Sunday, July 31, 2005 (A.G): IBM z9-109 Reprise Without Respite

Having thought about it, off-and-on for 5 days, while doing some digging around to get the material I needed to complete the new 'grids' I produced for the mainframe section, I have to say that I still standby what I said about the z9-109 last Tuesday [see below].  The z9-109 isn't worth getting too excited about.  Don't get me totally wrong on this.  I don't have anything against it and it sure looks like a decent machine.  It is just that it is not technologically new as some would like us to believe.  It is but a z990 mid-life refresh.  That is all.

Plus I am annoyed that IBM is back to their moronic game of changing brand names hoping that a new branding will magically bring them luck.  Talk about rearranging the deck chairs on the Titanic hoping that it would stem the flow.  IBM did it so many times on the networking front without avail and now they are doing it on the server side.  Just in case you have forgotten, cast your mind back to supposedly strategic IBM brandings such as NWays, eNetworks and SecureWays.  I am amazed we haven't had a NoWays as yet.

'Z' is such a dumb brand unless it is for a sleep related product.  I happened to see this picture on IBM's Web this weekend [and thus the source here is: IBM].  Notice the  'Z' at the top.  Looks to me as if this mainframe is snoozing.  So maybe IBM was right earlier this year when they said that mainframes are idle 40% of the time.  Maybe that is what the 'Z' is trying to tell us -- subliminally.  LAUGH.

Lets get to the brass tacks.  Cut to the chase, z9-109 is but a set of new z990 models.  The z9-109 and the z990 use the same MCM.  No wonder some at IBM were so bent out of shape that I started talking about the 16 PU z990 MCM so close to the announcement of the z9.  Ah!  Poor IBM.  I stole their thunder.  No matter.  As my charts in the mainframe section shows, the z9-109 models S08-S38 use the same 12 PU MCMs as do all the z990 models.  The only difference is that when it comes to the z9s, IBM does not insist on having 2 spares per MCM.  Instead sparing is only on the 1st MCM.  This thus gives you 10 CPs per MCM, on books 2 to 4 on the models, i.e. S18, S28 & S38, that have multiple books.  Bingo.  So no major breakthroughs here in terms of MCM technology.  And then we have the model S54 that finally gets around to using all 16 PUs on the MCM.  Wow, what a concept!

Yes, the 60 LPARs per machine is impressive.  But now that it is possible I am beginning to wonder whether we will ever want to have as many LPARs.  LAUGH.  Maybe never as much as 60 but it is nice to know that we can do it.

I will admit that the new Redundant I/O Interconnect (RII) capability that allows you to hot-swap a book out without disrupting anything, including I/O, made me stop and take note.  Yes, another big step when it comes to non-stop operation and that did make me glad.  SMILE.  But, and when it comes to the z9 I have a lot of 'buts' as you have noticed.  Can't we have this and the 60 LPARs on z990 models as well if we use the new books in the old z990 chassis.  Note the I/O [i.e. channel] configuration hasn't changed.  That reminds me of the z990, 16 PU RPQ that IBM told me about.  It is now beginning to all make sense.

Tuesday, July 26, 2005 (A.G): IBM Gives You the 'Zs' and NetManage Hopes You Are Asleep

Today's low-key z9-109 announcement from IBM really did make me yawn!  It was innocuous and irrelevant and the only thing about it that I found even vaguely amusing is that the model S54 (due to be available in November 2005) will indeed have 16 PUs per book (or MCMs).  That is funny.  Please, please read the z990 MCM entries below: particularly to the July 6 entry.  Please note Harv Emery's bold assertion that: "... MCMs with 16 working PUs. That has never been done because it was determined that it made no business sense to do it."  Less than 3 weeks ago I said: "So when IBM, as it must, announces bigger z990s with more than 16 PUs per book we will have to ask them to justify that.  OK?"  I am so disillusioned with IBM.

Please refer to my May 5 entry.  I said, and you can check that I am not making this up as I go along: "This 50% is underproductive and lazy!" as well as "For a start they should get rid of 75% of their so called marketing folks.  Let the products, many of them exceptional, speak for themselves".  Let me just say QED.

