MQ

Hi Yvonne,

I retired from IBM several years ago so I'm almost impartial.

I remember one customer saying to me "I want my applications team to write applications and not write infrastructure. Writing infrastructure is hard. I want IBM to do the hard stuff.  I want more bang for my buck"

Another quote is evolve or die

The short answer to your question is customer demand.

For example a common scenario was a customer as 3 linux servers each running with a queue manager.  Over the weekend, server 1 was shutdown for maintenance.
On Monday all the clients started work, and connected to qm2 and qm3.  At 11am server1 started up.

By lunch time, server 2, and server 3 were overloaded, and server 1 was doing very little work
If a client was to disconnect and reconnect, it may connect to qm1, but by default it stayed connected to same queue manager for the day.  .  As the clients disconnect overnight.  On Tuesday the work is spread across all 3 queue managers.

10 years ago you could solve this by having the clients disconnect after ever 30-40 minutes and reconnect.  This way as servers come online, they will get used.

Many customers asked if IBM could fix this.   They also did not want to change their applications ( for example change them, recompile them, test them, and put the changes into production).  In some cases the customers do not have the source of their programs!   This is known as Uniform clustering.

Another example is on z/OS.  Customers do 1,000 transactions a second through an  MQ application.   If there is a problem they would like to see the accounting data to see what is going on.  If the granularity of the data is 30 minutes you could get "average get time is 1000 microsecond" If you had a few mqgets taking 1 second.  The average (over the half hour) is 1001 microseconds which within the noise.  If you can get the accounting data every second anomalies will show up.

20 years ago people programmed in COBOL and C.   Then Java, now it is Go, Python, JSON, and "services" over HTTPS.
The current trend is containerisation...
TCPIP is not fast enough... please can MQ exploit the latest technology.

and so on... 

If you think what the IT world will look like in 5 years...  think what changes products (like MQ) will have to make, so they are ready when the future arrives.  ( how about the impact of quantum computing, and TLS/SSL...)

Colin

------------------------------
Colin Paice
------------------------------