| Author |
Message
|
| Gideon |
 Posted: Fri Apr 20, 2012 10:42 am Post subject: Qmgrs and RAID |
 |
|
Chevalier
Joined: 18 Aug 2009
Posts: 403
|
On a slow system to increase throughput of persistent messages, we want to put the qmgr on a RAID
Which RAID level should we use, any issues we should consider ? |
|
| Back to top |
|
 |
| Vitor |
| Posted: Fri Apr 20, 2012 10:47 am Post subject: Re: Qmgrs and RAID |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| Gideon wrote: |
| Which RAID level should we use, any issues we should consider ? |
The issue that the increased I/O associated with a RAID device is not going to help performance? Unless the current I/O is ridiculous slow / bottlenecked? 
How have you identified that that the poor performance is due to I/O, and what's the reasoning that RAID will help?
Apart from that, it's just a disk. You mount it & the queue manager uses it. RAID level to me indicates the level of resilience you want rather than performance.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| Gideon |
| Posted: Fri Apr 20, 2012 11:03 am Post subject: |
|
|
Chevalier
Joined: 18 Aug 2009
Posts: 403
|
| Wouldn't moving from a standard SAS drive to a RAID device set at RAID 0 with 8 drives increase the performance ? |
|
| Back to top |
|
|
| mvic |
| Posted: Fri Apr 20, 2012 11:06 am Post subject: Re: Qmgrs and RAID |
|
|
Jedi
Joined: 09 Mar 2004
Posts: 2080
|
| Gideon wrote: |
| On a slow system to increase throughput of persistent messages, we want to ... |
First read the MQ Performance Report for your OS. There's no need to read and understand all the graphs at first, but do read all the bits that describe how to get the best performance out of the kit you already have.
If you want more I/O performance, call your vendor's sales line, and get them to explain which of their products perform best.
I have no real idea what RAID buys you, in terms of performance. Again, get some performance information from your vendor's sales team. |
|
| Back to top |
|
|
| Vitor |
| Posted: Fri Apr 20, 2012 11:13 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| Gideon wrote: |
| Wouldn't moving from a standard SAS drive to a RAID device set at RAID 0 with 8 drives increase the performance ? |
I don't think you'll see a significant increase. I could be wrong; I'd imagine that if the RAID controller is doing all sorts of smart buffering & predictive reads it might but, as my worthy associate correctly points out, device performance questions should be directed to the vendor.
Taking into account I've never met a hardware vendor who sold a poorly performing product. According to the sales staff. Make sure any metrics they give you are consistent with the use case you're seeing for the queue manager's I/O.
I repeat my question regarding how you've demonstrated raw I/O is the issue & (for instance) eliminated poor application use of the queue manager.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| Gideon |
| Posted: Fri Apr 20, 2012 11:23 am Post subject: |
|
|
Chevalier
Joined: 18 Aug 2009
Posts: 403
|
| Could you please give me an example of an applications poor use of the product and how I could determine if that is the case |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Fri Apr 20, 2012 11:30 am Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9469
Location: US: west coast, almost. Otherwise, enroute.
|
| Gideon wrote: |
| Could you please give me an example of an applications poor use of the product and how I could determine if that is the case |
Could you please give me an example of why you believe that slow performance is due to disk latency OR lack of RAID? How did you measure performance? Has it always been slow? If it was faster before, what changed?
_________________
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 |
|
|
| Vitor |
| Posted: Fri Apr 20, 2012 11:39 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| Gideon wrote: |
| Could you please give me an example of an applications poor use of the product and how I could determine if that is the case |
Well:
- poor use of persistance
- poor use of UoW
spring to mind, assuming you're right and it's a pure I/O issue (which you've not yet offered any evidence to). From what you've said it's equally likely that the application is thrashing around internally & wasting CPU cycles, or that it's putting message to a large number of intermediate queues due to poor design.
Likewise if the queue manager is poorly configured this will hinder performance. Are the logs & the queue files on different physical devices and those devices on dedicated controlers? If you have high I/O being bottlenecked through a single controller that's not doing you any favours either. Are the logs single or triple written?
The list of tuning items goes on and on. You've already been referred to the performance report for your OS. And asked repeatedly how you've arrived at I/O as the problem. I'm not saying it isn't but it would be a shame to go through all this if the root cause is an app that's accidentally single threading.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| mvic |
| Posted: Fri Apr 20, 2012 12:06 pm Post subject: |
|
|
Jedi
Joined: 09 Mar 2004
Posts: 2080
|
|
| Back to top |
|
|
|
|
