new "performance test" accuracy

@soaringeagle
11 years ago
3,304 posts

i ran the performance test
it suggests a 2 core 2.8 ghz 2 gig vps with ssd drives
i have an 8 core 3.8ghz 8 gig dedicated server with sataII drives
the only diference isthe drives and it rates a 40 out of 1000 and says it will run very slowly
the host offers a maximim of 600 gigs in a ssd drive at a 125 additional cost if i upgraded to 2 ssd drives it would double my cost to over 500 a month for the server
i live on 700 a month

so how accurate is this could it really be the ssd drives making that big a difference

i did see slow connect and wait times in the pingdom tools tests

but youd think with the cpu and ram being 4 times the recomended it would rate better then 40 out of 1000

ps during the import im using over 660 gigs 73% of a 1tb drive
once live ill be using less...temporarily
and if things take off it wont be long till i need a tb

i looked at the jr recomended servers
they wanted 4 times as much money for 1/2 the power

besides the drives what else can i tweak to get better performance
would upgrading mysql from 5.5 to 5.6 help

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities
updated by @soaringeagle: 02/04/15 12:34:58PM

@brian
11 years ago
10,148 posts

If you're getting only 40 then something is wrong with your server - you don't have an 8 core machine OR you are running on "virtual" cores that are very, very slow.

My 1 CPU Linode personal VPS I use gets ~900.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
11 years ago
3,304 posts

its 8 cores and there is 1 thing the timeouts and max exe times on mysql and litespeed were all maxed out to allow the huge ning imports to run without stopping every 3 minutes
im upgrading mysql
and considering the smaller ssd drive as a cache drive

i installed lucid fusion last night hoping it would be a better search option but as soon as it was started b4 any configuration or anything the memory used was over 4 times as much so i shut that down till i talk to thier tech ppl and will more then likely remove it

question
the test.. since it runs in the browser if my home routers acting up again that could affect the results?
with the hundreds of gigs ive transfered the last month im surprised the router hasnt burst into flames

i think ill have someone else run the test from a diferent pc just to compare

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
11 years ago
10,148 posts

The performance test runs completely on your server - your browser has no impact on it.

Based on the performance issues I've seen you post over the weeks, I'm really suspecting your server is not setup properly OR you are not actually getting what you are paying for. If you are on a dedicated 8 core server with 3.8ghz CPUs, with 8G RAM you should never, ever get 40. The only way that would happen is if your disk subsystem was so slow that it brought everything way down.

One thing I can do for the next release is break out the CPU results and Disk results and show how they contributed to the overall score. But if your disks are so slow they are pulling your server down to a score of 40, you need to replace them.

How much are you paying each month? What are your disk requirements as far as space goes?

Hope this helps!

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
11 years ago
3,304 posts

im paying like 204 the current disk usage is 73-76% i think of 1 tb but thats cause im uploading my sites arechives into 2 diferent folders i had stared a 3rd but stopped it i might delete thgat and test

but i looked into your recomended they i think had a max of 350 gig drive i could be wrong but it wasnt big the 2 gig ram 2 core i think was over 400

they are doing some optimizxation tests trying to find the cause
they think maybe the ramdrive cache might have been configured wrong or something
but they are working on it trying to work out the issue

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
11 years ago
3,304 posts

one thing we found is although they told me ramdrive was mounted at a particular path and i configured the server that way no ramdrive was mounted
so that could be a significant cause

i upgraded mysql and litespeed and requested they mount the ramdrive
pindom tools testing from a freinds pc had 2-5 second loads everytime so thats definately a good speed for a kinda heavy page
but i think the ramdrive being set right shoiuld really improve that score

is there a details log i mean it says it runs several tests to determine the score
a details log that tells you wich tests lowered the score would surely help you find the cause and fix it

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
11 years ago
10,148 posts

soaringeagle:
one thing we found is although they told me ramdrive was mounted at a particular path and i configured the server that way no ramdrive was mounted
so that could be a significant cause

i upgraded mysql and litespeed and requested they mount the ramdrive
pindom tools testing from a freinds pc had 2-5 second loads everytime so thats definately a good speed for a kinda heavy page
but i think the ramdrive being set right shoiuld really improve that score

is there a details log i mean it says it runs several tests to determine the score
a details log that tells you wich tests lowered the score would surely help you find the cause and fix it

A 2 - 5 second page load time is really not good - not at all. Your page load should at max 1 second with Jamroom - unless you're just doing something really crazy it should never take 5 seconds.

