The photo above is of me close by the pit exit at Willow Springs International Raceway (WSIR), that up until recently wore the title of the fastest track in the west. The one take-away I have from the time spent out on road courses is that focus is paramount and improvement only comes with practice. Experience developed over the years is what sets racers apart.
And practice, practice, practice – there’s absolutely no substitute for the time spent behind the wheel! The most powerful car rarely clocks the fastest times, nor does the drivers’ age seem to be a factor. At the club level, drivers will often switch cars and “volunteer” to show newcomers the best lines and you can always recognize when any car is in the hands of a more experienced driver.
The origins of this post can be traced back to a post I wrote three years ago to this weekend, and posted on February 12, 2008 “’My Wish’ for NS Blades” and then followed-up more recently in a number of articles I have written over the past couple of months including a feature in the October issue of the e/Newsletter Tandemworld.Net as well as an earlier post to this blog on November 30, 2010 “Nothing seems to last …”
On the understanding that NonStop would be supported as part of the BladeSystem offerings, I suggested three years ago that firstly, in terms of wishes, mind you, HP BCS delivers on the slideware Martin Fink, Senior Vice President and General Manager, Business Critical Systems (BCS), unveiled as the “Shared Infrastructure Blades.”
With Blades a reality, my second wish was for the NonStop operating system, NSK, to be customer-configurable, perhaps even supported atop a hypervisor, such that customers could easily reconfigure the BladeSystem to support as many or as few NonStop processors as their applications required.
My third wish then pushed the barrier even further. Assuming an operator could reconfigure a BladeSystem to have as many or as few NonStop processors running as the applications required, then it would be easy for HP to provide an API, such that a workload manager could automatically reconfigure processors on the fly as transaction mixes determined.
Three years later, so how did I do? Obviously, the BladeSystem was announced and it supported NonStop, and today these have proven to be very popular systems. The drive to commoditization will only see the Blades become more universal in the months ahead so I suspect my first wish is close to fulfillment. The Engineering Prototype exhibited at a recent HPTF event certainly highlighted the potential from a shared infrastructure blades platform.
With respect to my earlier views on virtualization then no, they no longer make sense in my opinion – shifting NonStop further from the hardware lessens its ability to be NonStop!
As for the third wish, and perhaps even for all three, it’s going to depend on how solutions providers weigh the value of a hybrid platform and whether there’s a competitive edge for them as a result. In other words, new entrants into an industry vertical dominated by one or two vendors may not just step outside the box, they may smash it to pieces.
While one solution mandates a BladeSystem chock full of the same high-end Blades processors, a new solution may call for only a couple of high-end Blades processors and then complements it with cheaper systems – mixing Linux and even Windows with NonStop, for instance – then the price difference could be material.
Transactions will never uniformly represent the same value to any user – some transactions will be extremely valuable whereas others will be simply nice-to-have. The fiftieth time I check my 401K balance surely drive my services company nuts if every time it’s processed it’s been on the same system that handles billion dollar trades!
Between my second and third wish lies a really significant piece of work that I have covered several times, and yet has been getting a more encouraging response. Before you can manage, you first have to capture and report, I have been wisely counseled. This makes a lot of sense to me so I now anticipate the first forays into this area will be simple dashboard like solutions telling me I have 3 transactions of this type, 20 of that, and another 6 the other type. Processing will continue on a homogeneous Blade platform, but at least we will have mechanisms in place screening the workload.
If in my general enthusiasm I have not explained myself well, then what I would like to see is a chassis populated with commodity Blades that are physically identical and that can run any operating system, including NonStop, to provide a Hybrid platform and that the number of such Blades running any instances of an operating system can remain fluid – the number assigned in support of NonStop growing and ebbing solely on the mix of transactions arriving that really must be processed no matter what!
Furthermore, while the basics are in place, or are about to be, I no longer see HP driving the release of such packaged Hybrid platforms – it will be successful solutions vendors grabbing market share from the incumbents that will fuel such a possibility! It will be successful solutions vendors that attract infrastructure vendors to resolve the workload monitoring and management that could drive its eventual development.
However, it is now three years later and as I look back, I grade myself a B+ on the understanding that blades can populate a single BladeSystem and that it’s just a case of solutions vendors filling in some of the missing elements. But what do I wish for these days? What are my latest three wishes?
Expressed late last year as observations, let me start out by quoting Martin Fink, Senior Vice President and General Manager, Business Critical Systems (BCS), when he proposed in a recent email exchange “with every new microprocessor that becomes available to us, we continue to evaluate ServerNet and its impact on overall system performance. Certainly at some future date, we'll move to a standard interconnect – probably Infiniband (IB) as a possible alternative.”
Fink then added, “that's precisely in keeping with our stated strategy to have NonStop built from the same common, modular infrastructure as all the rest of HP's servers.” That will support a commoditized blade package, without a doubt and while there’s still no guarantee we will see every blade package with integrated IB support, a standard interconnect technology for all blades packages seems a certainty.
If that remains my first wish, I am now more than satisfied that one outcome of the commoditization could be NonStop shipping within every platform. This would be my second wish. Again, there are capabilities that could greatly benefit from the presence of NonStop. In much the same way as when IBM moved to “parallel sysplex” and required an instance of MVS to run “clocks, locks, and lists” HP could achieve something far better through the presence of NonStop in every Hybrid blade platform.
As for my third wish this time around, then it’s for greater cooperation with solutions and middleware vendors. We have already seen some partnership success in areas to do with security and manageability, but I believe there will be even more action on this front – the magnitude of what’s presented today in the software roadmaps really does call out for help from others to ensure implementation in a timely manner.
It may be a different set of circumstances from what a racer may find, but when it comes to my most recent three wishes, more than 35 years on, HP has the experience to master just about all I have covered. And for my part, it certainly would instill confidence across the community should these wishes eventuate any time soon!