Virtualization, and the support of virtual operating systems, has been a part of IT for a very long time. The concepts are pretty simple; I/O has been much slower than main memory so actions taken by the processor have happened more quickly than actions directed at storage. This resulted in a lot of processor wait time and processor time was viewed as expensive; what could we do while we waited? Initially this lead to the creation of a number of partitions all sharing virtual storage (VS) so much so that it wasn’t unusual to find large systems running five, seven, twelve and more partitions that, in the case of IBM mainframes, broke processing down into online and batch.
But even so, as systems became even more powerful there was still an abundance of processing power left underutilized. Operating systems supporting these multi-partition configurations became the norm from the early 1970s onwards and in time the amount of work, or processing, that these operating systems could perform seemed limitless. On the other hand the overhead of paging came at a cost. Or, as one IBM architect told me, if you want to run virtual operating systems and have it deliver the performance you need, make sure you give it access to lots of real memory. Yes, real memory certainly helped the virtual world.
Support for virtual operating systems are well known and comprehensively documented. NonStop has run as a virtual operating system for as long as I have been associated with NonStop and has supported some of the biggest and most important applications in the world. In fact, the broad acceptance of the concept of virtual spawned many innovations as very quickly it was realized that separating what was executing from any dependence on the hardware liberated the way we thought about programming. Before VS systems arrived, there were so many constraints over how large a program could be even as there were real issues over how long we could keep the on-line system up and running if we had large batch jobs scheduled.
Of course, this all harkens back to the primitive times of early data processing. A time before we even thought of transitioning from being a part of electronic data processing to where we supported management information systems. Again, this is all well known to many within the NonStop community but its implications were far reaching. If we were able to do everything virtually utilizing just one system – a large mainframe comes to mind – who could argue over the need to run a highly intelligent front-end as the earliest NonStop systems were called. Virtual Storage was joined by Virtual Networks such that flexibility of paths between end-users and data were almost limitless. Even at 9,600 and 19,200 bps.
However, these were primitive times and the world has moved on. Given how storage had been virtualized and with it, the main memory processors accessed, and how networks became virtual, it only needed one small step to be taken before we came to virtual machines. Conceptually, considering everything that goes into supporting our programs as being virtual right down to the computer, or machine, itself the prospect of running multiple virtual machines on just one machine became a reality. Processor speeds continued to trump I/O and up until recently, that difference in speeds only widened encouraging even greater prospects of not just running lots of machines but an array of different virtual machines that otherwise would have required their own separate computers or machines.
In the post of January 29, 2020, to the HPE Community blog, Virtual: For NonStop it’s a reality! HPE MCS Marketing Manager, Vikas Kapoor, said, “Virtualization is real. In a world where we have grown accustomed to virtual meetings and virtual tours and virtual wallets along with virtual cards, we have seen the progression from physical systems to virtual machines take hold in ways that mirror what we accept in the real world.”
More importantly, according to Kapoor, “When running Virtualized NonStop, anything associating with the backplane and the fabric is replaced by the hypervisor. In the case of VMware, IT is able to configure multiple virtual machines that include not just NonStop processors but also the all-important I/O servers supporting storage and communications, otherwise known as CLIMs. Should there ever be a failure of a hypervisor when running Virtualized NonStop, the system treats this as no different to a failure of a physical server with converged NonStop.”
Also Kapoor quotes Karen Copeland, Manager, HPE NonStop Product Management. “It became clear that our migration to an INTEL x86 architecture gave us an easier way to deliver fault tolerance to a broader market and with other operational methods. Of course we used input from our customers to guide us toward supporting Virtualized NonStop on VMware.” In the virtual world, the prospect of leveraging the inherent scalability of NonStop means systems can become as large as our imagination makes possible. In the new world of edge to core where clouds and even mini-clouds prevail, NonStop can now be anywhere and everywhere. As such, this represents perhaps the biggest step forward in the modern world of hybrid IT that NonStop has ever taken.
“On a clear day, you can see forever,” wrote the lyricist for the much loved musical of the mid-1960s. While there are limits to how far we can take this image when it comes to what NonStop can achieve, nevertheless virtual NonStop can provide true fault-tolerance out-of-the-box no matter what is underneath supporting the virtual machines hypervisor. With VMware already supported, a single physical x86 cluster could support production images of NonStop together with testing images where the latest release of the NonStop operating system is being validated / certified for production usage while all the time, multiple images of NonStop could be accessed by development teams.
Where the arrival of virtual NonStop steers today’s conversation is towards cloud deployments. Whenever HPE writes about the cloud experience for the most part HPE is highlighting the benefits of turning a combination of traditional systems and private clouds into a hybrid IT environment that takes on the appearance of a single unified cloud. Whether NonStop is running on the traditional systems or virtually, as a VMware supported virtual machine, or both, is not apparent to the end user. With virtual NonStop, NonStop is just one more resource being supported out of the IT organization.
We may not be able to see forever but what we can clearly see is a totally transformed way at looking at how best NonStop can serve the enterprise. Untethered from the presence of physical machines, virtual NonStop can be deployed in ways that best meet the needs of those enterprises whether configured as one virtual NonStop or as multiple virtual NonStop systems located throughout a global network. See what I mean? The opportunities, unlike what we can see, appear almost limitless! The only real question remaining for the NonStop community comes down to just one item – where will your virtual NonStop contribute best for your enterprise?