Thursday, April 9, 2009

When two tribes go to war ...

Last weekend I was in the Bay area again, where I found time to return to Los Gatos. For me the area between Cupertino, Saratoga, and Los Gatos holds so many memories and I seem to gravitate back to familiar haunts each time I’m there. It was early in 1988 when I began to commute between Cupertino and Sydney, Australia, and I would catch a return flight to the Bay on a Saturday, and head to Los Gatos first thing Sunday morning – for coffee at a small café on Santa Cruz Avenue. It’s still there, and this Saturday afternoon found me sitting by the front window.

Located not too far from the café is one of my favorite car dealerships in all of California – Ferraris of Los Gatos, now called Silicon Valley Auto Group - and with each visit to Los Gatos I somehow manage to find the time to stop by, and to check out the showroom. On Saturday, there were two Bugati Veyrons on display – not too shabby at all! And the picture at the top of the page is of the view across the lot from the rear showroom.

Get to know Jack (who has been there for as many years as I can remember), and there’s not a car he will not let you sit in, or pour over from every angle, for as long as you want. Business doesn’t look good, unfortunately, in this down economy and the foot-traffic was as light as I have ever seen it – with the market for exotics way off the highs of only a year ago. Where have the enthusiasts gone?

In the café, I had picked up the Wall Street Journal and read an article, “The End of the Affair,” that laments the end of our love affair with cars. “A key to any comeback by Detroit,” the subheading read “is connecting with car fanatics.” The story goes on to add “perhaps most worrisome for the industry is the recent disappearance of that iconic American character: the new car enthusiast, otherwise known as the tuner, motor head, speed freak, buff, nut, and zealot.”

All is not completely lost – they are still out there, watching as inventories rise and prices tumble. But today there is a new breed of car buyer who focuses on “practical matters such as price, fuel efficiency, and reliability,” a far cry from the past when sales made to the fanatic were solely driven by the “romantic allure of a new vehicle.” Two groups of car buyers, each responding to “internal wiring” the other group couldn’t possibly comprehend!

In the March 9, ’09 issue of BusinessWeek, was an article “The Next Net,” and the author observed that “the mobile Internet is generating oceans of data about people’s movements and predilections … companies … are sifting these data to place people in behavioral tribes.” The author then explained how one company’s computer, analyzing the data from mobile phones, after a few weeks, “usually has enough data to place (an individual subscriber) into a tribe – a group of people with common behaviors.”

Tribes? The thought that we could be grouped into tribes made me a little uncomfortable, but then again, thinking back to the way we buy cars perhaps we’ve always retained some genetic link with our past.

For as long as I can recall, the computer industry has fostered tribes – whether you believed in mainframes or minicomputers, or centralized versus distributed computing, or even networks with SNA versus TCP/IP. Preferences for any of these solutions influenced our behavior and we just naturally gravitated to other, like-minded users.

Tribes evolved from the communities unified by the deployment of the same solution. They tenaciously defended their position and aggressively fought to neutralize competing viewpoints, as if they were a threat to the tribe. Verbal sparing was often intense with emotions running high – participants who changed their positions were betraying the core values of the tribe and were subjected to public condemnation.

On April 7th, 1964, forty-five years ago this week, computing history was made with the announcement of the System/360 mainframe by Thomas J. Watson, Jr. Perhaps there’s no deeper division than exists between the mainframe tribe, and everyone else. As vendors have fallen by the wayside, there remains just IBM and HP as the lone dominant forces within the data center. I can recall Jimmy Treybig unveiling the latest Tandem product Cyclone, in October 1989, and proclaiming that Tandem would be “one of three” – alongside of IBM and Digital – as the only viable options for enterprise class computing.

With Jimmy as chief evangelist, innovative fault tolerant transaction processors captured the imagination of those of us working for Tandem Computers, and we enthusiastically championed the fundamentals of fault tolerance, scalability and data integrity. And today, our passion for continuous availability within the enterprise can be traced back to those times. With our behavior shaped this way, is it any wonder we became a tribe?

I was reminded of this as I sat through an early presentation on the z10 mainframe last year. The IBM speaker was talking about the reliability of the mainframe – and about the considerable improvements they had seen in the manufacturing process for the new multi-core Power chips. When asked about how many “nine’s of availability” the z10 provided, the presenter pushed back with “Nines! What is this obsession with nines! Leave the number 9 to the Beatles!” There was a little light laughter from the audience but to this day, IBM doesn’t talk as fervently about availability as does HP with NonStop.

