Friday, April 22, 2005 (A.G.):
Dual-Core Chips -- We Have Had Them Since 2001!

The PC world should be excited about the dual-core chips now being produced by AMD and Intel.  Dual-core chips reduce overall cost, consume less power and in-time reduce complexity.

The PC media, understandably, is having a field day ... actually more like a 'field month'.  But contrary to some of the claims in these 'tabloids' don't get misled that dual-core technology is brand-new and that the PC world is boldly venturing into uncharted territory.  This is not so.

IBM introduced the POWER4 (right), a 64-bit processor, dual-core chip in the Summer of 2001.  The IBM p690, that so eclipsed Sun's new unveiled Sun Fire 15K, was announced in October 2001 == and featured the dual-core POWER4.  Since then we have had the POWER4+ and as of last year the redoubtable POWER5.

Many of the issues now being addressed by the PC community such as relative performance, cache memory implication, bus design etc. have already been well thrashed out with the POWER4/5.  We also have various, extremely successful servers, with both single and multi-chip designs.  So if you want to really get into this dual-core stuff, it may help you if you begin by looking at the POWER4/5.

Obviously there is a ton of material on POWER4/5 on the Web -- especially at ibm.com.  There is even a POWER community.  That is all.  Just wanted to help out here.  Have a great weekend.

This was quite a week with those two BIG announcements on Monday.  Cheers.

Thursday, April 21, 2005 : Burt Gearhart, Applied Computer Technology, Rockwall, TX
Frank Pitt (of Attachmate) - As I Knew Him

Anu: I saw your IT In-Depth piece “The Pitts for Frank” and thought I would share some information about Frank Pritt when I worked with him at Harris Corp. Data Communications Division in Dallas, TX in the 70’s.

Frank was a Product Manager at Harris 70’s and I was a Section Head in Product Development for IBM compatible products. Frank was known to all of us in development as “Frank-O”. Harris Data Communications (HDC) produced a line of Front-end Communications Controllers and Remote Job Entry (RJE) terminals called “COPE” (Communications Oriented Processing Equipment) for IBM, CDC, Univac, Honeywell, and Burroughs mainframes. Frank-O was the product manager for the COPE Front-ends.

I worked extensively with Frank-O on a IBM 2701/2702, and 2703 compatible Transmission Controller called the COPE 4705 (a play on the IBM 3705 name). We OEMed the 4705 from a company called InterComputer in Phoenix. I inherited the testing and support of the 4705 and Frank-O inherited trying to figure out how to sell ‘the pig’.

The 4705 initially had many problems due to incomplete testing by my group and trying to compete with the higher function and power of the IBM 3705. But we eventually got the quality problems fixed and sold a number of the units.

Frank-O was helped extensively in his job by HDC’s General Manager (as were the programmers, engineers, sales force, field engineers, and the cooks in the lunch room). The General Manager knew more about each employees job than then the employee did. Frank-O grew tired of the constant meddling by the General Manager and decided to transfer to sales and became HDC’s salesman at Boeing Aircraft in Seattle.

Frank-O was almost an instant success on the Boeing account and was quickly earning more money than the General Manager. This angered the General Manager so he decided to cut the Boeing account in half, with Frank-O and Jeff Leffingwell being the salesmen. Frank-O still did pretty well.

But, he noticed what was being done with PC’s and the IRMA card in Emulating 3270 Terminals at Boeing. Frank-O was not one who fell off of the pumpkin truck very often, and immediately saw the opportunity. He quit Harris and started Attachmate with Jeff and his wife, Julia.

I also decided that I had received enough help from the HDC General Manager and started my own company, Applied Computer Technology in October of 1980. I decided to focus ACT on SNA testing, since that was one of my biggest problems as Director of Product Development at HDC. One day my phone rang and it was Frank-O who had heard about ACT’s SDTF (SNA Development Test Facility) product and wanted to purchase one. Frank-O’s only problem back then was that he did not have the money to pay for it. So we worked out a deal where he would pay for it over four months --- Ha Ha dream on Burt!

Every month for the next year I had discussions with Frank or Julia about getting paid. I eventually did get paid all the money. Frank-O’s sales talent began to show and shine as Attachment grew and grew. Frank-O decided to hire a President and other management to run the day to day affairs of Attachment. He then migrated from Seattle to Southern California where he spent his time Roller Blading and Attachmate was spending it’s time going down the Tubes. Frank-O was then forced to return to Attachmate and take over the Wheel again.

I have not talked with Frank-O for many years was surprised by your story that he waited so long for some well deserved time off and maybe a return to Roller Blading.