What are you using a ramdrive? The only thing that would help on a server by using a ram drive would be to set MySQL's /tmp partition to the ram drive, but then if you get too busy and run our of space on that ramdrive you will have issues.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
11 years ago
3,304 posts

cache on ramdrive and the front page is large now because of html not being disabled on theforum list
im goinbtg to addthat code later todaty wich should reduce the page size by a 3rd at least

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
11 years ago
10,148 posts

soaringeagle:
cache on ramdrive and the front page is large now because of html not being disabled on theforum list
im goinbtg to addthat code later todaty wich should reduce the page size by a 3rd at least

Jamroom's cache is in the database, so JR is not going to use an HTML drive unless you've symlink'd the jamroom/data/cache directory to your ramdrive - just be very careful you don't run out of space or Jamroom will not function correctly.

Just a recommendation too - the $160 a month Linode plan here:

https://www.linode.com/pricing

with 384G SSD, 8 CPU cores, 16G RAM and 16TB transfer would be an insanely fast platform for Jamroom - seriously. It would be cheaper than what you have now, but would run circles around the performance. Just an FYI. I'm a huge fan of Linode and their platform - I've never been on anything faster for hosting.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
11 years ago
3,304 posts

hmm checking some test tools seems the images and transfers arent geting compressed

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
11 years ago
3,304 posts

brian:

soaringeagle:
cache on ramdrive and the front page is large now because of html not being disabled on theforum list
im goinbtg to addthat code later todaty wich should reduce the page size by a 3rd at least

Jamroom's cache is in the database, so JR is not going to use an HTML drive unless you've symlink'd the jamroom/data/cache directory to your ramdrive - just be very careful you don't run out of space or Jamroom will not function correctly.

Just a recommendation too - the $160 a month Linode plan here:

https://www.linode.com/pricing

with 384G SSD, 8 CPU cores, 16G RAM and 16TB transfer would be an insanely fast platform for Jamroom - seriously. It would be cheaper than what you have now, but would run circles around the performance. Just an FYI. I'm a huge fan of Linode and their platform - I've never been on anything faster for hosting.

the problem being the drivespace is nowherenear sufficient

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
11 years ago
10,148 posts

Got it... well hopefully you can get to the bottom of what's going on with your current setup.

I'm working on updates to the performance tool that will break it down by processor, database and filesystem so it will give you a better idea of where the server might be struggling at.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
11 years ago
3,304 posts

to get the minumum hard drive space i need right now id have to upgrade to the 20 cote plan at double the cost
they arent scalable you cant just add more drive space without adding more cores and ram and doubling and trippling theprice everytime u need another 120 gigs
i need 1 that alows u to just add more drive space as needed nothing else
add more ram as needed
and add another 30 a month not another 200 a month with each upgrade

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
11 years ago
3,304 posts

are ssd essential
all the ssd drives are outragiously expensive for the size i need
could it be configured somehow to use a couple hundred gig sdd for often accessed resources and the 1 tb sata for vids music etc
although just the photos alones around 90gig.. 97 i think

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
11 years ago
10,148 posts

soaringeagle:
are ssd essential
all the ssd drives are outragiously expensive for the size i need
could it be configured somehow to use a couple hundred gig sdd for often accessed resources and the 1 tb sata for vids music etc
although just the photos alones around 90gig.. 97 i think

No - SSD's are not essential, but they are around 100 times faster in random access than a regular spinning disk drive. Since web servers by their nature tend to be lots of small random accesses, SSD's are a fairly big benefit. However I've setup several dedicated servers for customers that are using 2TB disk drives and they are very fast as well, so you should be OK.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net
updated by @brian: 12/31/14 11:07:36AM

@soaringeagle
11 years ago
3,304 posts

im a happy camper now
i got a server with 1 tb ssd .. double the ram (threw that in free)
faster port speed same cpu .. scalable can add more cores etc more drives as needed

20 dollars more a month
only diference is its not managed although they will do migration, and configuratuion for the 1st month

i expect that performance score to jump from 40 to 940

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
11 years ago
3,304 posts

Single Quad-Core E3-1240 3.4Ghz 8M 1333 w/ HT
16GB RAM
1TB Samsung SSD - Main Drive
1TB SATA2 Enterprise 7200RPM
CentOS 6 64Bit
cPanel/WHM Standalone - Requires CentOS
No Firewall
Customer Managed - w/ Ping Monitoring
/28 IP Space 11 Usable - SSL Certs
Piscataway, New Jersey
Automated DDOS Detection + Null Route
10TB + Free Incoming
1000Mbps Uplink

