Ted,
Thanks for your reply. My comments below.
> I don't understand the uniqueness argument!
[Ron Hawkins] While they may or may not be important to you, let me offer
some examples:
HDS - AT-TIME split for Shadowimage
IBM - SSD drives recognized by DFSMSdfp
EMC - Infomover (FTP pipe)
> Our MIM control dataset, and later GRS, was/is in the CF.
>
> We gave up on disk years ago.
[Ron Hawkins] Perhaps my statement "the throughput of the Sysplexes that
used MIM" was a little ambiguous. When multiple SYSPLEXES share datasets
using MIM you cannot use a CF. That is the case for this customer.
> If the dataset in the CF, what does this feature mean?
[Ron Hawkins] Do you still beat your wife? :-) Well, it means nothing for
MIM in the CF, but the same site is also in the process of loading some
medium size datasets (2-8GB) into Cache Residency Manager to improve the
performance of 200-300 jobs that read these datasets sequentially at the
same time. This feature is all about reducing response time and improving
throughput.
> I haven't had the privilidge of working with SSD since the old STK 4301's.
[Ron Hawkins] You may not be aware that SSD is the generic term for
FlashDrives. I agree with you that SSD used to be memory on the end of a
channel, which is exactly what the MIM example I discussed earlier is doing.
I suggest you read
http://www.zjournal.com/index.cfm?section=article&aid=1226.
> If you say so.
[Ron Hawkins] You place so much emphasis on response time that I thought
back-end performance benchmarks would be a critical part of making sure "all
things are equal." Don't trust me: I'm a vendor. Look at your own line by
line comparison of Back-End performance.
> The feature is not as important as response.
[Ron Hawkins] My experience is that PAV has a direct impact on response
time. It is service time that it does not affect. Do you have a different
experience?
> The important point is response.
> NOT specific details.
[Ron Hawkins] Some sites find the number of FlashCopies sessions supported
important, and some do not. So for your site 1 second to copy 15 datasets
with zero response time counts the same as five minutes. That's OK, but it
is an exceptional tolerance for response and throughput. I use FlashCopy to
create my PAIO test datasets with 16 targets from one source. It's an 80%
reduction in IO, a huge saving in MSU (formats), and saves me hours of time.
>
> Since I never discussed it, I can't answer a 'have you stopped beating
your
> wife' question.
[Ron Hawkins] You said:
* "the big three vendors products and have found no major
technological differences"
* "There is NO one feature, other than cost, that will make
me pick one over the other"
* "All that matters is space, performance and such features
as..."
If you take some time to understand it, Mainframe virtualization speaks to
all three of your points: (a) It is a major technological difference; (b) It
can significantly affect cost; and (c) it provides space, performance, and
the features you mentioned. It was not meant to be a loaded question so let
me rephrase it. Do you think that all Mainframe Storage offer the same
features, functions, speeds and feeds, such as Mainframe Virtualization?
> What I said was, with all features equal, price is the driving factor.
[Ron Hawkins] I cannot find that statement, or anything similar in your
email. In fact I believe you make points that counter that argument, viz:
* "one is as good as another, as long as you are buying current
technology"
* "There is NO one feature, other than cost, that will make me
pick one over the other."
> If one vendor doesn't have the feature I need, that one is not in the mix.
>
> What I never said, was that exluding required features was on the table.
[Ron Hawkins] I think that the two quotes above mean that required features
were never considered ahead of cost and relative "newness of the platform."
> If I need the feature, I will, of course, require it.
> But, I have never seen a feature that drives one vendor over another.
>
> Maybe (most likely) I was unclear, but what I meant was, I have never been
in
> a situation where a specific feature was so needed that it was a deal
breaker.
[Ron Hawkins] Yeah, that's what I said you said.
> The bottom line:
>
> If the features are 'equal', regardless of implementation details, cost is
the
> over-riding feature.
[Ron Hawkins] I have no problem agreeing with that. Your previous email
seemed to take a contrary position.
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