<< Read Burt's fascinating bio. >>

Wednesday, April 20, 2005 (A.G.):
The Pitts for Frank

Whatever you may think of Frank's handling of Attachmate during the last decade it would be very remiss of us if we did not acknowledge that Frank was, in the 1980s, a visionary and a pioneer.  He, more or less, single handedly got PC-based host access chugging along.  The merger with DCA, that created a $400M superpower of the day, was typical of Frank's then creativity.  In the early 1990s he also acquired CSI, a boutique publishing house, and was the proud publisher of "SNA Perspective" -- a very prestigious publication of its day (even though it had me as a major contributor!)

Though I have spoken at a number of high-profile Attachmate sales events, I don't remember ever meeting Frank.  That was OK.  We probably don't have much in common and he definitely would not have wanted to hear my opinions on how he should be running Attachmate.  He was savvy like that.

I very much doubt that this was the exit strategy that Frank would have liked.  He should have gone public a long time ago, taken the cash ... and retired.  This in many ways is a capitulation to its nemesis.  As such, I kind of feel bad for Frank, though he will no doubt be consoled by what he makes on this deal.  He now should have more time to pursue his passion for antique car racing.  Maybe he will call me and invite me to navigate for him.

Well Frank, I for one wish you the best and goodbye.  You created an institution.  You were a legend.  I will try and remind people of Attachmate, ten years down the road and maybe even mention you.

Tuesday, April 19, 2005 (A.G.):
Autonomic Serendipity

Discovering Opalis with their exciting take on automated management realized via autonomic computing related concepts was serendipitous to say the least.  They popped up as a Google AD on guruge.com because of the references in their to my work on autonomic computing.  As soon as I read the one line description I knew I had to talk to them.  As the interview clearly demonstrates I was not in anyway disappointed.  It was Serendipity.  I wear ring that has the word ‘Serendipity’ engraved on it.  So this, a word whose origins hark back to Sri Lanka, was special.

I had just finished writing two white papers on managing mainframe-centric IP networks.  In those I talked a lot of about automated alert generation.  What I like about Opalis is the automated, policy-based alert processing – without human intervention.  Obviously the notion of automated, intelligent (or heuristic as we used to call it) management is not new.  IBM, with NetView and SystemView, started talking about lights-out management (LOM), i.e. unattended data center operations, over 20 years ago.  It, however, in my opinion is still but a holy grail when it comes to large data center operations – particularly when mainframes are involved.  Opalis, moreover using IBM sponsored de facto standards such as the Common Base Event (CBE) format, is showing us what can be achieved in this space.

Autonomic computing technology has the potential to reinvent data center management.  We all complain about the complexity (and thus the associated cost) of contemporary data center management.  So here is a way for us to finally make some great strides.  Now you can see why I think this was Serendipity.

Monday, April 18, 2005 (A.G.):
Adobe: The Macro Economics

Having gone on record, last Tuesday, as saying that Adobe Acrobat is the flakiest piece of software that I have had the misfortune to encounter in 39 years of using computers, I, for one, am not thrilled that Adobe intends to acquire Macromedia.  This is not good news, in any shape or form, to the faithful legions of Marcomedia users (with me being a loyal, but relatively ‘light’ user) in particular, and to the graphics designer community in general.  The media claim that a combined Adobe/Macomedia will have more clout to compete with Microsoft is ‘hokey’ to say the least.

For a start, this acquisition is going to further reduce competition and that is always bad given that this is the bane of today’s desktop software industry.  If Acrobat had some credible competition, Adobe would have been forced to rectify the manifold problems that plague this product.  Now we will see more job cuts and product withdrawals.  In reality, there is more product overlap between these two companies than between either of them and Microsoft.

In my opinion, there is no significant overlap between the products of these two companies and Microsoft.  Where there is some overlap [e.g. FrontPage], the target audiences are very different.  Microsoft offerings are invariably targeted at ‘casual’ users while Adobe/Macromedia DreamWeaver, PhotoShop, FreeHand etc. are the tools of the trade of dedicated graphics professionals.

So I don’t get how this acquisition is really going to impinge on Microsoft.  Instead, what we are likely to see is the demise of FreeHand, Fireworks and GoLive – with the last, indubitably, being the most expendable.  The long-term competition between Illustrator and FreeHand was extremely positive and productive and ensured that designers and artists were extremely well served.  Now we will lose this competition and with it a lot of jobs.

This is a sad day for the software industry.  I am not going to be consoled even if Adobe resurrects its long beloved image of Simonetta Vespucci per Sandro.  Given Adobe's track record they will probably try to animate it using Flash!

