Monday, September 22, 2014

Those Special Occasions!

I’m posting this on the road as I drive to MATUG in Philadelphia, PA. However, before we packed the bags and began the drive, we had an opportunity to celebrate with friends and when it comes to NonStop, I was reminded of earlier posts and yes, I do get it!

Had a good friends, Brian and Jan, over for the weekend; a mix of business and good meals with fine wine. In this case the friend, Brian, is someone with whom I can share my business plans as well as my interests in cars – yes, when asked where he was by a former neighbor, Brian simply responded “I’m with the Buckles, driving Corvettes and drinking good red wine”. As it turned out, he was here for his birthday as well, and on a special occasion like this, it called for something equally as special from the wine cellar.

As I pulled the cork from a 1993 Silver Oak – a great California Cabernet Sauvignon – I have to admit, I was apprehensive. I was real careful removing the cork as it had dried, and once I had it out of the bottle I decanted the wine, just to be sure. Left to breathe for an hour, perhaps longer, it proved to be everything we were hoping it would be – not a single rough edge on this drop of red as we tentatively took a first sip. It was suggested that the wine could be kept 15+ years but I was really surprised as we had laid it down for 20+ years and it tasted a lot better than when first we opened a bottle, back in the mid-1990s.

I couldn’t help noticing that in a post of nearly two years ago, back on December 26, 2012, I had featured another good bottle of wine that I had opened. The occasion that time had been the Christmas holidays, and again it had been a good bottle of Cabernet Sauvignon - 1994 South Australian bottle from Wynns.
The post, Yes, I get it!

In addition to wine review featured comments from a race car driver who had observed how the team had tried to “Control costs, restrict options, standardize on certain parts, increase production, amortize investment over a wider base, increase value for money” and at the time, I suggested we could say much the same thing about NonStop.

Continued investment in NonStop is a very delicate balancing act – there’s a need for new features even as there’s a need to provide an integrated hardware and software product for a fair and reasonable price. The viability of NonStop is still very much tied to it being cost effective when viewed against the competition, even if we all agree that NonStop has very little by way of competition. It was analysts at IDC who came up with an “availability spectrum”, categorizing all systems as either Availability Level 1, 2, 3, or 4, with AL1 referring to systems not shipped as highly available and AL4 referring to fault-tolerant servers.

IDC describes AL4 as being “the highest-availability level, connoting that the end-user experiences no perceived interruption based on the use of fault tolerant servers. In this level, the combination of multiple hardware and software components allows a near-instantaneous failover to alternate hardware / software resources so that business processing continues as before without interruption.” Without delving into the finer points of takeover versus failover, I think readers will appreciate the simplicity of this definition, especially as it comes at the “problem” from an end-user experience perspective.

Control costs! Standardize on certain parts! Increase value for money! It’s hard to argue against the motivation or drive behind these goals. For readers who have regularly attended user events anywhere in the world where NonStop product management has been present these goals should be well-known by now. I can’t recall a presentation by a manager or executive who hasn’t pointed to a roadmap slide and hammered home the continuing pursuit by all within the HP NonStop team to increase the value for money.

The argument in favor of going with MIPS was all based around the unsustainable business model of continuing with custom chips. The subsequent move to Itanium followed a similar line of reasoning, this time the thought being that Intel had far deeper R&D pockets than MIPS (then a part of SGI between 1992 and 1998). However, even as Itanium continues to underpin modern NonStop servers, increasing value for money over time meant that an even more popular chip would be required – something that was not just a standard but deployed broadly enough so as to “amortize investment over a wider base”.

The move to support the Intel x86 architecture certainly is a positive move in this direction, and having a portfolio of products all utilizing x86, economies of scale will certainly be present. But there’s more to this story and as I wrote two years ago (long before NonStop announced plans to support x86), I do get it!

It’s not that NonStop development is hedging on either performance goals or pricing, as I am sure we will hear a lot more about all of this at the NonStop Bootcamp in November, but I have to believe the x86 will result in performance improvement for some solutions, if not all, even as I have to believe the price will be more attractive. When you look at Intel’s roadmap for the x86 you quickly realize that the x86 is a substantial family of chips – Xeon comes in many flavors, if you like, with some Xeon chips focused on client-side processing whereas others are focused on the server side.

Good performance, value pricing, and of course, a whole new fabric interconnect that holds the promise of even faster processing speeds combined with the possibility of building hybrid computers made up of different systems (from an OS perspective) and again, all based on standards. Certain parts just have to be standardized these days – the chips, the memory, the interconnect fabric and NonStop is certainly delivering on this promise. As middleware and solutions vendors come to terms with the new technology, the prospect for even greater optimization is apparent and even though there may be changes coming, I get that, too!

When x86 rolls out, there’s no doubting the special occasion it will represent. In some quarters, it will warrant a bottle or two of champagne being opened. And from my perspective, a well-deserved celebration, for sure! However, when it comes to controlling costs, this is where the NonStop users will share an equal burden. There’s much that they can do to control costs and much of that has to do with the cost of human resources – in achieving AL4, NonStop development has provided an integrated hardware and software “stack” and the higher up the stack NonStop users elect to go, the more they can control the costs.

This too includes those vendors who are now in the midst of porting new applications to NonStop. If existing NonStop users or new users to NonStop elect to skip the middleware available today, they make leveraging the benefits of an AL4 system difficult. It may be viewed as clever, or even ingenious, to dump a new product on NonStop with minimal touch points to the stack NonStop development provides, but I am sure five or ten years out, with the development lead no longer present, it will look little different to instances of the same solution on other systems. If you are going to run it on NonStop then take advantage of NonStop top-to-bottom, otherwise little by way of controlling costs will materialize. System savings will pale in comparison to the human costs involved.

Increase value for money! Adding support for the Intal x86 is going to be a big help in this department. Standardize on certain parts! Embracing an interconnect technology as NonStop development proposed via InfinBand (IB) as well as manufacturing producing an universal blade as a result is also going to be much appreciated. Control costs! Leveraging NonStop development expertise, fully utilizing the integrated stack and letting the OS ride the changes well is our cue for NonStop users to do the smart thing – it’s almost as if we are learning to love Pathway all over again!

The bottom line here, of course, is that after four decades (or thereabouts) we are still talking about NonStop and still thinking up new ways to exploit the best implementation of AL4 available today, Sure, IBM mainframes today now have what IBM is calling Geographically Dispersed Parallel Sysplex (GDPS) that is an extension of Parallel Sysplex in order to support mainframes located, potentially, in different cities. But this is at what cost? And in the end, with what added value?

Complexity abounds with this model of IBMs and yet, it’s so simple for NonStop users. Can we execute a takeover where systems are separated? NonStop users familiar with the recent additions to Pathway (TS/MP) 2.4 and later, understand how easy it is to deploy NonStop in this manner with far less complexity and at a lot lower price point – TCO remains important as do the number of nines, and even though IDC gives the IBM system z mainframe running in Parallel Sysplex mode an equivalent AL4 status to NonStop, it’s at such a high price point that no, I just don’t get it!

I still have some good wine left in the cellar and I am sure that my friends, Brian and Jan, will find a way back for more. There will be more pictures taken of empty bottles, too, to be featured in stories to come I am certain of that. However, when the new NonStop systems begin to ship I too will make sure there’s a little champagne on the side that I will be only too happy to open and to toast the NonStop development team for a job that is sure to be well done. NonStop on x86? NonStop in IB? NonStop in hybrids and even Converged Systems well, what else can I say but that I get it, too!

No comments: