Operating System - OpenVMS
1827870 Members
1156 Online
109969 Solutions
New Discussion

Summary output from MONITOR

 
SOLVED
Go to solution
Chaim Budnick
Regular Advisor

Summary output from MONITOR

I would appreciate it if someone could please review the mon IO and mon disk summaries and comment:

MONITOR IO:

I/O SYSTEM STATISTICS
on node ALFTRD From: 13-OCT-2004 15:03:32
SUMMARY To: 13-OCT-2004 15:04:44

CUR AVE MIN MAX

Direct I/O Rate 327.66 284.01 132.66 442.33
Buffered I/O Rate 327.00 244.97 162.66 432.33
Mailbox Write Rate 3.33 2.55 0.33 7.66
Split Transfer Rate 0.00 0.00 0.00 0.00
Log Name Translation Rate 235.00 275.28 35.66 797.33
File Open Rate 14.00 10.47 2.33 32.00

Page Fault Rate 325.33 321.71 26.66 1169.00
Page Read Rate 72.33 78.01 1.00 293.66
Page Read I/O Rate 42.00 44.77 1.00 167.66
Page Write Rate 0.00 0.00 0.00 0.00
Page Write I/O Rate 0.00 0.00 0.00 0.00
Inswap Rate 0.00 0.00 0.00 0.00
Free List Size 64294.00 63792.95 63018.00 64294.00
Modified List Size 4362.00 4385.33 4362.00 4393.00

MONITOR DISK:

OpenVMS Monitor Utility
DISK I/O STATISTICS
on node ALFTRD From: 13-OCT-2004 15:02:01
SUMMARY To: 13-OCT-2004 15:02:46

I/O Operation Rate CUR AVE MIN MAX

ALFTRD$DRA0: ALPHASYS 11.33 26.46 9.33 49.33
ALFTRD$DRA1: USERS1 51.00 22.92 6.66 51.00
ALFTRD$DRA2: USERS2 67.33 90.17 58.33 123.66
ALFTRD$DRA3: USERS3 89.66 43.59 8.00 127.66
ALFTRD$DRA4: USERS4 21.33 16.55 5.66 44.33
ALFTRD$DRA5: USERS5 0.00 0.00 0.00 0.00
ALFTRD$DKC0: DAILY1 0.00 0.00 0.00 0.00
ALFTRD$DKC100: DAILY2 0.00 0.00 0.00 0.00
ALFTRD$DKC200: WEEKLY1 0.00 0.00 0.00 0.00
ALFTRD$DKC300: WEEKLY2 0.00 0.00 0.00 0.00





7 REPLIES 7
Willem Grooters
Honored Contributor
Solution

Re: Summary output from MONITOR

I _guess_ you have some performance problem, but no-one can tell more if we don't know what's going on on your system, and what type of system you have.
So we need at least VMS version, and a dump of what's on (SHOW SYSTEM will probably do for a first impression)
At first glance on these figures, I would say you have a number of problems in this particular timeframe:
* Disk IO, particular to disk USERS3 and, in a lesser extent, disk USERS2
* High pagefaultrate (wll, what's 'high')
* High log name translation rate.

Do you have a program on your system that creates processes at high speed? There has been a discussion on that in this forum before on that. That could explain the high pagefault rate.

Do you have a program running that fills an array 'the wrong way'? I've seen very high pagefault rates due to this issue.

But as stated, there's more info required.

Willem
Willem Grooters
OpenVMS Developer & System Manager
Ian Miller.
Honored Contributor

Re: Summary output from MONITOR

A high I/O queue size is a better indication that disks are overloaded.

However as Willem said, what is the percived performance problem and what are the normal figures for your system.
____________________
Purely Personal Opinion
Chaim Budnick
Regular Advisor

Re: Summary output from MONITOR

The OpenVMS version is V7.3-1

The system is primarily (almost totally) used for a DSM application.

Chaim
Hein van den Heuvel
Honored Contributor

Re: Summary output from MONITOR


Doesn't look too bad for a minute snapshot.
If those AVG IO rates (DRA2, and to a less extent DR3 :-) keep up over a longer time (30 minute main production window), then you may want to try to tune. Can you give more buffers to DSM? There appears to be memory enough.

The worst coutner I see is the page read IO.
Image activations? Install (more) images shared?

Also, those DRA devices, are they simple disks or stripes? 100 IO/sec is significant for a singkle disk, but a walk in the park for 4 or 5 disks striped/raided.

What was the CPU usages at that time (MONI MODE)?


Also... what made you look? If it was a suggeastion of a performance problem, then can you try to articulate that better? Is there is response time / batch job time you cna use as a reference for before/after tuning?

Cheers,
Hein.

Antoniov.
Honored Contributor

Re: Summary output from MONITOR

As Willem pointed it seems your machine makes heavy I/O.
Direct I/O and Buffered I/O are high values;
it seems disk USER2 and USER3 make too high I/O.
Your application doesn't use mailbox.
Perhaps you have many process starting because you have Page Fault Rate, Page Read Rate and File Open Rate very high.

Antonio Vigliotti

Antonio Maria Vigliotti
Lawrence Czlapinski
Trusted Contributor

Re: Summary output from MONITOR

Chaim, determine whether the performance complaint is valid. What is the performance complaint? Is it specific to certain processes or systematic?
1. Measure the before change performance and after change performance.
2. Can you post a MON PAGE? This will give a better indication of what types of soft faults you're getting.
3. Since you have available memory, I would recommend running WORKSET.COM (attached)
Lawrence Czlapinski
Trusted Contributor

Re: Summary output from MONITOR

Chaim,
1. If you have applications that are used by multiple users, make sure they are installed (INSTALL)/HEADER/SHARE and consider use of /RESIDENT.
2. Since your hard faults are all page fault read i/os and you have available memory, you might consider raising WSDEF and WSQUO for tasks that have working set total pages greater than WSDEF and WSQUO. By getting more pages quickly, the processes page fault less. You might also look at the percentage of pages which are global for the processes with performance problems. My Oracle user global pages are over 60% of their total pages. My Natural user global pages are typically 55-60% of their total pages. With a lot of users, there is a good chance that the global pages they need are already in memory.
Lawrence