Friday, April 15, 2005 (A.G.):
Sunsets at Armonk, NY

IBM's and Sun's below par results for 1Q2005, trotted out over the last two days,  certainly rattled the market -- with IBM taking much of the blame for today's precipitous drop in the Dow.  In reality I didn't think IBM's numbers were that disturbing.  Given that they have been claiming all year, until yesterday that their servers are grossly underutilized, I am not surprised that some customers delayed placing big orders!  q.v. April 14 entry.  That last statement, if you haven't worked out my flippancy by now, was somewhat tongue-in-cheek.  I am not letting IBM off the hook that lightly for that faux pas.

Methinks, and I hope I am dead wrong, that IT spending for 2005 isn't going to be as lavish as everybody has expected it to be.  Yes, indubitably, we saw a marked up-tick in IT spending sentiment towards the last few months of 2004 that spilled over to 2005.  But lets face it, we were well overdue for a bit of relief given the drought that we had experienced since late 2001.  What we had to determine was the driving forces that were going to sustain this increased spending.  SOA definitely isn't one of them, despite what others may be deluding themselves with!  I actually believe that all this fuss about SOA is a reason that spending has stalled.  SOA is another technology, supposedly strategic, that now needs to be evaluated and justified.  But think about SOA.  The whole rationale of SOA is that of reusing existing applications -- in situ.  Yes, we might need more capacity for the new SOA-based composite applications, but not as much as with prior methodology.  That reuse factor, what I referred to in my last book as software recycling, now kicks in.

All this talk about grid and utility computing isn't helping matters either.  If I was an IT manager I am sure that I would start nervously looking over my shoulder to see if the CFO is going to tap on the shoulder and ask "are you sure we don't have a ton of unused computing stuff sloshing around ...?"  I don't have time, anymore, to read the Journal ... the Wall Street one.  Do any of you know whether they have done a piece on grid computing yet?  If not, it will only be a matter of time.  And then the fur will fly.

Sun has decided to jump straight into the utility computing fray with its $1/hour deal.  No wonder they continue to feel the pain.  Sun, as I have maintained, for the last 3 years -- ever since IBM's POWER4 started to make its mark -- has problems.  It doesn't have the resources, any longer, to effectively fight all the battles that it has now fight.  Java, though successful, has not lived up to all of what Sun and others expected it to do.  It galls me to say it but .NET has a lot of fans.  And then there is Linux.  If that wasn't enough, they also have to contend with the POWER5 and IBM's new consolidated i5 mid-range that supports AIX, Linux and OS/400.  Sun needs a Gerstner.  Somebody that can take an unemotional look at their overall business and get the senior managers to make some hard choices.  That might take some time.

Have a great weekend.

Thursday, April 14, 2005 (A.G.):
IBM Retracts the "Mainframes are Idle 40% of the Time" Claim!                   read the claim

On my very 1st entry on Monday, April 11th, I complained that I was miffed at IBM's claims that mainframes were IDLE 40% of the time. I then contacted a few friends at IBM and told them that I was making an issue of this.  Today, 3 days later, I get a very nice e-mail from IBM saying that the statement was indeed erroneous and that they have taken the article down.

<< It is still preserved here for posterity (in the Mainframe Archives section in Mainframes) for those that insist on reading it to make sure that IBM really did make these preposterous claims. >>

That was a great New Year's gift for me -- today being Sri Lankan New Year.

IBM rightly points out that utilization is in the 70 - 90% range. I knew that.  Strange that IBM's VP of Grid Computing didn't do anything about this. But he isn't a mainframer like me.

Wednesday, April 13, 2005 (A.G.):
WAN Bandwidth Optimization, Riverbed and John Doe

WAN bandwidth, notwithstanding the misleading impression created these days by the Web, is neither free nor infinite – even on the Internet.  Obviously a vast majority of Web users, many of them plodding along on a 56Kbps dial-up links, have no concept of what bandwidth is – and, in reality, they should not have to worry about such arcane plumbing details related to networking and interconnects.  That, of course, is the abiding beauty of the Internet.  But, the profligate squandering of Internet bandwidth, by us all, continues to amaze me and occasionally bother me (such as when I received nearly 50MB of PowerPoint presentations via e-mails from a vendor).  Obviously things would be very different if we all had to explicitly pay for the Internet bandwidth we use – as with ALL other utilities such as electric, phones/cell or gas.

