Getting Hyper And Converged With IBM i
May 14, 2018 Timothy Prickett Morgan
The hallmark of the System/38 and its progeny, the AS/400, iSeries, System i, and IBM i platforms, is that these machines came fully integrated with all of the operating system, database, management, and development tools necessary to run a modern business. Integrated did not mean that these pieces were all sold as a single bundle, mind you, but they snapped together with good fit and finish and allowed companies to not have to become masters of the system code and could therefore be craftsman for the application code that actually ran the business.
The AS/400 really set the pace for this, bypassing the DEC VAX and Hewlett Packard 3000 competitors in the minicomputer space while at the same time offering a converged System/36 and System/38 platform for these two very distinct sets of customers. IBM got to kill two birds with one stone, and at its peak in the late 1990s, there were 8,000 ISVs peddling over 20,000 distinct applications on the box, which had a very health ecosystem of about $15 billion a year.
The integrated platform goes by a different name these days. For the past decade, at the hardware level an integrated system was called a converged system, and the idea was to bring together compute and networking, of call things, into a single and usually a blade server form factor, with both compute and networking virtualized and managed from a single control plane. Cisco Systems, the leader in datacenter networking, got the ball rolling here with its Unified Computing System in March 2009, but arguably HPE was working on similar ideas with the VirtualConnect for its Itanium-based blade servers even earlier. Whatever the case, everyone reacted to Cisco, not HPE, and before you knew it we had IBM coming out with FlexSystem converged hardware and PureSystems converged hardware with application specific tunings to make it easier to build and deploy applications. This Flex/Pure line never really took off from a commercial standpoint for IBM – it did as well as anyone else, mind you – and it was sold off as part of the Lenovo acquisition of the System x business in 2014.
At about that time a different kind of integrated system started to take off, particularly at midrange and large enterprise customers who, decades earlier, might have been typical AS/400 shops for their transaction processing workloads. This next kind of converged was called hyperconverged, and with this, a cluster of machines running a server virtualization hypervisor are networked together into a seamless whole for running virtual machines and at the same time a clustered, virtual storage area network file system is added to the compute cluster for those VMs to feed upon. It is a server-storage half-blood, and once again, HPE was the innovator of sorts here through tis acquisition of LeftHand Networks, which had a virtual storage array that ran on normal X86 servers, but Nutanix really took the idea another step and created what is now called the Acropolis platform. Last July, I told you all about the Converged Systems line, and specifically, the CS821 and CS822 systems that marry the Nutanix Acropolis platform to Power8 machines for created hyperconverged platforms, and in February of this year, I also told you that IBM has follow-ons, based on the Power9 processors, in the works to support the Nutanix stack. I am not going to go through it all again, but I want to remind you that IBM has its own OpenKVM variant of the KVM hypervisor (controlled by Red Hat) on Power8 and Power9 machines, and that Nutanix has its own variant of KVM, called the Acropolis Virtualization Hypervisor, as well. Everything I just said is just background for what I really want to talk about, as is the next paragraph that I am extracting from what I said nearly a year ago:
“First of all, if Nutanix can be a virtual SAN for KVM instances running on Linux on Power, there is no reason that it cannot be a kind of hyperconverged Integrated File System for IBM i, much as the actual Integrated File System from OS/400 V3R6 was based on OS/2 Parallel File System that was quite good back in the mid-1990s when IBM borrowed it to give ASCII files and desktop server operating systems something familiar for them to talk to. I see no reason why IBM i and AIX cannot be propped up on the OPAL firmware and the OpenKVM hypervisor, and therefore also be able to support Linux partitions on a single server or a cluster of servers that could support IBM i and AIX in virtual standalone mode above a distributed virtual SAN on a bunch of Power-based systems. What is good for Linux is good for IBM i and AIX, and in fact, porting these two platforms from the earlier PowerVM hypervisor to OpenKVM and its Acropolis KVM brother makes a lot of sense given the fact that HCI platforms are aimed specifically at midrange shops that do not have a lot of IT people.”
Which brings us to today. As part of its launch of the “Boston” Power9 servers for data analytics and data ingest workloads last week, IBM put out a short statement, in announcement letter 218-200, saying that it has ported AIX 7.2 to the Acropolis hypervisor from Nutanix, allowing AIX to run alongside the Power variants of Linux from Red Hat, SUSE Linux, or Canonical that are also supported on Power processors. The AIX licenses are being sold through Passport Advantage on a monthly subscription basis, and are available in AIX Standard Edition in “cloud ready” raw disk image format for the CS821 and CS822 machines. The Prism cluster manager that comes with the Acropolis platform can create, deploy, maintain, and retire AIX images just as it can Linux images and the AIX 7.2 instance is binary compatible with the version of Linux running atop IBM’s own PowerVM hypervisor, which means companies do not have to recompile their applications to make the switch to the Acropolis hypervisor and the Acropolis stack. This software was available on May 11.
