|
View previous topic :: View next topic
|
| Author |
Message |
DJN
New User
Joined: 04 Jun 2009
Posts: 6
Location: INDIA
|
|
|
|
Can anyone explain how we pay IBM ? Is it according to CPU time only. or Elapse time also contributes to it.
Because if I try to reduce a total Elapsed time of job will it reduce cost paid to IBM? |
|
| Back to top |
|
 |
dick scherrer
Moderator Emeritus
Joined: 23 Nov 2006
Posts: 19244
Location: Inside the Matrix
|
|
|
|
Hello,
Someone in your organization (Purchasing?) has a contract that specifies what resource usages are charged.
The conditions are completely specific to your contract.
Reducing elapsed time will most likely require reducing i/o or cpu use (either inside the problem process or in database processing) - both of which could reduce the cost of the process. |
|
| Back to top |
|
|
enrico-sorichetti
Superior Member
Joined: 14 Mar 2007
Posts: 10873
Location: italy
|
|
| Back to top |
|
|
DJN
New User
Joined: 04 Jun 2009
Posts: 6
Location: INDIA
|
|
|
|
| dick scherrer wrote: |
Hello,
Reducing elapsed time will most likely require reducing i/o or cpu use (either inside the problem process or in database processing) - both of which could reduce the cost of the process. |
I know we calculate the cost saving by Cpu time reduction. Its very directly proportional. How well it for I/o wait time. Do I pay for waiting for a dataset ? |
|
| Back to top |
|
|
DJN
New User
Joined: 04 Jun 2009
Posts: 6
Location: INDIA
|
|
|
|
| enrico-sorichetti wrote: |
for coffee I usually get charged by the time it takes me to drink it,
longer time ==> more satisfaction ==> higher price 
rough approach.. elapsed = cpu + wait
why would You want your organization be charged for the wait time
( job waiting for a tape mount and the tape operator having a coffee break) |
Thats funny. I will then perfer to have a quick sip
I want to know if i reduce a IO(wait time not the CPU time) operation time for a Batch job will turned to some available dollar saving. |
|
| Back to top |
|
|
Robert Sample
Global Moderator
Joined: 06 Jun 2008
Posts: 8696
Location: Dubuque, Iowa, USA
|
|
|
|
How you pay IBM is entirely a site-specific function, and depends upon many, many factors.
Reducing I/O wait time would have absolutely no impact upon your elapsed time if your job is CPU-bound; if you don't know whether or not the job is I/O-bound or CPU-bound then you cannot predict the impact of reducing I/O wait time.
Furthermore, accounting systems typically recover the cost of the machine while it is doing work -- not the elapsed time, but the actual CPU usage (and possibly including I/O, etc). This is because elapsed time has only the most tenuous relationship to CPU time -- elapsed time depends more on the number of processors being used, the number of address spaces running, the workload for the various address spaces, channel contention, device contention, presence of zIIP and zAAP processors with work that can be done on them, etc, etc, etc.
My imagination is limited -- I am not sure I can think of a worse way to tune a machine (and in wanting to reduce costs you're talking about doing tuning) than by measuring elapsed time. |
|
| Back to top |
|
|
Itanium
Active User
Joined: 22 Jan 2006
Posts: 114
Location: India
|
|
|
|
remebering an instance where one of our job was in execution queue for a long time (nearly 1.5 hrs) but the total CPU time taken was 2 - 3 mins. This resulted in that job holding an initiator for nearly 1.5 hrs.
That job was actually updating an IAM file, we tried to add BLSR to the file which reduced the execution time of the job to less than five mins, but there was no huge CPU time saved.
Still the initiator was freed early which helped in other jobs to use that and execute.
There are certainly gains even if you try to reduce the elapsed time. |
|
| Back to top |
|
|
Anuj Dhawan
Superior Member
Joined: 22 Apr 2006
Posts: 6250
Location: Mumbai, India
|
|
| Back to top |
|
|
|
|
