| Author |
Message
|
| dotaneli |
 Posted: Wed Aug 13, 2008 3:48 am Post subject: MQ 6 Linux vs Windows Benchmarking |
 |
|
Voyager
Joined: 19 Oct 2005
Posts: 99
Location: Israel
|
Hello everyone.
Where can i find information comparing MQ 6 benchmarks of Win vs. Linux?
Plus - where can i find benchmarks that help me consider if moving to 64-bit MQ will improve my performance?
thank you.
Eli. |
|
| Back to top |
|
 |
| exerk |
| Posted: Wed Aug 13, 2008 3:55 am Post subject: |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
The SupportPacs link at the top of the page will take you to where V6 benchmarks on various platforms are published, however I don't think there are any specific platform v platform benchmarks; I think it would be relative anyway unless you had two identical hardware setups to test against.
_________________
It's puzzling, I don't think I've ever seen anything quite like this before...and it's hard to soar like an eagle when you're surrounded by turkeys. |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Wed Aug 13, 2008 4:49 am Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
I remember reading somewhere in IBM land that on identical hardware, Linux will be slightly faster for MQ than Windows. I think it was the Linux Performance Support Pack actually.
Linux x86 and Windows are one of the few O/Ses (the only?) that can run on the exact same type of metal box, making the comparison fair.
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| exerk |
| Posted: Wed Aug 13, 2008 4:59 am Post subject: |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
| PeterPotkay wrote: |
| I remember reading somewhere in IBM land that on identical hardware, Linux will be slightly faster for MQ than Windows... |
Not surprising...thank you for that snippet of info Peter.
_________________
It's puzzling, I don't think I've ever seen anything quite like this before...and it's hard to soar like an eagle when you're surrounded by turkeys. |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Wed Aug 13, 2008 6:50 am Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9472
Location: US: west coast, almost. Otherwise, enroute.
|
Slightly faster should not be the business case for your platform choice. (I suspect you already know this.)
If you are at or near the edge of capacity, you will have only slightly delayed the inevitable.
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Wed Aug 13, 2008 6:56 am Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
Agreed. All things being equal, slightly faster can be the decision maker. However, different O/Ses immediatly bring into play so many variables, many specific to each shop for each O/S that things are FAR from equal in this comparison.
But, if you are trying to make a case for Linux, it helps to point to the doc and say "See, it will be faster!" 
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| dotaneli |
| Posted: Thu Aug 14, 2008 12:13 am Post subject: |
|
|
Voyager
Joined: 19 Oct 2005
Posts: 99
Location: Israel
|
Hey,
thank you all for your replies.
However - the dilema is over, Windows is the winner since the out client already has knowledge using mq and windows.
Now what bugs me, is how can i advise my client about new machine's hardware?
He wants to know how many CPU's he needs.
The client is looking at peaks of 60K messages pre second, average message size is 2K.
I can, ofcourse, calculate the needed disk space (each queue's max depth + logs etc...)
But what i really need to know - how many cpu's and how much RAM? let's assume that only windows and MQ runs on that machine.
P.S. - some connect to MQ via servers and some via client connections...
thanks. |
|
| Back to top |
|
|
| David.Partridge |
| Posted: Thu Aug 14, 2008 12:21 am Post subject: |
|
|
Master
Joined: 28 Jun 2001
Posts: 249
|
|
| Back to top |
|
|
| dotaneli |
| Posted: Thu Aug 14, 2008 12:42 am Post subject: |
|
|
Voyager
Joined: 19 Oct 2005
Posts: 99
Location: Israel
|
10x...
i already saw that.
However - that shows benchmarking, not a formula for calculationg the neede CPU's and RAM.
And - a correction. 60K messages per minute
Anyone? Formula to calculate? assume messages are all persistent.. |
|
| Back to top |
|
|
| zpat |
| Posted: Thu Aug 14, 2008 2:46 am Post subject: |
|
|
Jedi Council
Joined: 19 May 2001
Posts: 5866
Location: UK
|
Double post.
Persistent throughput is limited by DISK performance usually.
Use fast, cached disks, separate for logs and queues, on dedicated disk hardware with the RAID optimised for writes. |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Thu Aug 14, 2008 6:08 am Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9472
Location: US: west coast, almost. Otherwise, enroute.
|
The number and size of messages per second is only a part of the equation. Concurrency of workload is also part of this. What is the service level agreement for a single transaction - assuming you are using the usual request-reply application architecture.
In the request-reply design the requesting program MQPUTs a message to request queue, the replying MQGETs request message, does whatever back-end work (database update, for example) and MQPUTs reply message, and the requesting program MQGETs reply message, commits the changes (or backs them out), and communicates the results to the user.
A performance monitoring tool can capture clock-time and cpu required between the start and and end of the transaction. Calculate how many transactions a single processor can process in the time allotted by the SLA. If more tranactions must be processed concurrently, add enough processors to accomodate the backlog, PLUS some additional capacity for peaks.
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| Back to top |
|
|
|
|