However, this ‘flat-charge’, use-as-much-as-you-want bandwidth model of the Internet does not extend to corporate WANs.  Administrators and users of private corporate WANs know this to their cost and chagrin – irrespective of whether they are using ATM, Frame Relay, T-1/T-3 links, ISDN, or even VPN.  Nonetheless, corporate WAN traffic, bolstered, for a start, by the ever burgeoning volume of e-mail, continues to sky-rocket impacting overall network performance and in particular the response times of mission-critical applications.  Thus it is no real surprise that WAN issues plague most corporations with any type of distributed operations whether these be branch offices, overseas manufacturing plants or dispersed divisions.

IT managers are expected to deliver LAN-quality throughput across such dispersed networks, but without spending too much on equipment or additional bandwidth.  But WANs, as networking professionals know, tend to use protocols, now often referred to as ‘chatty’, that curtail throughput.  Plus, WAN speeds will always lag behind the networking speeds available on today’s 100Mbps Fast Ethernent or Gigabit Ethernet LANs.

So to try and economically overcome this prevalent, productivity-sapping WAN bandwidth shortage, more and more companies have been looking at WAN ‘accelerator’ technology.  Obviously, this technology, though sometimes incredibly impressive, is still governed by the laws of physics (though I have encountered at least one Israeli company that would lead you to believe otherwise).  So in the end it has to end up relying on four basic techniques: compression, caching, protocol optimization and prioritization.  There are limitations to all of these and reliance on widespread caching, in particular, bothers me.

Against this backdrop, I am intrigued by the rapid strides been made by the 2002 start-up Riverbed [www.riverbed.com].  They claim to have got their first 100 corporate customers within 10 months of their first products (predictably called ‘appliances’) hitting the market.  Since then they have also won much praise, media coverage and awards.  I have looked at their story … and having worked with some other products of this ilk, I know this technology quite well.  I have to say that I was impressed.  So if you are looking for a corporate WAN bandwidth expansion solution spend some time looking at Riverbed.  You might find it rewarding.

Here is my ‘chuckle’ of the day.  Somebody showed me a Microsoft presentation which was by “John Doe”.  That really did tickle me.  I just wonder how many Product Managers out there would be better off if they were known as John or Jane Doe (or even Roe) rather than by their real names.  Laugh.  It is good for you.

Tuesday, April 12, 2005 (A.G.):
Whitewashing Adobe and Managed Source

When I saw the headline “Adobe Launches Linux-Compatible Reader” on Google this morning, with my first zap of caffeine for the day, my first reaction was that this was but a “April Fool’s Day” press release that inadvertently got held up for two weeks!  No Adobe Acrobat Reader for Linux?  How could that be possible?

Then I read the story.  Not even a tempest in a teacup.  More like coffee grinds in a Java pot.

There was a Acrobat 5 for Linux.  Adobe bypassed Linux with V6 and have just released V7 for Linux.

Well no great loss here.  As somebody who uses Adobe Acrobat, both to create and read documents, 7 days a week, I would be plenty happy with a V5 – provided it worked.

But Adobe has done us all a favor.  It highlighted the need for an Open Source Acrobat – or to be more precise Open Source PDF.

I should have thought of this earlier because I detest Adobe Acrobat.  In my 37 years of using computers, I can safely say that Adobe Acrobat is the flakiest piece of production-level software I ever had the misfortune to use … and this from a guy who, as an IBMer, even used IBM software in the 1970s! (Laugh. It is good for you.)

Great news.  It appears that there are some very real Open Source PDF offerings out there.  It didn’t take me long, thanks to Google (and an excellent December 2001 article that it dug up for me from a Cameron Laird), to find references to PDFlib, Etymon and ReportLab.  True, that none of them offer a direct alternative to Acrobat Reader per se, though from what I can see the “PDFLib + PDI” combo will let you read and manipulate PDFs.  That is OK with me.  I don’t have Acrobat Reader either.  Instead, since I have the full blown Adobe Acrobat 6.0.1 Professional, it is this that kicks in when I click on a .pdf.  So for me, PDFLib + PDI will be similar.

How many of you use Acrobat to create documents, capture Web pages or print Web pages.

Don’t get me wrong.  It works and when it woks well it can be a delight.  But it is flakey.

We all like to beat on Microsoft, deservedly, for the lack of robustness and resilience of Windows and Office.  But I have to be fair.  Windows XP Professional (without SP2) is livable.  Office and FrontPage 2002/XP can, even with the hoops I make them jump through each day, go for weeks without totally screwing up my day.

But combine Word and Acrobat.  It is like pouring water into hydrochloric acid.  V6 improved things.  After a year of struggling and having crashes every day … I yanked Acrobat out of Office.  Alak, I still have to use Acrobat.  But not for long.  I am contacting PDFLib as soon as this is posted.