$224.00/Month

that im happy with
they will do full migration and config making sure its all set up right and running smoothly
after the 1st month its self managed

but this should give me a 940
im guessing
im assumming imports will be faster too and uploads probly not alot but ifit helpsa at all that would be great

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
11 years ago
10,148 posts

If that is a bare metal server, you should be around 1500 or so - definitely higher than the 2.8GHz I test with here. Note that the performance test only tests a SINGLE processor. I'm not sure yet how I want to go about handling multiple processors, but I'll figure something out since a quad core server is going to be able to handle more than a single CPU at the same speed.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
11 years ago
3,304 posts

also it needs an output and log of what the heck the score means as far as individual tests

like it needs to tell you the 40 is based on
a 300 for memorty access
700 for net connect time
200 for disk access speed
40 for mysql querries

sort of like the windows 7+ or was it vista + rating that tells you what needs to be upgraded to improve your performance score

just saying 40 oyut of 10000 doesnt give u a single clue where to look for performance improvements

its a great tool no doubt and im happy to have found out now that a server change will help that much more
but it really needs some output wether in the page a report of each test run.. or a log file

put it on the back burner tho i think the seemless pagingation and site builder finalizations more important
but what id say is worth considering is
something along the lines of
now testing ram access speeedd..... done results 33 ms performance value 800/1000
now testing disk access speed..... done results 243 ms performance value 600/1000
now testing network port speed
now testing ssl verification
now testing 1st byte transfer time

etc etc

so it tells you what its testing qas its testing it gives you the actual results and how those results affect the overal ranking

that....that would be an amazingly useful tool
then you can look say ah ok i need to upgrade the network card or get a faster cpu or add more ram etc etc

it doesnt just say...your server sux you should consider 1 thats actualy got 1/4 ofthe power you already have.. but instead tell you what about your severs lacking and offer suggestions on improving
like consider getting a server with a faster data transfer rate or upgrade your network card..etc

cause when i had the server admins run the test they were like..ok 40..but what the hecks that 40 mean...

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
11 years ago
10,148 posts

soaringeagle:
also it needs an output and log of what the heck the score means as far as individual tests

like it needs to tell you the 40 is based on
a 300 for memorty access
700 for net connect time
200 for disk access speed
40 for mysql querries

sort of like the windows 7+ or was it vista + rating that tells you what needs to be upgraded to improve your performance score

just saying 40 oyut of 10000 doesnt give u a single clue where to look for performance improvements

its a great tool no doubt and im happy to have found out now that a server change will help that much more
but it really needs some output wether in the page a report of each test run.. or a log file

put it on the back burner tho i think the seemless pagingation and site builder finalizations more important
but what id say is worth considering is
something along the lines of
now testing ram access speeedd..... done results 33 ms performance value 800/1000
now testing disk access speed..... done results 243 ms performance value 600/1000
now testing network port speed
now testing ssl verification
now testing 1st byte transfer time

etc etc

so it tells you what its testing qas its testing it gives you the actual results and how those results affect the overal ranking

that....that would be an amazingly useful tool
then you can look say ah ok i need to upgrade the network card or get a faster cpu or add more ram etc etc

it doesnt just say...your server sux you should consider 1 thats actualy got 1/4 ofthe power you already have.. but instead tell you what about your severs lacking and offer suggestions on improving
like consider getting a server with a faster data transfer rate or upgrade your network card..etc

cause when i had the server admins run the test they were like..ok 40..but what the hecks that 40 mean...

Well... 40 out 1,000 means your server on cpu, database and filesystem tests (combined) scored 4% of our baseline (which is shown on the screen). The next version will break down processor, database and filesystem individually so you can see which one is pulling you down if it is.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
11 years ago
3,304 posts

yea thats all im askin is a breakdown

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
10 years ago
3,304 posts

ok i like the improvements in the new performance tests however now the performance test and dashboard seem to produce wildly diferent results
remember i was getting a score between 48-42 (dropped to 17 or 18 during migration)
now im getting a 57 (still on old server mind you had to replace the ip block due to blacklisted main ips)(will have new server results relatively soon)

although my ratings now higher at 57 the cpu reported in both the dashboard and performance tester are way off at 0.43 when they are 3.3 or 3.4 ghz i forget

it shows 8 cores just the ghz measurements way off

good news is the mysql ratings way way above baseline it was probAbly the filesystem affectine my score so badly (new server flies when tested on my wb blogs it was instant u couldnt see it load it was too fast couldnt test jr cause some datsa needed resyncing)

ill attach a ss
basicly.. big improvement overall in the tester but.. inacurate readings on the core ghz