So, whatever IBM can do for AIX, it can do for IBM i. And given that a hyperconverged platform, merging IBM i, AIX, and Linux workloads onto a single cluster that is managed by Acropolis, is something that is probably going to be appealing to IBM i shops, we not only think that IBM should think very hard about porting IBM i to Acropolis. We think Big Blue should seriously consider buying Nutanix and selling it to each and every one of the 125,000 IBM i shops in the world as the new Integrated File System.
Nutanix is one of the darlings of Silicon Valley and one of the first of the so-called Unicorns that in the recent wave of innovation surpassed $1 billion in valuation well ahead of going public. Nutanix has been a public company since December 2015, and I talked to the company last week over at The Next Platform about the new goodies in virtual networking and system and application management that Nutanix is adding into Acropolis.
Interestingly, the company is also adding platform level services that allow for Oracle and Postgres database deployments on clusters to be completely automated, with the open source MySQL and MariaDB derivative as well as Microsoft SQL Server in the works. By adding IBM i to AIX and Linux on the Acropolis platform, companies could have a single compute and storage substrate for all of their databases, and it would be so interesting if IBM could talk Microsoft into making a Power variant of the Linux version of SQL Server available, allowing for Power Systems to be a converged database platform under the management of Acropolis.
Do you see how this can all snap together, and IBM ends up with a modern midrange platform, adding new stuff and keeping all of the good stuff it already has?
Nutanix has just north of 9,000 customers today, and is going to report its most recent financials next week. The company sells appliances with its Acropolis stack embedded, and the vast amount of its customers still use the VMware ESXi hypervisor as their underlying hypervisor, but because of the high cost of VMware’s vSphere virtualization stack, about 30 percent of the base has moved to the Acropolis variant of KVM, which is bundled into the Acropolis stack and is free. (VMware will never be native on Power, so we don’t care much about it here at The Four Hundred. But inasmuch as all of those Windows Server workloads could be pulled onto Power iron on the Acropolis platform, either through Linux or through a Windows Server re-port to Power, should that ever happen, we care.) Nutanix is sitting on $918 million in cash and investments, and has done $900 million in software and support bookings in the trailing twelve months through its second quarter of fiscal 2018 ended in February. In the second quarter, the company had $355.9 million in bookings, up 57 percent, and $286.7 billion in revenues, up 44 percent. The company lost $61.5 million in the period, but is heading towards operating break even and actual profits are probably not too far away.
In other words, buying Nutanix is already expensive, and it is going to get more so soon. The company has a market capitalization of $9.71 billion as this story goes to press, but there is no reason that IBM could not make back the dough through future Acropolis sales and its own market capitalization boost after doing a Nutanix deal. It might take as much as $12 billion or $13 billion to buy Nutanix, mind you, making it the largest acquisition that Big Blue has ever done. But the company doesn’t really have a unified midrange strategy, and this would give it one as well as a way to tap into the Windows Server and VMware markets indirectly. All IBM would need to do is hook up with Supermicro as its preferred X86 server vendor and hand off those hardware deals while telling everyone that its own Power Systems are the better option.
This strategy sure makes a lot more sense than just sitting there, talking about data analytics to a customer base that is largely still doing transaction processing and web applications and trying to manage a cornucopia of disparate server platforms and applications. Here is the real math that IBM needs to be doing. If Nutanix can drive nearly $300 million in revenues against nearly $400 million bookings with 9,000 customers, then pushing that base up to 125,000 customers – the entire IBM i base – is around $4.2 billion per quarter, or $16.5 billion per year at an annualized rate. That is the size of the ecosystem of the AS/400 at its peak, and that is not counting the underlying hardware, which might easily be as much dough.
Yeah. Let that sink in.
If all that IBM did was host some VM applications and provide virtual desktops on Power Systems as well as IBM i database workloads, this would be a huge boon. Nutanix is worth whatever it takes for IBM to buy it, and only IBM has the issue I am talking about with its “legacy” IBM i and AIX bases. Dell owns VMware and pushes vSphere plus vSAN and HPE has SimpliVity, and these two are not really interested in peddling Nutanix anyway. Dell had a reseller agreement that predates the EMC/VMware acquisition, and it is still in effect, I know. Either will sell iron to a Nutanix shop to close a server deal, mind you. They aren’t stupid.
My first criticism is that IBM didn’t invent Acropolis and even see that it was necessary. My second is that IBM spends its money on worthless stuff to do financial engineering and fritters away an installed base – well, actually two if you count both IBM i and AIX – that wants Big Blue to cut a path to a future. My third is that it didn’t buy Nutanix many years ago, when it would have been less expensive, if it didn’t have the foresight to invent something like Acropolis. There is no open source server-storage hybrid, or we could suggest IBM commercialize it.
Still, there is no time quite like the present, and hope springs eternal.