In my last post, When windows open, only the swift make it through, I wrote of how I would be developing a new presentation for an upcoming webinar IR was hosting: NonStopX- The New Goal. It was to be the third webinar (in a series of three) and I was pleased with the audience, as I was with the quality of questions asked. It's always good for those giving a webinar to know their audience is fully engaged.
All three webinars can now be viewed on demand. If you found the final webinar of interest then you may like to read the latest post to the NonStop community blog, Real Time View, Turning up the heat! While the topics covered under the general heading of vision, strategy and goals one focus area generated more questions than any other and that was the topic of NonStop running in a virtual environment.
For as long as I can recall, inserting a hypervisor between NonStop and the hardware, running NonStop inside a virtual machine, was an anathema to all die-hard NonStop system managers and yet, here we are. NonStop, up and running in a virtual machine, after all! NonStop, as a platform, becoming a guest of a hypervisor. The ramifications for application performance monitoring solutions, including Prognosis, are significant. Historically, being able to run an OS as the guest of a virtual machine allowed customers to run multiple instances of their platform. Given the propensity of enterprises to continue doing so, Prognosis will be required to keep track of many instances of NonStop running on a single physical machine.
In his December, 2015, keynote presentation at HP Discover London Martin Fink, EVP & CTO, Hewlett Packard Enterprise (HPE), talked of issues of scale when he talked about data and the problems that come with virtual machines. "There's more data and with more data there are more apps and this winds up meaning more infrastructure to where you need more people," said Fink. "How many IT devices can one person manage?" Not surprisingly, Fink laid the problem squarely at the door of virtual machines adding that, the concept of a "virtual machine is a highly inefficient" approach (for optimizing resource sharing) and among its drawbacks is that virtual machines "prevent you from taking advantage of all of your infrastructure."
So what is propelling HPE NonStop development to go down this path? What's the value proposition for NonStop running in a virtual machine? Isn't there fundamentally an escalation of risk associated with NonStop no longer being tied directly to the hardware? At issue here is that within some market verticals the decision has already been made – run virtually, or not at all. And for NonStop, one such vertical happens to be among the biggest users of NonStop; the Telco and mobile phone communications marketplace.
While Fink didn't call out NonStop as participating in an entirely new for the industry, what Fink called the ContainerOS, he had already acknowledged that work was under way within NonStop development labs in support of NonStop running in a virtual machine. In his November, 2015, keynote presentation at the NonStop Technical Boot Camp San Jose, Fink talked of NonStop, "Running in a virtual (environment) on Linux." Fink then told the audience, "As an important proof point, we can absolutely get there" confirming for all in attendance that this was no longer theoretical.
However, this wasn't the end of the story as Fink then asked, rather provocatively, "Wouldn't it be cool to bring the NonStop value proposition to Linux and bring to market (more) powerful hybrids – a powerful combination." In these comments I was to later find out, the reference to Linux was to a Linux that had been enhanced to run more or less as a hypervisor – Kernel-based Virtual Machine, or KVM. This is a full virtualization solution for Linux on x86 hardware and KVM is an open source software project that fits in nicely with HPE's clearly stated goals of staying focused on Linux and open source.
This doesn't mean that NonStop too is becoming open source but rather, where appropriate, some components of NonStop that, broadly speaking, enhance availability and potentially scalability, may find their way into the open source community. Long term, this is clearly one of the goals of HPE for NonStop. It's hard to argue against Fink when he says, "Wouldn't it be cool to bring the NonStop value proposition to Linux" should this eventuate. This should give vendors who are well-versed in all things NonStop an immediate advantage as products materialize.
The drive to virtualization highlights the coming intersection between NonStop and two major monitoring trends – the drive to greater automation together with the need for single pane of glass visualization. As Fink noted, "How many IT devices can one person manage?" isn't so much a lament over future restrictions as so it is a rallying call for smarter technology. As Prognosis continues its upward trajectory moving beyond simply displaying alarms and alerts to provide prescriptive and self-healing capabilities, one person will be able to manage so much more than they do today.
This is an important consideration for any enterprise thinking of upgrades to the monitoring solutions. Just this week I found a reference to the blog post, Big Data is Dead. All Aboard the AI Hype Train! The post was based on the premise that there's a limit to just how much technology can be used to help extract meaningful data from data warehouses, stores, lakes and streams. "But for now and the foreseeable future, the best way to attack your business problems is still done the old fashioned way: creative, smart, and curious people who can ask the right questions and know how to get them answered. Big, dumb algorithms and warehouses of data are useless without them."
The properties we will count on, as being crucial for our monitoring solutions, will continue to focus on reducing the numbers of smart folks needed to run an enterprise's information systems. However, there is a limit, and no substitute for creative, smart, and curious people. Introducing virtual machines into the equation will likely compound the problem – adding something entirely new as we see what's coming with ContainerOS compounds the problem further. It will be impossible to restrain the tide from coming up the beach and virtualization inroads into platforms particularly as enterprises transform to hybrid infrastructure means NonStop will not be excluded. Visualize this? Fortunately, with Prognosis I believe we will be able to see it all!