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities
updated by @soaringeagle: 01/02/15 12:58:21PM

@soaringeagle
10 years ago
3,304 posts

just to be certain 1st thing after dns change will be atest on new server as comparison..

but since the dashboard used to read corectluy b4 the core updatr and doesnt now i suspect its the tester still needs a tweak

overall tho..excelent work once that bugs fixed

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
10 years ago
10,148 posts

You're going to get slightly different readings almost every time, so you want to run it multiple times to get a "feel" for what the average score is on your server.

Your MySQL performance is killing you though - 138 seconds to run the test is very, very slow. Have you tuned your MySQL install at all?

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net
updated by @brian: 01/02/15 01:21:37PM

@soaringeagle
10 years ago
3,304 posts

could that have to do with the max exec times i hadto set for imports

dns is propegating to the new server so should have sa new test run then
but isnt the odd cpu results a concern

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
10 years ago
10,148 posts

soaringeagle:
could that have to do with the max exec times i hadto set for imports

I don't think so. The only thing I can think of that would make the score for MySQL be so low is:

1) MySQL is running on another server over a very slow network link (not likely in your case)

2) MySQL is running with the "defaults" for settings - i.e. you've not set your key_buffer size, or any of the InnoDB settings. Here is the [ mysqld ] settings section from our Jamroom site here (tuned for a 4Gig Server):

key_buffer              = 256M
max_allowed_packet      = 32M
thread_stack            = 192K
thread_cache_size       = 8
table_cache             = 1536
table_definition_cache  = 1536
open_files_limit        = 2048
tmp_table_size          = 256M
max_heap_table_size     = 256M
thread_concurrency      = 8
sort_buffer_size        = 1M
read_buffer_size        = 1M
read_rnd_buffer_size    = 768K
max_connections         = 200
innodb_buffer_pool_size = 512M
innodb_flush_method     = O_DIRECT

On our master server here we run through the DB tests in about 3.5 seconds.

3) There something erroring like crazy on the MySQL server which is keeping it from properly being able to handle requests.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
10 years ago
3,304 posts

ok as soon as propegations done on the new server ill double check those actualy i can set them as soon as the php modules are done reinstalling but what would u recomend on a 16 gig ram server

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
10 years ago
10,148 posts

for 16Gig I would go:

key_buffer              = 1024M
max_allowed_packet      = 32M
thread_stack            = 192K
thread_cache_size       = 8
table_cache             = 1536
table_definition_cache  = 1536
open_files_limit        = 2048
tmp_table_size          = 256M
max_heap_table_size     = 256M
thread_concurrency      = 8
sort_buffer_size        = 1M
read_buffer_size        = 1M
read_rnd_buffer_size    = 768K
max_connections         = 200
innodb_buffer_pool_size = 1024M
innodb_flush_method     = O_DIRECT

Should be good.

Hope this helps!

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
10 years ago
3,304 posts

helps alot thnx
ill set that in a few min they are resyncing the db so dont think its a wisde idea toi change settings till after

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
10 years ago
3,304 posts

so the lower numbers best right im gettinbg a 300 now but mysql is still at 20.something
were setting those setting
but the cpu ghz speeds are being misread

says 1.6 instead of 3.8

but thats better then the ,34 or whatever the old server said

i think as soon as the mysql settings get tweaked it will fly

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
10 years ago
3,304 posts

wow
even before tuning the database its made a massive diference the ning imports are flying now i think it will import more in an hour then it used to in days

still tge database scores around a 20 but getting those settings tweaked and its going to fly

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
10 years ago
3,304 posts

its alot faster but i know i can get alot more out of it
any more mysql tweaks u can think of

oh
and the 64k photo import that used to take days was done on 1/2 an hour or less
might have been 10 minutes

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities
updated by @soaringeagle: 01/02/15 08:04:45PM