But this got me thinking about Linux again.  Some of my view about Linux and say IBM have already appeared here.

I still think there is too much lip service and too little genuine commitment.

I sometimes think that Linux has become the ISO OSI of the 21st century.

Ironically many Linux folks probably have never heard of ISO OSI.

But remember for 20 years (and I know I because I was counting) it was going to be the next big thing … very, very soon.  Then.  Poof.  The Web comes along and OSI joins the dodos.

So is this what will happen to Linux?  It will pave the way to become sideswiped by something ‘better’?

And the last thought of the day.

Enterprise Open Source is an oxymoron.

When it comes to the enterprise it will be MANAGED SOURCE.  A subtle but big difference.  OK?

Monday, April 11, 2005 (A.G.):
IBM’s idle mainframes, Farabi and the New Year

No, that is not a typo.  I really do mean IDLE mainframes and IBM mainframes at that.

But some background first.

It has become Pavlovian -- already.  It is close to mid-April and I keep on expecting to hear that IBM is going to be making a meaningful mainframe related announcement.  To be fair there is no real tradition, as yet, that April is to be red letter month for mainframes in IBM’s revised and now very methodical (if not mechanical) release calendar.

Last April was obviously special.  April 7^th, 2004 was the 40^th birthday of IBM mainframes.  To commemorate that special day IBM unveiled the variable-capacity z890 and the still beguiling zAAP “Java-grinder” – though in my opinion the USPS was somewhat remiss in not helping IBM out by issuing a commemorative stamp to mark this landmark, especially since they used to be such a big user of IBM technology.  Talk about misplaced priorities.

And prior to that in 2003, albeit in early May as opposed to April per se, we had the ground-shaking T-Rex launch.  So the funny thing is that I had, until just reminded by a cousin, forgotten that April 14^th is the Sri Lankan New Year but remembered that IBM makes mainframes announcement around now.

This said, I am of two minds (and I am acutely aware that some of you question whether I have even one mind let alone the capacity to be of two) whether we need a mainframe announcement, at all this year – unless it has to do with more granularity on the Capacity On-Demand (CoD) features or real performance numbers for zAAP.  I have to confess that any mainframe announcement will undoubtedly make my day and immediately give me the grist for my next Deep Blue column in zJournal that is rapidly becoming due.

But, I am miffed with IBM vis-à-vis mainframes on two counts.  One serious.  The other not as.

Methinks they are letting the side down and making the life of us, pro-mainframers, unnecessarily difficult.

My main gripe is IBM’s unsubstantiated and bewildering claim that mainframes are idle 40% of the time.

This claim has now been around IBM’s Website, albeit in the esoteric Grid Computing hinterland, since at least February (when I first saw).  It is still there today: http://www-128.ibm.com/developerworks/grid/newto/index.html 

I keep on expecting … nay hoping that IBM will have the decency to change it especially since I talked about this claim, explicitly, with Ken King, IBM’s V.P. for Grid Computing at the Globus Conference in Boston on the 9^th of February.  Though Ken was not familiar with this posting per se – he was also, to my surprise (or was it chagrin) not perturbed by it.  But then again he had already told me that he didn’t have much experience with mainframes.

Anyhow, I guess you understand my pain!  If mainframes really are that idle then the last thing we need to worry about is bigger and better mainframes.  Lets just work out how we can increase the workloads – without IBM having to put quite a few ‘idle loops’ into z/OS 1.6 to jack-up utilization.

That also brings me to the zAAP.  I have been looking at it, quite closely, for the last few days because zAAP support is a featured highlight of the new WebSphere Application Server for z/OS V6.0.1.  More on this, in these pages, this week.  But if we have all this idle capacity why squander $125,000 on an ‘unrated’ [i.e. no published performance specs.] zAAP.  Lets just use the capacity that is just sitting there idling.  [My tongue could be in my cheeks.  SMILE.]

Well, on Sunday I had the pleasure of having brunch with my friend, of many a long year, Michel Kassar – the president and the co-founder of Farabi Technology in Montreal.  I hadn’t broken bread with Michel in a very long time and it is always good for me to chat with him because Michel, as a true hands-on 20 year veteran of host access, knows more about the Web-to-host market than most – especially when it comes to the nuances of the European and Middle-East markets.  While he vouched for the interest in SOA he, to my delight, confirmed a nagging doubt that I had been harboring for the last few months.  Though there is a lot of buzz about programmatic adapters, much of the traditional base, especially the huge AS/400 market, is still very focused on conventional ‘screen-scraping’ centric techniques.  Hhmmm.  Food for thought.

Thank YOU.