Getting rid of the zSeries branding is STUPID.

They are wasting our time and stockholder equity.  They should all be laid off along with the name.  Who cared.  eServer and all the other stupid ySeries brands were just IBM affectations.  We who live with this stuff still called it mainframes, Unix servers and AS/400s.  Do I have a tab that say 'zSeries'.  Today was enough to give you the Zs.  This was not a new family.  It was but a refresh; one that we knew was coming.  I can't get excited about this, though I will update my charts etc. once I get enough Zs.

To make it worse, this enough to drive you to Zs announcement came on the heels of NetManage's lackluster results.  You know there is no real antonym for 'synergy'.  We need to come up with one to describe NetManage.  They excel in making revenues disappear.  Give them 2+1 and they will conjure up 1.5 but try and make you believe that it is really 5.1 written backwards.  Others may be asleep at the wheel BUT despite my old age I can still be alert.  Revenues are down and for the second quarter they conveniently forgot to mention that Librados rather than boosting revenues is making revenues disappear.  They already did it once, quite spectacularly with Wall Data.  They took two $100M companies and created one that barely makes $48M.  And they are doing it again.  So a bottle of the best bubbly if you can come up with an antonym for 'synergy' that reflects NetManage's unique skill of taking 2+2 and coming up with 1/2. 

Sunday, July 24, 2005 (A.G): PureEdge Acquisition Extends IBM's Confounding Spree

On July 19th, the day after they announced their 2Q 'lets quickly gloss over the declining revenues' results, IBM informed us (slightly more boldly than the last time) that they now plan to acquire the PureEdge -- a Victoria, BC-based entity that is supposedly a leader in e-forms technology.  All I can say is here we go again!

This is IBM's 4th inane acquisition of privately held companies since May 10th: the others being Meiosys [June 23], Isogen [June 16] and Gluecode [May 10].  These, indubitably, are all fine companies in their own way.  But they do not warrant being picked up by IBM.  When talking about Isogen I pointed out the "why buy the cow when you have unrestricted access to the milk" argument along with what used to be Gerstner's criteria as to how IBM should prioritize their M&A efforts.  This is what irks and bothers me.  To use the somewhat hackneyed though still useful Nero adage, Palmisano is amusing himself buying up these piddly 'never heard of them' companies while IBM's revenues head South.

The financial community, mesmerized by the slight increase in income, totally ignored the fact that 2Q revenues were down and bought IBM's glib explanation that this was due to the exclusion of 2 month's worth of PC revenue.  [$0.55B PC revenues for April were included in these numbers.]  The revenue numbers are worrying particularly since we all agree that there has been an up-tick in IT spending over the last year.  And this continues to come back to my point.  By all means focus on M&A but do so along Gerstner's guidelines that resulted in Lotus becoming a part of IBM.  Don't waste time and money buying 'no name', privately held companies.  Instead think about buying Sun, Adobe or even Novell.

Sunday, July 17, 2005 (A.G): Forming A Queue For The Enterprise Service Bus

Posting Eric Newcomer's latest interview last weekend, which among other things talked about IONA's open-source ESB initiative [viz. Celtix] and how an ESB differs from our prior notions of a composite application server, proved to be unexpectedly salutary -- especially since I had also spent a lot of time, just prior to that, talking about SoftwareMining's COBOL-to-Java conversion solution with a very large U.S. bank.  I am beginning to believe that ESB and actual legacy software conversion à la SoftwareMining [replete with business process extraction as a 'no-charge' by-product] are going to become the new business drivers when it comes to EAI.

If I am not mistaken, in the last few weeks, we for the first time, are hearing the term 'ESB', from a plethora of companies including BEA (with its AquaLogic), Sun, IONA and IBM, with more frequency, resonance and conviction than that much maligned acronym 'SOA'.  It makes sense.  ESB facilitates SOA-based solutions particularly in the context of EAI.

