Reader Feedback On The Power Systems-IBM i Road Ahead
June 27, 2011 Timothy Prickett Morgan
I have always liked your articles.
For large shops doing MSHA using replication, we would love partition mobility, but the high volume of I/O our internal storage performance gives us seems to also preclude any chance at the mobility we need. Note: IBM SANs could not provide the performance required when we tried the shift way back in 2003 and to this date nobody wants to risk SAN performance–even with EMC technologies.
If I could have SAN replication functions (and clustering functions) with internal storage (and its great performance) without having to implement iASPs, then we would be doing site to site mobility now (in fact, we would be doing it yesterday).
But since iASPs don’t cut it for us and we need the best performance of internal storage we are not going to SANs any time soon (of course). Plus, the way IBM has its cluster stuff implemented–dependent on iASPs–we can’t do clusters or mobility.
IBM needs to make our internal SANs (that’s what they have been since the System/38 days) functional enough so we can take advantage of mobility with or without our data replication methods.
I wish IBM would walk away with that message and not the old “hey, we aren’t going to SAN any time soon” part of the message. But they seem far too invested in iASP and external storage to consider what the very small and the very large shops really need to implement mobility today.
Thanks for listening,
While I am not surprised that the performance of SAN arrays in conjunction with OS/400 was bad eight years ago, I wonder if the situation has not improved somewhat. All of the components–disk drives, Fibre Channel interconnect, disk controllers, cache memory, and so on–are bigger, stronger, and faster, even if you need to link through the Virtual I/O Server (VIOS) to attach many arrays to the IBM i operating system. It seems to me that it really doesn’t matter much at this point, however, and that you might be a perfect candidate for the addition of flash-based solid state disks to your IBM i boxes. That all depends on your budget and how easily and transparent hot data can be moved from disk to flash and back again. All server makers make this sound easy, but I just can’t quite believe it.
If you need both high I/O and high capacity in your storage arrays, you are kind of in a bind.
It is exceedingly interesting to me that the needs of the low-end customer–who doesn’t want SANs or iASPs–and the high-end customers–who needs internal arrays for performance reasons–are aligned in a direction that IBM is not going. I can’t remember the last time that happened. Maybe when System/38 and System/36 customers both wanted cheaper and more powerful iron, resulting in the AS/400.
We are in a bind of sorts in that we have a huge investment in internal with SSDs and SATA drives.
We are also in a bind because of the climate after the iASP and external issues were so devastating that people still cringe at the thought.
The last thing is that IBM hardware techs in Rochester know our I/O signature very well and even the current products don’t quite make it for use plus the cost of replacing our current internal investment would never be approved.
Thanks again for listening and perhaps somebody you know will hear this message. 🙂