Thursday, July 22, 2010

Investing in NonStop continues ...

I spent this weekend in the Bay area and visited with my friends in Cupertino. I passed buildings on Pruneridge Avenue, Fremont Avenue, Stevens Creek: all familiar addresses to anyone who has spent time in Cupertino and Sunnyvale and who has had history with Tandem and now, NonStop! Of course, there’s still the Duke of Edinburgh, a pub that’s already been written about in earlier posts and where, if the walls could talk, so many details of much that has been implemented on NonStop could be heard! A lunch in this pub is always hard to pass up!

A few weeks ago I had the opportunity to spend a lot more time on the HP campus on Pruneridge Avenue. On that occasion I had the chance to tour a temporary data center where a team of folks under the watchful eye of John Donelan, another Australian mate of mine, had assembled a very a large NonStop server. The server was destined for the new Austin data center, and was a part of the data center convergence program that Randy Mott, HP’s CIO, has been pursuing. I came face-to-face with a massive Neoview 512p sever. The picture above is of me alongside one of the rows that housed 128 NonStop processors.

The scale of the system was impressive. Four rows of cabinets housed the 512 processors while some 5,000 plus 600Gb disk drives were to provide 1.5 plus petabytes of usable data storage. There was an element of symmetry about the configuration that I particularly liked – each row was made up of eleven cabinets with a central cabinet housing essentially the engineering console, flanked either side by four cabinets, each with eight boards housing 16 processors and enough CLIMs to support just the population of disks deemed optimal for Neoview’s Enterprise Data Warehouse (EDW), that was central to the Business Intelligence (BI) applications it would support. On the outer extremities of each row were separate cabinets populated with additional processors, CLIMs and disks. Forty four cabinets in all and, if I was mistaken, the temporary data center selected was in Building 44!

When the system was shipped to Austin shortly after I was given the tour through its rows, it would go through additional testing before being loaded with data and then made available to any ODBC client via HP’s global network. Hard to grasp that it would become just another network-accessible resource. Good news for all those interested in all things NonStop is that a second Neoview 512p system is being assembled in Singapore and that it would eventually find its way to Houston. Whether HP would require a third system of this size for it’s Atlanta data center wasn’t much more than a budget issue and I wouldn’t be surprised to hear shortly that yes, indeed, a third system was on its way. Randy Mott has demonstrated repeatedly that he wants each of the three data centers to mirror the others.

Looking at this system of forty-four cabinets made me realize just how many Intel Itanium processors were involved. Many more than most user installations get to see but, all the same, the developers at Intel must be very pleased in the knowledge that servers of this size were being used. I have no idea why Itanium continues to draw questionable comments in some quarters but in realizing just how much data could be processed reliably, and immediately, put aside any doubts I may still have had over this particular chip architecture. There have been NonStop systems as big as this shipped to other customers, I have been told, but I suspect they are running in sites where I couldn’t as easily view them – so seeing this set-up first hand, as briefly as I did, will be something I will remember for quite some time.

For several decades now, applications available for NonStop did it all – from processing transactions to checking tables and files to recording pertinent information on a data base, the HP NonStop Server was involved every step of the way. For some market segments where applications were available on NonStop, the market was extremely kind. According to The Standish Group in the recent paper, Megaplex: 35 Years of Evolution from the TNS1 to the Integrated Data Center, the authors even went so far as to suggest that “in markets where applications do exist such as payments, cell phone registration, and stock trading, the NonStop enjoys an almost-monopoly.”

Modern applications are being architected such that components run on different platforms, each with different price points that reflect the platforms properties whether application suitability, performance, or scalability. At a time when so much discussion is centering on the ability to deploy cheap commodity hardware that can be easily scaled-out, however, data bases are getting bigger and the properties of NonStop better appreciated. According to The Standish Group paper on Megaplex, “the future is much more interesting for NonStop applications. Five percent of a typical application could benefit from a highly reliable environment …putting the 5% of high-value applications in a NonStop environment will give the reliability and availability that may be desired, without forcing the other 95% to run on the more costly platform.”

Flipping through magazines that have piled up in my office I glanced at the August 2010 issue of Harper’s Bazaar. Yes, among the car and motorcycle magazines there were copies of Architecture Digest, The Robb Report, The Economist, and yes, a selection of Fashion magazines. What caught my attention this time was an advertisement for L’Oreal featuring the entertainer Beyonce promoting a new “infallible, never fail, lipcolour” that lasts for 16 hours! Apart from the obvious tautology this represented, it struck me how often the CIOs and data center managers that I have talked to have spoken earnestly about how they longed for technology that was infallible, never fail, no matter what happens! And for much longer than 16 hours! The NonStop architecture, in its most recent, modern, commodity based iteration is beginning to make sense to many of these CIOs – not for the whole application but for the most critical 5% that they just cannot afford to ever loose. Within this small percentage, it is the data base that is most often talked about!

At this year’s HPTF event in Las Vegas, in presentations by NonStop Enterprise Division (NED) executives, there were references made to new investments in the NonStop SQL/MX data base and that these investments included bringing some of the technology, added in support of the Neoview application, back into SQL/MX. According to Randy Meyer, NED Director of Product Management, “there are certain features and capabilities inside Neoview that are beneficial to NonStop customers. It behooves us to find cost-effective ways to make those accessible to our NonStop customer base in the most efficient way.”

For all of those attending it was hard to escape the message that NonStop was again, becoming very serious about its data base support. NED is investing to find ways to better leverage current Neoview technology advances into the more traditional OLTP / ODS data base environments of today, as well as in ways to optimize it for those future, most-critical, 5% high-value applications that will find their home on NonStop. It was also hard to miss the positioning being done in support of NonStop and its role in hybrid configurations, even stretching as far as its potential future at the very center of private enterprise cloud configurations.

This posting is not about Neoview. Many of us familiar with NonStop still view Neoview as nothing more than a very large application on NonStop. True, today’s Neoview systems, in order to meet the requirements of the BI applications that depend upon it, run a distinctly separate release of the OS where the processors and I/O are configured far differently from what any NonStop customer would use,. What this posting is about however, is the continued investment being made in NonStop. Seeing that the data base is the benefactor of some of this funding sends a very positive message into the marketplace. Nothing could cement the future of NonStop more surely than visible improvements to the NonStop SQL/MX implementation!

As The Standish Group wrapped up its first Megaplex paper, it remarked on how “in a tarot card reading the overturning of the Grim Reaper card usually solicits a conspicuous gasp. If we were doing a tarot reading on the NonStop System, The Standish Group would not be predicting the platform’s death, but instead, its rebirth. The rebirth for NonStop will be the Megaplex. In the Megaplex, NonStop will become the heart of the operating environment service. Much of the NonStop technology will be used by all the operating environments within the Megaplex.”

Perhaps pushing a little further than I would have proposed, although some of my posts towards the end of 2007 and into early 2008 that culminating with the post "My Wish" for NS Blades addressed something very similar, The Standish Group is definitely on the right track. It may take a few more years to see all the pieces “fleshed out” but what is visible today gives us more than enough clues as to where this will all possibly lead.

5% of enterprise applications needing NonStop may not sound all that much for some users, but in today’s global marketplace, the math could prove enormous. Let’s all hope that HP truly sees this possible future for NonStop and begins to value NonStop as highly as those of us aware of where this is all going have already begun to do!

No comments: