From: Mark D
I can't imagine how IBM I pricing/performance could beat a
blade center or especially a rack of say 10 1u servers. 

When you speak in terms of 10 1u servers, that's something of an unfamiliar notion in the IBM i world. I suppose you're talking about servers with at least 4-cores per server? 40-cores total? That's a lot of power! Why would you need that many servers?

The majority of IBM i shops run single servers with 4 or fewer cores, and something like 95% run single servers with 8 or fewer cores. IBM i makes it possible to integrate and manage otherwise disparate workloads so that you don't need that many servers and that many cores. Granted there are some companies running IBM i on 256-core servers. But those are HUGE organizations.

If you run native workloads (stuff compiled in ILE languages), and native interfaces (database, system APIs, data queues, data areas, output queues, integrated file system, etc.) you get the added bonus of your applications utilizing fewer CPU cycles per work accomplished. You don't need much hardware, relatively speaking.

How do IBM i customers do more with fewer cores, less total memory, and less storage? Say you have an environment consisting of IBM i native, PHP, Java, PASE, and DB workloads? IBM i is able to manage each environment safely, securely, reliably, efficiently, with fewer bottlenecks, using a choice of green-screen or "navigator" interfaces for administrative purposes, and various other user interfaces for applications, without needing to distribute them across server farms. IBM i is a better operating environment for managing otherwise disparate workloads.

Under most real-world workloads, IBM i offers lower total cost of ownership. Why distribute workloads across server farms?


Return to Archive home page | Return to MIDRANGE.COM home page