There are a lot of factors in TCO evaluations, and initial investment and service costs are only a part of it. Having gone through an evaluation of a Z900 as a replacement for multiple smaller boxes, it turns out that the amount of space saved in the machine room and the associated reduction in A/C, electricity, and management staff can justify the cost. The MTBF of 35 years is rather cool too.
As an example, a 32-processor system can potentially replace 100 ~ 200 smaller systems depending on the load characteristics.
The beauty of a mainframe is that there are
fail-over dies on the ceramic base in case of a failure. The hardware handles the switch from a damaged
section to a live standby system without having
to go to the OS level. Also, resource management
can be finely tuned - switching memory or CPU resources from one VM to another is relatively easy and is all done in software. No opening
of boxes and pulling out/putting in memory is
required.
A mainframe can be a good fit for companies which have large numbers of
servers and are looking to consolidate, and at the same time as increasing reliability.
There are a lot of factors in TCO evaluations, and initial investment and service costs are only a part of it. Having gone through an evaluation of a Z900 as a replacement for multiple smaller boxes, it turns out that the amount of space saved in the machine room and the associated reduction in A/C, electricity, and management staff can justify the cost. The MTBF of 35 years is rather cool too.
As an example, a 32-processor system can potentially replace 100 ~ 200 smaller systems depending on the load characteristics.
The beauty of a mainframe is that there are fail-over dies on the ceramic base in case of a failure. The hardware handles the switch from a damaged section to a live standby system without having to go to the OS level. Also, resource management can be finely tuned - switching memory or CPU resources from one VM to another is relatively easy and is all done in software. No opening of boxes and pulling out/putting in memory is required.
A mainframe can be a good fit for companies which have large numbers of servers and are looking to consolidate, and at the same time as increasing reliability.