I spent the weekend in Cupertino. So much has changed over the past twenty years that the town doesn’t look familiar – not the buildings, as the structures are the same, but the names on the tombstones lining the streets. Some familiar names remain, but most I just didn’t recognize. Who would have guessed Apple would have taken over so much of the city, and who would have guessed that so little from the glory days of Tandem would remain. Kind of depressing, after a fashion.
It was a cold weekend and snow had fallen on the ridgeline of the mountains to the east of Silicon Valley. The blustery conditions only seemed to amplify the gloominess of the place and as I slowly drove down Vallco Parkway I noticed no security fence blocking the eastern entrance to Building 3, so I pulled into the rear parking lot. With no guards visible, I walked through the rear gates and onto the basketball court adjacent to the swimming pool, behind the canteen of Building 2.
The picture at the top of the page, taken in the fall of 2008, is of this area with the Tandem chevron clearly visible at the bottom of the pool. But last weekend, as I looked back across the same area, there was no pool. The photo below is of what the area looks like today – look closely and you will see the safety gate still visible on the right of the photo. But the Tandem swimming pool wasn’t there – only a circular grassy area where it used to lie.
A former Tandem employee who follows this blog emailed me earlier this year to tell me that the pool had been filled in, but I couldn’t bring myself to believe it. Not possible! Too much tradition – surely, if nothing else remained of Tandem’s presence in Cupertino, there would always be the “blue” Tandem chevron visible at the bottom of the pool. But today, it’s just another grassy retreat. I have always been a fan of Bruce Springsteen and all I could hear in my mind was a verse from his anthem, My Hometown:
“They’re closing down the textile mill, across the railroad tracks
Foreman says these jobs are going boys and they ain't coming back”
There comes a time when manufacturers rationalize processes and standardize components. They embrace broader, company-wide development and manufacturing. And Simplify! Simplify! Simplify! Everything that goes into a modern server is re-visited, time and again, to ensure more standard components can be used. Complexity will always be present, but standard interconnects, standard channels, standard communications augmenting today’s industry-standard microprocessors that are well integrated, and well tested, provide more value over time than anything that is custom-built.
There has been considerable coverage in the press of late about changes taking place among the major IT vendors. Cisco decided to dive deeper into the data center and unveiled its new server product, what it called ‘Project California.’ According to one analyst interviewed by the electronic publication MarketWatch (March 13, 09) “if your customer needs something complex and integrated, then you're best off if you can supply it all yourself as a solution … then that's a competitive advantage." But this really can’t be called news, and shouldn’t come as a surprise. Users have been asking for greater simplification for many years – and Cisco customers would certainly benefit from any plans Cisco has to reduce the “number of moving parts” in the network!
And then we read of rumors about IBM buying Sun - for very little premium, given the cash Sun has in the bank and the profit being generated by the StorageTek unit. And if this acquisition goes ahead, it will catapult IBM into the dominant position in the open system space – but will IBM customers relish fewer vendors in the open systems space, and what of its mainframe offerings? While Cisco may be addressing technology complexity, IBM looks to be doing nothing for the traditional mainframe user.
Just as I consider it highly likely that the NonStop Advanced Architecture (NSAA) products from HP will evolve to become a “bespoke” solution, could the IBM mainframe too become a bespoke solution targeting the needs of a very select group of clients? If it does, would the mainframe become further isolated and outside of IBM’s mainstream R&D and unable to leverage IBM’s much better funded open systems efforts?
The Tandem Buildings at this end of Vallco Parkway witnessed the emergence of a culture we associate today with many Silicon Valley high-tech companies. Dress-down Friday’s with its afternoon beer busts started at the back of these buildings, and it became the place to watch “First Friday” TV broadcasts. So many ended up in the pool as night wore on that it became in icon for all that it meant to be a Tandemite.
But the pool’s gone, buried underneath a grassy lawn, as has the need for pricey custom hardware. And for the user community – is this necessarily a bad thing? Bringing NonStop into the mainstream (of HP’s product roadmaps), and riding commodity technology lifecycles, certainly brings with it better price / performance metrics and if what we see right now is any indication as to where this is all headed, then no, it’s not a bad thing.
Above, and to the right I have included a picture of the IBM z10 package and a little lower and to the left, a picture of the HP BladeSystem package that includes NonStop. Check out how the z10 now comes with two stand-alone Linux-based hardware consoles. It’s getting harder to tell the difference – HP uses ServerNet (nicely integrated, mind you) for NonStop interconnects, whereas IBM uses Infiniband (cables clearly visible on this pre-production model, hanging over the filters), and on which the "handprint of ServerNet" can be found, for mainframe interconnects.
But whereas HP now uses industry-standard blades packages across all its product lines – including NonStop, IBM shies away from blades for its mainframes. All other product offerings from IBM have a blades option, but not the mainframe. IBM’s BladeCenter? HP’s BladeSystem? Very similar – and really, IBM’s mainframes today bare more resemblance to other manufacturers’ large servers than they do to the IBM mainframes of a few years back. Why wouldn’t IBM pursue a blades option? Why no mainframe blades?
Of course, IBM does use something similar – a technology it calls “Books” but to everyone else these look just like fat blades. When it announced the zSeries 990’s in May 2003, IBM packaged a single thermal conduction module (TCM) into each Book and supported up to four Books per mainframe. This packaging continued, with the announcement of the z10 in March 2008, with IBM upgrading the chips in the Book and this time (unlike what we saw in the previous zSeries 990’s), its dual-core chips all seemed to work.
Blades, with their cutting-edge approach to standardization and commoditization, could easily scythe through IBM’s mainframe marketplace. Not so much delivering a single fatal cut, but with thousands of slices, as one after the other, mainframe applications are “modernized” and moved onto less complex platforms. Today’s multi-tier architectures, anchored by the data base and front-ended by web and application servers in support of the Internet and other web-based clients, has not helped the mainframe cause.
Some users have begun to offload the data base and this will prove interesting to watch. After all, the mainframe is still a great place to keep the data base! But moving it off the mainframe may only accelerate its demise. As Tim Rathbun, Executive VP of Goldengate reminded me this week “weaken the necessity to anchor the data base on a mainframe, and you quickly remove the need to have a mainframe at all!”
I remain uncomfortable with the concept of Books however, and have commented about them in previous blog postings. Check out the posting of January 21, 2008 “HP and IBM? Moving in opposite directions ...” For me, books belong in a library, and I have to question whether IBM’s Books can compete with blades long term – or will there be mayhem in the “library” as IBM’s hold on the high-end marketplace gets slashed? It’s unclear to me if IBM can remain innovative with its mainframe relying on this packaging, and if the mainframe should go away this time, “they ain’t coming back!”
HP has its own issues when it comes to innovation. Many within IT may contend that HP has a lot of ground to make up to catch IBM as far as innovation goes. But looking at the battles that will be fought in the coming years – IBM will have to let go of its current mainframe packaging and try to catch up with the rest of the industry. Pinning its hopes on proprietary Books is a loosing cause for IBM, and can only stall further innovation for the mainframe.
As difficult as it may be, IBM really does have to fill in its own swimming pool. Not possible? Too much tradition! All well and good for IBM to shout, and remain a contrarian, but competitor’s blades may land too many cuts from which the mainframe cannot recover. And cutting deeply into the Books as one application after the other is modernized, may just precipitate such an end.
Leave the Books in the library and sharpen their blades messages! After all, IBM has never shied away from a fight before over their ability to innovate – and I just can’t see them walking away from a fight like this!