@brian
10 years ago
10,148 posts

soaringeagle:
but the cpu ghz speeds are being misread
says 1.6 instead of 3.8

JR gets the CPU MHz directly from /proc/cpuinfo or /usr/bin/lscpu - if JR sees it wrong then it means your OS is seeing it wrong, so you want to check that out.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
10 years ago
3,304 posts

no it happened at the update that added v 1 of the performance test
on both servers it was reported wrong
before the performance test was added it worked just fine

also according to your 'optional ) mysql tweaking guide i found on here the values u gave were for 8 not 16 gigs so i doubled them but didnt affect performance still getting around a 21 18 at the best

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
10 years ago
10,148 posts

soaringeagle:
no it happened at the update that added v 1 of the performance test
on both servers it was reported wrong
before the performance test was added it worked just fine

None of the CPU detection code has changed. Try this - log in to your server via SSH and run (as root):

lscpu

If you get no output or "command not found" run:

cat /proc/cpuinfo

and send the output to me at support [at] jamroom [dot] net.

Thanks!

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
10 years ago
3,304 posts

sent

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
10 years ago
3,304 posts

that does look odd im going to call them about that but the old server dropped from a 3.3 to 0.34 with that update the new servers suposed to be 3.8 but reports 1.6

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
10 years ago
3,304 posts

am i reading this right
are some processors diferent speeds then others

Total processors: 8

(snipped)

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
10 years ago
10,148 posts

Jamroom will use the output of lscpu BEFORE /proc/cpuinfo, and your lscpu reports 1600mhz. Most likely this is the "idle" speed of the CPU OR it could be getting throttled due to a thermal issue - either way I would contact your hosting provider and have them check.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
10 years ago
3,304 posts

there was a thermal issue with the drive..twice so they just swapped the drive
hmmm

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
10 years ago
3,304 posts

ok they are looking into the thermal issue overall but 1 last optimization question..
i hope 1 last anyways

upgrading mysql to 5.6.. says it has alot of performance improvements etc
sounds like just whats needed
but they say the only way to go back is a os reinstall if it doesnt work out

so whats your opinion is it worth upgrading
im really considering it ..but wondering if its worth the risk

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
10 years ago
10,148 posts

Use whatever the stable version is that has been created for your distribution. This means that if you are on CentOS, use whatever is installed via "yum" - i.e.

yum install mysql-server

This is likely what you are already using, and I would not change it.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
10 years ago
3,304 posts

yes in whm 5.5 is installed but has a option to upgrade to 5.6 wich says it has several performance enhancements
its tempting but ill wait and see what the thermal issues resolve

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
10 years ago
3,304 posts

any fine tuning options you can think of i'm all ears
i can get the db perfomance to toughly 18.something and overall sometimes to 405ish
how u managed 900 is beyond me except maybe if the sites more..in use the db cache is more primed

heres what i got

[mysqld]
default-storage-engine=MyISAM
innodb_file_per_table=1
max_allowed_packet=268435456
open_files_limit=65535
key_buffer              = 4096M
max_allowed_packet      = 32M
thread_stack            = 192K
thread_cache_size       = 8
table_cache             = 1536
table_definition_cache  = 1536
tmp_table_size          = 1024M
max_heap_table_size     = 1024M
thread_concurrency      = 8
sort_buffer_size        = 1M
read_buffer_size        = 1M
read_rnd_buffer_size    = 768K
query_cache_size        = 64M
max_connections         = 200
innodb_buffer_pool_size = 4G
innodb_flush_method     = O_DIRECT
wait_timeout=10

i was up to 4 am...again
going over many many mysql tweaks all of wich didnt seem to help some hurt
note
i have 8 cores
16 gigs
the tips u offered earlier seemed to be for 8 gig
ps the cores are correctly reading 3.3ghz now

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
10 years ago
10,148 posts

Do you have an SSD in your new server? If so, make sure MySQL is using the SSD drive for it's tmp files.

Remove your max_allowed_packet line - you've set it wrong. In fact you don't need some of the things you have added. It should look like this:

key_buffer              = 1024M
max_allowed_packet      = 32M
thread_stack            = 192K
thread_cache_size       = 8
table_cache             = 1536
table_definition_cache  = 1536
tmp_table_size          = 1024M
max_heap_table_size     = 1024M
thread_concurrency      = 8
sort_buffer_size        = 1M
read_buffer_size        = 1M
read_rnd_buffer_size    = 768K
query_cache_size        = 128M
max_connections         = 200
innodb_buffer_pool_size = 1024M
innodb_flush_method     = O_DIRECT

You're not going to need those other settings with Jamroom. Also - there is no need to go above a gig in your Key Buffer or InnoDB buffer size - you're just going to be wasting RAM there, which would be better used on Apache processes.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net
updated by @brian: 01/04/15 06:43:21AM

@soaringeagle
10 years ago
3,304 posts

ill check on the tmp it does have ssd
remove the default storage and all?

