"My Wish" for NS Blades

This weekend I drove out to the Southern Californian coastline, and it was a real change from earlier weekends spent in Colorado. Sure it’s winter across the US but out along the coast, it was balmy days in the mid 70s. I was driving the roadster, with the top pulled back, and the picture here is of me beside the Pacific Coast Highway –or PCH as it is known locally - just north of Malibu. The drive reminded me of those articles you find in the airline’s magazines – three perfect days in “wherever” - and I had days about as perfect as you could experience anytime in LA.

During the previous week, I was emailing folks in the UK as I was preparing to catch up with them next month. As these emails had been flying around, I made a minor mistake in the subject line – rather than proposing to meet for “drinks”, I was suggesting we meet for “rinks”! Of course this led to Sean Bicknell of XYPRO, and a BITUG committee member, to respond “I’ll bring my ice skates then shall I!?!?” Driving the PCH, enjoying the sun and thinking how freezing it was back in Colorado, led me to recall the lines of the song “All Star”, from the first Shrek movie, by Smashmouth: “The ice we skate is getting pretty thin, the waters getting warm so you might as well swim.”

Talking about skating on thin ice, I have been taking a really close look at HPs press releases, and have pulled down a number of Roadmap presentations from the ITUG web site. I have also been deep into other HP web site pages. I really wanted to get a better understanding of HP’s plans for NonStop. I am putting together a new presentation to use at ITUG events, and it all started as I recalled a comment made at last years’ HPTF&E by Martin Fink.

Martin was talking about how pleased he was that NonStop was being openly accepted by BCS development. He related that, as the blades developers working on the new cClass bladeserver chassis began to leverage the IP of the NonStop group, they asked Martin “Can we call it the NonStop Backplane”? Martin told the audience his response was “why not – go right ahead”! And today, if you go to the web and pull down the pdf of the HP BladeSystem c-Class architecture “technology brief”, you will find on page 8 a paragraph titled “NonStop signal midplane provides flexibility”.

As we continue to hear more about the first bladed architecture products, something a lot of us simply call NS Blades, it too will be using this chassis. When I asked a well-placed NonStop engineering manager whether this was the case he responded “the NS Blades system does indeed use the new c-Class platform (BladeSystem c7000) enclosure”, adding enthusiastically “everything to blades”!

Predicting the future at HP, particularly when it comes to the future of NonStop, is a risky business and I have spent my time out skating where the ice is thin. I haven’t always been right but I have been close most of the time. But rather than talking about what will be coming from product management, it may be more interesting to lay out my own wishes for the NonStop family.

I began to lay out some of my thoughts in earlier blogs and late in January (The check-in desk two-step!) I wrote about Mission Critical applications and of their long association with the NonStop platform. I went on to add in the next posting in January (NeoView; a new view?) how transaction processing will always have mission critical elements and that the support of these elements will always benefit from the NonStop architecture. And then in an early February posting (Disruptive technologies and radical innovation) I reported that NS Blades have the potential of becoming a classic disruptive technology in that its support for any OS, including NonStop, means that your choice of OS can be arbitrary and driven solely by the demands of the application.

I have three items on my wish list and all involve the NS Blades. I am working with the belief that HP will begin shipping NS Blades sometime mid-to-late summer. I am also confident that these will utilize the c-Class blade enclosure (BladeSystem c7000). And so my first wish is to see HP BCS deliver on the slideware Martin Fink first unveiled as the “Shared Infrastructure Blades” package. This is where any mix of NonStop, HP-UX, Linux, and Windows Server OS’s will be supported.

The importance of this feature – and getting the manageability all sorted out – is that customers will be able to deploy any mix of web, application, and data base servers in a single enclosure. NS Blades customers will be able to add additional blades as growth in any one area dictates. Indeed, I am expecting that they should be able to reconfigure and “redeploy” under-utilized blades to other OSs as workload mixes change over time. The manageability component is very important – you only want a single command interface to all the blades. A separate console, per OS instance, kind of misses the point. And I am wishing for a programmatic interface into all of this as well.

My second wish is to see a hypervisor introduced where NonStop can be configured as a “guest OS” in much the same way z/VM is used on the IBM mainframe. The trick here is to see this introduced without marginalizing the traditional association between NonStop and the hardware with respect to being fault tolerant. In other words a processor, within a guest NonStop configuration, needs to know when to initiate a take-over. Perhaps exposure to the hardware may not be right answer these days – as so much self-checking and correction goes on at the chip level – but NonStop users will not want to see a hypervisor becoming a potential “single point of failure” either. While it may not be as simple as configuring a number of hypervisor processors under the management of a check-pointed “hyper-mon” process, my own logic wants to take me down this path.

If you check out hypervisors on wikipedia, you will see references being made to Type 1 (native, or bare-metal) and Type 2 (hosted) hypervisors. I am a lot more interested in the Type 1 implementations, and while some OSs may want to run the hypervisor as a guest of the OS (e.g. Microsoft Virtual Server – a virtual machine on the Windows Server 2003 operating system), traditional NonStop users may view the properties of Pathway providing something similar!

But it is my last wish where I really want to go out on a limb. If you assume Martin is successful and a shared infrastructure blades package becomes available with a native, or bare-metal, hypervisor (NonStop as a guest with no loss of its NonStop attributes), then wouldn’t it be advantageous to users if interrogation of the incoming transactions would direct mission critical transactions to NonStop, important informational but not quite mission-critical to a Unix or Linux, and voluminous inquiries to Windows? A variation on today’s workload balancing products, but supporting a transaction profiling capability that once set up, learns about the overall mix of transactions, and automatically adjusts the OS configurations on the fly? No longer would you agonize over the ratio of processors assigned to any of NonStop, Linux, or whatever – the system would learn enough over time and adjust the composition accordingly!

Returning from the drive along the PCH, the song “My Wish” by Rascal Flats began to play and as I listened to the verse: “And if you're faced with a choice, and you have to choose, I hope you choose the one that means the most to you. And if one door opens to another door closed, I hope you keep on walkin' till you find the window”. Followed by the chorus “Your dreams stay big, and your worries stay small”. I just knew they were singing about NS Blades!

With the flexibility inherent with the design of NS Blades, and the doors this will open for us all, I can’t see placing any limits on how big our dreams will be! Or, as Smashmouth sang, “My world's on fire how about yours, that's the way I like it and I never get bored”!