When I talked about AquaLogic I pointed out that WRQ/Attachmate, SEAGULL and Jacada all claimed to be BEA partners, and as such could avail themselves to BEA's Java-centric ESB without in anyway appearing to be backtracking or being disingenuous.  I did a quick traverse across these vendors, as well as NEON, a few minutes ago, to see what they currently have to say about where ESBs fit into their EAI strategy.  Well you should go and check too.  While a few claim to have tenuous relationships with little known entities offering ESBs, it is safe to say that ESBs are not central to their strategies as yet.  This probably explains why SEAGULL abruptly stopped dealing with me on an interview I was hoping to do.  I had made the mistake of asking them about ESB.  Laugh.  I don't blame them.  I would be scratching my head too if I was in their shoes at this juncture as how best to tie together their disparate kit-and-caboodle especially if they don't have the equivalent of an ESB.

This said ESBs are still new and unproven so to speak.  So I worry that ESBs will become the latest excuse as to why new EAI projects have to be put on hold again.  This is where COBOL conversion with business process extraction slots in.  It is here.  It is intuitive.  It is not that difficult and it works.  So we now have another option.

Saturday, July 16, 2005 (A.G): IBM Powers Up the Layoff Machine

My May 5th BLOG entry, which addressed the news of IBM's latest round of planned layoffs, was entitled "13,000 Layoffs At IBM Just Isn't Enough!" -- and that was way before I got involved in trying to sort out this uncalled for z990 MCM controversy! [See below.]  Laugh.

Of late this site has been getting a lot of visits from IBMers, from around the world, looking for information on layoffs and 2Q results.  Thanks to them (and where they were coming from) I have learned some very interesting things.

I think somewhere in the murky corners of my cranium I did know that there were one or two ex-IBMer organizations that were hounding IBM over pension rights etc.  But since IBM claims that they handed over my pension to ITT UK, and I gave up trying to locate which STC organization would own up to being the offshoot of ITT UK, I had never really bothered to track down these organizations.

But then I found Alliance@IBM.  You need to go and visit them.

I am also discovering that, yet again, I might have been right on the mark -- and in the vanguard.  It appears that IBM may layoff more than 13,000 people.  While I do hate to see anybody lose their job, IBM, as I have maintained of late, once again has too much deadwood that has accumulated since Lou's perspicacious and much needed 'slash-and-burn' a decade ago.

It appears that the cuts will be deepest in Europe, with IBM UK (my old stomping ground) likely to be one of the hardest hit -- with UK mainframe experts being particularly vulnerable.  That is a shame since there are some worthy mainframe experts in the U.S. who should also be put out of their misery.  Remember the "mainframes are idle 40% of the time" faux pas?  q.v. April 14 BLOG.  There are many interesting stories, some of them at Alliance@IBM, such as the one about IBM UK mainframe folks being made to train their South African replacements.  There is some irony in this (and that is even without most of you understanding the cricket based rivalries that exist between these two countries).

While laying off US and European workers, IBM is said to be ramping up its Indian workforce.  I find that amusing too and I hope that 'Indian' in this context also includes Sri Lanka.  There is irony here too!  In the 1970s the Indian government unceremoniously booted out IBM because IBM would not relinquish their ownership rights.  [Some of you may also remember that IBM had to withdraw from South Africa during the 1980 trade embargo.]  So we are indeed seeing a new cycle.  I am all for it.  Out with the deadwood.  In with more people with commitment, drive and motivation.

Friday, July 15, 2005 (A.G): IBM's Harv Emery Pleads the 'Faith'?

We are still trying to get to the bottom of the z990 MCM issue [July 6], in particular to get Harv Emery to confirm his claim to a large bank that my BLOG entry is wrong and that a z990 MCM only contains 12 functional PUs -- despite published documentation, as well as explicit confirmation, from IBM that shows that Harv Emery may be the one that got things wrong.

I do not have any problems with being wrong especially since all I was doing was regurgitating IBM data.  Plus as I learned a very long time ago there is no shame in being wrong provided your motives were impeccably honorable.  So all that we (and in this instance I am including my contact at the bank) want to know is what is the truth here?  It doesn't matter who is right or wrong.  All that we want to know is whether it is possible to have more than 12 functional PUs within a z990 MCM as IBM has indicated.