Bill Highleyman, the author of many papers on the subject of availability emailed me recently to say “I don’t see much coming out of (IBM) on continuous availability. Sysplex is about it, and it is not used very much I believe.” And when I approached a major solutions provider familiar with IBM's approach when facing an existing HP NonStop user, and asked how IBM positions their mainframe against NonStop, the response was similar to Highleyman's. "Continuous availability isn’t part of IBM’s pitch when comparing the mainframe to NonStop," was the response. Instead, “IBM has the mindset of productivity through highly configurable tools, without regard to performance (or availability)!” It is this kind of rhetoric between the two companies that drives their adherents together as tribes.

In the posting of March 5, ’09 “Tough Neighborhood” I wrote how “the standards set by early Tandem products were uncompromising when it came to availability … everyone who worked on Tandem was part of a community focused on continuous availability.” I then added how “NonStop hides so much magic inside the box that today, many CIO’s have become reluctant to move off of NonStop.” This is a far cry from the messages coming from IBM where it’s productivity and performance that are given more prominence.

And over less issues have “two tribes gone to war!” to paraphrase the song by “Frankie goes to Hollywood.” Perhaps not in the conventional sense, but IBM wants to retain supremacy inside the data center and with NonStop, HP has a legitimate challenger with greater availability straight out of the box. IBM wants to do everything in its power to marginalize the product, labeling it legacy and proprietary!

Yet it is IBM that continues to manufacture the mainframe from proprietary components. In the posting of March 26, ’09 “Knives in the Library” I wrote of how IBM has no plans to adopt industry-standard blade packaging continuing to rely on outdated, proprietary “Book” packages. While HP can deploy Blades in support of all of its operating systems, only the mainframe can use IBM’s Books and this will severely limit future mainframe’s open story. Perhaps the IBM tribe prefers to ignore open systems after all!

The war between the tribes is sure to heat up further in a deteriorating economic climate. And I am reminded again of one of my favorite quotes from Road and Track writer, Dennis Simanaitis. who wrote in September, ’08 of how “there are plenty of maybes, a fair amount of hard news … and even more hype!” I am a new car enthusiast, and from my behavior, easily recognizable as a member of the car fanatics tribe.

IBM may want us to leave the number 9 to the Beatles, but for the NonStop tribe, availability will always remain a “call to arms” and, should IBM elect to ignore it, restraining the combatants may be difficult. And should we be surprised? They have filled in the swimming pool but the skills we developed growing up in the tough neighborhood that was Tandem remain, and we stay as focused on availability as ever. There is probably no greater “worse case scenario” today than a “down system” – surely, we should be doing everything in our power to eliminate the system itself as a likely source of an outage.

Building reliable systems is all well and good, but if the package cannot deliver continuous availability, then it is under-serving its core constituency and for the tribe that is mainframe, this could be a vulnerability that generates the hard news nobody at IBM cares to read about.

And from this member of the NonStop tribe, this isn’t more hype!

3 comments:

Maria Olivero said...

So now instead of the tie tacks that said "HFIADS" (how fast is a down system) maybe new ones should be issued as "HPIADS" (how productive is a down system)?
Rot's o ruck IBM!

Anonymous said...

Ever heard of Parallel-SysPLEX or GDPS?

With DB2 Data-sharing and Hyperswap you can achieve Continuous Availability.

Banks in the US and Canada have operated Workloads with 100% uptime for many years - on particular Canadian Bank has had their workload up for 12 years and have never had an outage.

Richard Buckle said...

Yes - I have (re: Parallel-SysPLEX or GDPS); but my experience / observsations suggest that the set-up and management of such an environment is considerable and the resultant cost pushes the solution way beyond what you routinely get with NonStop.

Everything is possible - but you are talking about an architecture with origins in batch workloads, and everything we see today has been layered on top, or thrown alongside, and compromises have been made with every step. Even within IBM there's not really a "reference model" for deploying Sysplex and every customer is approached differently.

But yes, I have talked to IBM customers who do have uptimes stretching for years - but it's always come with a significant price tag.