Hello Brian, Some questions:
- Did your research took into account the common storage overhead? other synchronization overhead? - I know that CPU speed of the 50 Mips proc. is equal to the 100 Mips proc. The difference is in total capacity. But its like the supermarket. You can have 2 cashiers, but if one can do the same work in a time frame that the 2 others do (No doubt that the single CPU can do twice the work in a heavy duty system). My point is that looking at a specific TRX, it may take less, but the total machine performance will be same or better (no overhead, see 1 above). The average throughput is equal or better. - You didn't take into a considering a demanding application that needs more then what a single CPU can give. For example, if you have a 2 way machine, each of 50 Mips, and you application needs 51 Mips, You will pain more then in a system with a one way machine. As you can't by a machine in a single mips step, the total capacity driven by a 2 way machine selection is much higher and the (software) costs are much, much higher. - looking at your CICS example, hoping your users don't just hit "enter", CPU is the least factor in a user - machine environment looking at the elapse time. IO, interval-control wait, record level sharing, So much reasons to wait - so many opportunities to free the CPU to others... - MVS handles TCBs (and SRBs) not big bulks line "CICS" "TCPIP" the TCB and old/new PSW swap can occur in the same or other address space, depending on priority. If you set the serving address spaces higher then CICS, you will not need to wait. - Another interesting example is the engine efficiancy. Do you use bus (51 passangers) or your own car? Corrent me if I am wrong. Wish to hear more about this issue. Thanks to those who already responded, Itschak Some questions: - Did your research took into account the common storage overhead? other synchronization overhead? - I know that CPU speed of the 50 Mips proc. is equal to the 100 Mips proc. The difference is in total capacity. But its like the supermarket. You can have 2 cashiers, but if one can do the same work in a time frame that the 2 others do (No doubt that the single CPU can do twice the work in a heavy duty system). My point is that looking at a specific TRX, it may take less, but the total machine performance will be same or better (no overhead, see 1 above). The average throughput is equal or better. - You didn't take into a considering a demanding application that needs more then what a single CPU can give. For example, if you have a 2 way machine, each of 50 Mips, and you application needs 51 Mips, You will pain more then in a system with a one way machine. As you can't by a machine in a single mips step, the total capacity driven by a 2 way machine selection is much higher and the (software) costs are much, much higher. - looking at your CICS example, hoping your users don't just hit "enter", CPU is the least factor in a user - machine environment looking at the elapse time. IO, interval-control wait, record level sharing, So much reasons to wait - so many opportunities to free the CPU to others... - MVS handles TCBs (and SRBs) not big bulks line "CICS" "TCPIP" the TCB and old/new PSW swap can occur in the same or other address space, depending on priority. If you set the serving address spaces higher then CICS, you will not need to wait. - Another interesting example is the engine efficiancy. Do you use bus (51 passangers) or your own car? Corrent me if I am wrong. Wish to hear more about this issue. Thanks to those who already responded, Itschak ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