As I have stated, I sought confirmation from IBM using my now 'main man' at IBM.  I do not need to specify who he is BUT anybody who is familiar with this site should be able to work out who he (and yes it is a he) is given that his name and picture appears more than once in these pages.  SMILE.  He, modest as ever, didn't try to answer this question himself since he thinks of himself as a 'software guy'.  When I informed him of the increasingly sad Harv Emery saga he told me that he got his confirmation from the 'distinguished fellow for z900 milli code'.  I think it is safe to assume that this person can easily pull rank over Harv Emery.  [Also note the graduation to milli code from the now old micro-code.]

Anyway I have e-mailed Harv Emery multiple times asking him to clarify his statements.  I have even left him a few v-mails.  Then today my contact at the bank informed me that there were to be two sessions in the upcoming Boston SHARE, and here I quote him: "the now infamous (and I take it so far quiet?) Harv Emery."  I e-mailed and called Harv Emery again and was amused to see that his v-mail greetings was out-of-date.  It is a small thing BUT I always find that it is an indication of one's discipline.  Well I finally got this e-mail from Harv Emery that I reproduce in its entirety to avoid any confusion:

"See you at Share.

Semper Fi!

Harv

zSeries Hardware Support Team Leader, WSC, Advanced Technical Support
VM: IBMUSM33(EMERYH)
Telephone: 301.240.2655"

Suffice to say it is a very curious e-mail.  Semper Fidelis.  Not just the motto of the U.S. Marines (as many assume) but also that of the City of Exeter in Britain, and the Royal Navy warship: HMS Exeter.  It is Latin for "always faithful".  I was confused by this response and immediately asked Harv Emery to explain.  What is he being faithful to?  The truth?  I had already discounted that this was in anyway a veiled reference to the august Marines.  I have had close dealing with some and used to even participate in an annual toast to commemorate their creation.  If I am not mistaken, I even paid for half the wine tab one year -- and here we are talking about a gathering close to 100 people (with my lapse of memory most likely associated with the amount of wine I consumed).  In my experience honor was sacrosanct to a Marine.  Needless to say Harv Emery has yet again clamped up.

So I will now reproduce one of the e-mails I received from my contact at this (as yet) unnamed bank -- and all I have done here is cross out the bank's name and expunge the first somewhat harsh sentance:

"My suspicion is that he answered the question based on his own base knowlege (or lack of knowledge) without doing any real research, and he simply didn't know the truth about the MCMs.  Now he's been caught, and he's in a very difficult position.  If he responds to you or to XXXXXX, he has to admit that either 1) he didn't know what he was talking about or 2) that he did know the truth and chose to mislead XXXXXX.  Either way, he loses - no upside for him, or for IBM.  He's probably been contacted by some more enlightened individuals within IBM (thanks to you!), he now knows he was wrong, and he's probably just hoping that if he ignores this, that you and I will just go away and leave him alone."

But we are determined to get to the bottom of this, before SHARE.  Please help us in this honorable endeavor to ascertain the TRUTH.  It is not about who is right or wrong, it is all about what is true and false.

Wednesday, July 6, 2005 (A.G): Who Can You Believe At IBM re. z990 MCM

The June 22 and May 24 BLOGs both dealt with the intriguing issue of whether the z990 MCM has more than the 12 PUs that IBM has so far admitted to.  Since I am a glutton for making sure that my facts are at least 98% right and sound at least 102% credible, I stated in the May 24 entry that I had seen an article in IBM's "Journal of Research and Development", Volume 48, Number 3/4, 2004 that led me to believe that a z990 could have more than 12 PUs.  I even reproduced a diagram from that article which I will show again here.

Try this link for an online copy of that article: http://www.research.ibm.com/journal/rd/483/winkel.html

If that doesn't work, use Google and do a search on "z990 MCM".  Click on "Images".  You will see the MCM picture reproduced here.  Click on that and get to the article.  [I am genuinely sorry that I have to give such detailed instructions but as you will see there are many who still do not seem to know how to use the Web for anything more than checking out free porn!]