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
10 years ago
10,148 posts

soaringeagle:
ill check on the tmp it does have ssd
remove the default storage and all?

Yeah you don't need that.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@brian
10 years ago
10,148 posts

And just to be sure - you're restarting your MySQL server after you make any my.cnf changes?

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
10 years ago
3,304 posts

ofcourse
im starting to get to be a pro at editing and restarting ive done it often enough haha just not at getting the settings right as theres soooo much conflicting info out there

some say to give inobdb buffer as much space as possible between 25 and 50% of your ram
etc
thedy might have been setting up a server to just act as a mysql server alone though
wasnt too clear

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
10 years ago
3,304 posts

no matter the settings 387-405 seems to be the max i can squeeze out of it
with mysql best performance at about 18.85..ish
often 19.45
the colors yellow blue..is there a green that i should hope to achieve
ive seen performance test scripts but they seemed old and outdated like for mysql 2. something to 3.something
doses the test test cache vs disk reads
does having mysql running longer with a more primed cache affect the readings

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
10 years ago
10,148 posts

soaringeagle:
does having mysql running longer with a more primed cache affect the readings

Maybe a tiny bit - but caching is purposefully disabled for the test (since we want to test raw MySQL performance).

I get ~500 on my laptop (2 year old Macbook Pro) so I'm going to guess that your MySQL is NOT using your SSD for the tmp table OR the SSD in your server is a bit on the slow side.

I really don't know. I didn't setup your server, so really can't give you any more advice than I already have given you. I would just expect a server of that size to perform much better.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
10 years ago
3,304 posts

they say that mysql is set to use ssd for everything
is there a way to check that config

id be more then happy to give you admin on my site ssh and whm root access and whatever else youd need to figure this out
youve already been so responsive and helphul just dont see whats the issue the ssd or filesysten checks always arounbd 0.73-0.79ish
since baseliunes 4.0 i assume thats pretty damn good and its justthe database thats really slowing things

also
does the number of modules and skins installed affect the test at all

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
10 years ago
10,148 posts

soaringeagle:
does the number of modules and skins installed affect the test at all

No - that does not impact it at all.

You can send me your root login and Jamroom login and I will check it out tomorrow - I'm actually out of my office right now but will be back in tomorrow morning.

Thanks!

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
10 years ago
3,304 posts

ill send it to support at jr

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
10 years ago
3,304 posts

sent info but b4 u start let me know and ill stop my ftp transfers

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
10 years ago
10,148 posts

soaringeagle:
sent info but b4 u start let me know and ill stop my ftp transfers

Thanks - I won't be able to check that out until tomorrow morning - I will respond to your email when I am checking it out.

Thanks!

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
10 years ago
3,304 posts

i already knew that

just keep in mind that your idea of morning might not be the same as mine haha if i dont get to sleep till after 6 am again that is

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
10 years ago
3,304 posts

just a note
im installing jr on a news site fresh install nearly empty database and i get a significanly improved score 440 and db about 17
still not close to that bsaseline
but seems the size of the database on the other site makes a diference
and .. there are most likely a few databases no longer in use that i can empty or remoive

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
10 years ago
10,148 posts

I would recommend running the test multiple times to get a feel for the average - it's not uncommon for it to jump or fall by 10%.

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
10 years ago
3,304 posts

have been let me know when you want to try ill shut down the ftp

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@soaringeagle
10 years ago
3,304 posts

andif u want access to both sites i can do that too

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

@brian
10 years ago
10,148 posts

soaringeagle:
have been let me know when you want to try ill shut down the ftp

Sorry - have had a bunch of things to take care of here (typical Monday morning) - it will be another hour or so but I will email you back.

Thanks!

--
Brian Johnson
Founder and Lead Developer - Jamroom
https://www.jamroom.net

@soaringeagle
10 years ago
3,304 posts

well thats good cause my morning started at 1 pm
take your time wasnt trying to rush u

--
soaringeagle
head dreadhead at dreadlocks site
glider pilot student and member/volunteer coordinator with freedoms wings international soaring for people with disabilities

new "performance test" accuracy

Tags