Author |
Message
|
ivanachukapawn |
Posted: Thu Nov 16, 2006 5:48 am Post subject: sizing data and log (linear) file systems |
|
|
 Knight
Joined: 27 Oct 2003 Posts: 561
|
I'm having trouble locating the documentation/sizing formula for sizing the file systems for MQSeries 6.0 queue managers.
I know
Messages/Day = 1,000,000
average size = 1k
and I need to determine requirements for Log and Data file system sizes.
Thanks |
|
Back to top |
|
 |
jefflowrey |
Posted: Thu Nov 16, 2006 6:00 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002 Posts: 19981
|
Persistent messages?
Messages staying on queues for very long? _________________ I am *not* the model of the modern major general. |
|
Back to top |
|
 |
ivanachukapawn |
Posted: Thu Nov 16, 2006 6:12 am Post subject: |
|
|
 Knight
Joined: 27 Oct 2003 Posts: 561
|
Yes - persistent messages. Retention period = 3 days. I've just not been able to locate the documentation for the formula. |
|
Back to top |
|
 |
jefflowrey |
Posted: Thu Nov 16, 2006 7:15 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002 Posts: 19981
|
|
Back to top |
|
 |
mvic |
Posted: Thu Nov 16, 2006 10:26 am Post subject: |
|
|
 Jedi
Joined: 09 Mar 2004 Posts: 2080
|
ivanachukapawn wrote: |
Yes - persistent messages. Retention period = 3 days. I've just not been able to locate the documentation for the formula. |
What type of system and OS ?
To (mis-)quote past comments from Jeff, memory and disk space are cheap compared to the downtime you experience if you run short of either of these.
So my very basic advice is to specify a system with a LOT of headroom, both in memory and disk, to protect your investment into the future. |
|
Back to top |
|
 |
ivanachukapawn |
Posted: Thu Nov 16, 2006 10:54 am Post subject: |
|
|
 Knight
Joined: 27 Oct 2003 Posts: 561
|
MVIC,
I agree with you completely. I intend to have plenty of headroom. I will be building highly-available (using both HACMP and VCS) MQ infrastructure in multiple environments, Solaris, AIX, and possibly Z/OS.
Linear Logging will be involved.
I was hoping to find a forumula which would provide a network file system size recommendation for both data and log file systems, given the following:
M = # Messages a day.
S = Average size of message.
R = # retention days
P = # of primary
X = # of secondary
Z = log page size
ignoring # of Primary and Secondary and log page size and adjusting via a multiplier of 5, I'm thinking about:
(M * R) * S * 5
I'd like to get a reasonably accurate estimate by using a formula and then I'm perfectly willing to multiply it by a factor of 4 (or more) to get the headroom you recommended. |
|
Back to top |
|
 |
mvic |
Posted: Thu Nov 16, 2006 12:18 pm Post subject: |
|
|
 Jedi
Joined: 09 Mar 2004 Posts: 2080
|
ivanachukapawn wrote: |
I'd like to get a reasonably accurate estimate by using a formula and then I'm perfectly willing to multiply it by a factor of 4 (or more) to get the headroom you recommended. |
This is an exercise only you and your business can really decide on. Your formula seems OK to me, but bear in mind any future expansion of your system that might be required by the business - will they want this queue manager to handle more messages, bigger messages, longer retention period, many queues used for similar purposes in parallel...
Another thing to think about: will you use linear or circular logging... if linear, how will you archive your inactive log files. |
|
Back to top |
|
 |
ivanachukapawn |
Posted: Thu Nov 16, 2006 12:37 pm Post subject: |
|
|
 Knight
Joined: 27 Oct 2003 Posts: 561
|
Thanks. I understand that requirements will change and the network file systems need to be sized to deal with that eventuality.
We're dealing with enterprise applications that will be use this infrastructure and consequently, we'll be using persistence and linear logging.
we'll have linear log maintenance scripts run every hour 24/7 which will discover which old linear log files can be safely deleted and delete them. |
|
Back to top |
|
 |
bruce2359 |
Posted: Thu Nov 16, 2006 1:07 pm Post subject: |
|
|
Guest
|
Is 5 in your equation a fudge-factor? If all the messages are 1k, the average message size will also be 1k. If the size varies, I'd use the maximum message length instead of average.
Data and log images include the Message Descriptor; log images include the log header (refer to the doc url obove).
Think about how long it would take (worst case scenario) if 3 million messages are lost because of a media (disk) failure. When MQ restarts, it will demand tha all logs since the last good backup of the data be mounted so it can perform redo-undo processing to bring the data (messages) back to a consistent state. Size isn't everything. |
|
Back to top |
|
 |
|