> It is said that running the same DB2 batch job on an z/OS consumes more > CPU, or in other words that z/VM if far more economical in use of CPU > resources. Is that a fact and if so, can anyone give me a good indication > about the factor z/OS is more expensive...
It is definitely true that z/VM and Linux guests (DB/2 for z/VM is a orphan product; all the real comparisons are to DB/2 on Linux as a z/VM guest) is more economical in its use of CPU, but that's not the whole picture for DB/2. DB/2 on z/OS certainly consumes more expensive CPU resources (in that it requires standard engines rather than the vastly cheaper IFLs running z/VM and Linux guests). But, be careful comparing the two: DB/2 on z/OS has access to a lot of system enhancements that are not available to DB/2 on any other platform (access to zIIP processors, subsystem utilization enhancements, etc). DB/2 on VM or Linux can't use those things. In my head, it's more a comparison of how much bang you can get for the same buck. You will get substantially less total processing capacity for the same amount of money using DB/2 on z/OS (you need to buy standard engines for z/OS, and you may take a hit on other z/OS software costs as that will change the model number of your machine). You also trade some functionality limitations in UDB and the necessity of using network access from batch jobs to a remote database for the "everything in one place" configuration of DB/2 on z/OS. If you have the right levels of z/OS and DB/2 (ie VERY modern), the zIIP levels that advantage a bit. It's not just a financial or performance (x is nn% faster than y) issue. You can build the case that for X amount of money, you get more total capacity in that X euros buys you 3 IFLs vs 1 standard engine or 3 zIIP engines (if you have sufficient DB/2 workload to merit more than one). If you are a small shop with a sub-uni processor configuration (rare for z/OS), then the fact that an IFL or zIIP are the same price and both run at full rated speed for that engine type may level the playing field a bit. The fact that zIIPs only benefit DB/2 workload makes them less attractive to me; I'd really rather have an IFL which I can use to possibly deliver other interesting services rather than just speed up DB/2.
