Talk:IBM System z10

Price?

How much is it?

—67.88.123.2 (talk) 17:28, 17 October 2008 (UTC)

The baseline system is said to be one million dollars. The actual cost for the customer will vary. Will the customer buy the machine outright or lease it from IBM? Also, the particular configuration of the machine (how many main processors, how much central memory/RAM, how many of the various specialty processors, sales discounts, etc etc etc). So to be able to quote a price in the article would not be accurate for everyone. --TreyGeek (talk) 00:33, 18 October 2008 (UTC)

Hiperdispatch

Hiperdispatch a name for work similar to that already done on other operating environments as well. Is it reasonable for completeness or NPOV to mention that there is prior art in SMP systems from other vendors? Jsavit (talk) 14:39, 5 December 2008 (UTC)

My being not too familiar with the topic did what I normally do, search wikipedia: HiperDispatch. The article focuses on the IBM mainframe. So, if other vendors did it first, the hiperdispatch article is incorrect. --TreyGeek (talk) 14:45, 5 December 2008 (UTC)

I looked at the HiperDispatch article, and didn't see any claims of "done here first" (I may have missed it). Prior art I know of is Solaris Memory Placement Option helping NUMA since Solaris 9, and Solaris 10 support for CMT processors to dispatch work on warm caches also for some years. Perhaps a brief note belongs here? I agree with the comments by Henriok in the discussion page that it needs to be improved. The only thing worrisome from a factual error is the claim that WLM is involved. I don't think so - WLM is oriented to more coarse scheduling issues than whether to dispatch a TCB on one logical CPU or another to get better memory latency at the nanoseconds level. Also worth mentioning that this feature of z10 is z/OS only, and of no value for z/VM, z/Linux or other non-z/OS workloads at least for now. Jsavit (talk) 15:25, 5 December 2008 (UTC)

I have to admit, while I glanced at the HiperDispatch article I didn't really read it. I later noticed that has only one edit (the original creation by the same person with the latest edits in the z10 article). If HiperDispatch on the z10 works similarly to some other system I would agree that it may be appropriate in this article to mention that. --TreyGeek (talk) 02:22, 6 December 2008 (UTC)

What is a "book"?

The text uses the word "book" in an odd way -- I assume this is jargon for "the mainframe", but would someone be so kind as to clarify?

The text in question:

There are more physical PUs in a book than characterizable PUs.

TIA, -- Erik Anderson, 98.125.239.148 (talk) 05:55, 16 February 2009 (UTC)

This is explained in no great detail later on: "The machines are powered by one to four "books," each comprising memory cards (including multi-level cache memory) and a multi-chip module (MCM) of processing units (PUs)." -- Henriok (talk) 10:36, 16 February 2009 (UTC)

I can see how hiding the definition to book down in the middle/bottom of the article can be unclear. I've moved the definition up the start of the processors section. It also resolves an issue I noticed where we refer to PU without saying what PU stands for. Hope it is more clear now. --TreyGeek (talk) 16:31, 16 February 2009 (UTC)

Mantissa Corp, z/VOS, and x86 virtualization

I posted this on the z/VM talk page. However, it appears to be very stubby and I'm not sure how many people are watching that page. So, forgive the double posting, but depending on what is actually announced it may be relevant for a mention in this article.

I'm not sure how significant this will be.  But next week the Mantissa Corportation 
will be announcing their z/VOS product.  The announcement about their upcoming 
announcement says that with z/VOS "you can deploy and manage native x86 Windows® and 
Linux images under z/VM."[1]  It will be interesting to see what real information 
they can provide next week and integrate it into this article and possibly the other 
System z related article.  

--TreyGeek (talk) 04:06, 26 February 2009 (UTC)