread per question topic in future.

 
Average of ratings:Useful (1)
Picture of shrikant gakhare
Re: deleting posts
 

Thanks All

Issue resolved and this thread is closed.

regards

Shrikant

 
Average of ratings: -