In this article, under the heading MCM, IBM has this to say (and here I do nothing but cut-and-paste and Figure 2 is the picture reproduced here):

"The MCM is a 93-mm × 93-mm high-performance glass-ceramic substrate. The floorplan shown in Figure 2 illustrates the 16-chip MCM, which consists of eight dual-core processor chips labeled PU0–PU7 in the figure, four L2 cache chips denoted as SD0–SD3¹ creating 32 MB of shared second-level cache, a system controller chip (SCC), two memory controller chips (MC0, MC1), and a clock chip (CLK), which provides the clock distribution as well as pervasive function for the MCM. The MCM was designed using the methodology described in [1, 5] with upgraded noise-checking tools and routing capability."

In the June 22 entry I said that I had received confirmation from IBM that the z990 MCM did indeed have 8 dual-core processors and hence the possibility of having 16 working PUs providing that none were compromised during the manufacturing process.  I won't divulge my sources, since there is no need, but lets just say that he is a well know name in mainframe circles.  It was also IBM that told me about the RPQ.

I received an e-mail last week from a reader (and BIG Iron club member) that he had received from a Harv Emery, zSeries Technical Lead at the Washington Systems Center. This is what Mr. Emery states, and again I just cut-and-paste the e-mail I received:

"The original blog entry is wrong. There are only 12 functional PUs on a z990 MCM. Of the eight PU chip sites, four are populated with chips with two PUs (dual core). The other four are populated with chips that have only one PU. This notion probably arose because early engineering and design work did allow for the POSSIBILITY of building z990 MCMs with 16 working PUs. That has never been done because it was determined that it made no business sense to do it.

RE sparing: On a z990 MCM, the two spare PUs in each book reside on a single dual core chip. The spare PUs in this chip can be used to spare a failed dual core chip (both PUs) or can be used individually to spare failures on chips with one PU.

And, finally, for completeness, the z890 MCM has only five functional PUs. Five of the eight PU chip sites are populated with PU chips with a single PU. The other three PU chip sites are empty,"

So who at IBM can we believe?

I know the amount of checks, reviews and approvals that articles that appear in IBM's "Journal of Research and Development" have to go through.  Like me, the editors of that publication appreciate veracity and credibility.  So Mr. Emery is emphatically contradicting IBM's own published article.  He also makes two amazing claims (and one of them, amusingly, isn't even that my blog entry is wrong).  He claims that there is no business sense in having 16 PUs on a z990 MCM.  PLEASE make note of that.  So when IBM, as it must, announces bigger z990s with more than 16 PUs per book we will have to ask them to justify that.  OK?  Mr. Emery also states that IBM will go to the trouble (and expense) of making two very different chips: one with dual cores the other with single cores.  No wonder the financials looks iffy.  As is my wont, I have asked Mr. Emery to respond to his claims, albeit via the reader who forwarded his e-mail to me.  That was on June 28.  I will keep you posted since our goal here is the indefatigable pursuit of the truth.

Monday, July 4, 2005 (A.G): Microsoft Wins Again

The $775M that Microsoft agreed on Friday, July 1, to pay IBM to settle their long-running antitrust dispute seems like a lot of money.  With the distraction of the BIG, holiday weekend nearly upon them most people, in particular the financial talking heads, appeared to think that this was an appropriate (and somewhat resounding) slap on the wrist to Microsoft for raiding the desktop OS cookie jar.  My reaction was this is but bagatelle.  Thus I am pleased that at least a few in the media have pointed out that this settlement includes 'damages' for the loss of OS/2, since most people in today's financial sector don't even know what OS/2 was.

$775M is nothing!  IBM loses 20 times that, year after year, for not having its own desktop.

By hijacking OS/2, Microsoft didn't cripple IBM.  It castrated them.  IBM still feels the pain.

Agreeing to the additional $75M credit so that IBMers can use Microsoft software, in particular MS Office, is adding insult to this injury.

$775M!

Microsoft wins again.