From Integrated Systems To Disaggregated And Composable
January 6, 2020 Timothy Prickett Morgan
As IBM midrange shops well know, there used to be a number of suppliers of integrated minicomputer systems that included all of the hardware and software that was needed by a company to automate its bookkeeping and operations. Many of them – notably the Hewlett Packard 3000 and the Digital VAX and Alpha lines – are gone, and the IBM i on Power platform, the descendant of the System/38 and the AS/400, is in many ways the last of its kind.
But there are other kinds of integrated systems, and we have discussed the market for these machines – included integrated systems, certified reference systems, and hyperconverged systems – before and we like to keep track of how they are doing because it gives us a sense of the appetite for any kind of integrated system. And we think these kinds of stacks will play out on future disaggregated and composable infrastructure in the future.
It is important to define these machines before we talk about how well or poorly they are selling. So let’s review some definitions. Hyperconverged infrastructure means taking a cluster of virtualized systems and running a virtualized storage area network on the same cluster where the virtual machines supporting applications are running. Nutanix and VMware are the dominant suppliers, but there are a few others. There is a new sub-category of hyperconverged storage that has come into the market, what IDC calls disaggregated hyperconverged, where the storage is on a separate appliance from NetApp, Dell EMC, IBM, or Pure Storage. By breaking storage free from the nodes, compute and storage can be scaled independently from each other – one of the big drawbacks of earlier hyperconverged systems. Certified reference systems and integrated infrastructure is comprised of machinery that has compute, storage, networking, and system management all woven together and ready to deploy. Integrated platforms add layers of systems software and sometimes application software on top of that, such as Oracle’s Exadata systems or IBM’s now defunct PureApplication platforms. The way IDC does the math on these systems, the whole system is counted as one thing; IDC does not track individual compute or storage nodes. As we have pointed out before, this is how people talk about these systems – it is one system, not a cluster of systems – so that makes sense to us.
If you add up all of these various kinds of systems, then the world consumed $3.75 billion in various kinds of converged systems in the third quarter of 2019, the latest statistics available from IDC. (It takes a while to come up with the data for these converged machines because they often involve components from multiple vendors.) That is a 3.7 percent jump in sales compared to the same quarter in 2018. That growth was mostly due to the growth in hyperconverged systems, as you can see in the chart below:
In the third quarter, sales of hyperconverged systems, which dominate the converged systems category with 53.8 percent of total revenues, rose by 18.7 percent to just a tad over $2 billion. Integrated platforms, which is probably the category that is most like the IBM i platform (but as far as we know the IBM i sales are not included in this data), had a 6.5 percent decline to $475 million year on year. Integrated infrastructure and certified reference platforms, which tend to have multiple vendors participating in the stack, accounted for $1.26 billion in sales, but declined by 8.4 percent and accounted for a little more than a third of the market for converged systems.
By the way, you can build a converged system on one of the major public cloud, and none of this data takes that into account. This is all data for on premises gear, and not including the on premises gear sold by Amazon, Google, or Microsoft.
What is clear as 2019 was coming to an end is that converged systems are somewhere between niche and mainstream, and integrated platforms were a little too vertically integrated for most customers but certified reference systems and integrate infrastructure, which leaves more room on the software side, are more popular, and that hyperconverged compute and storage has gone from noise in the data only five years ago to having the largest slice of the pie. But, that said, hyperconverged is not for everybody.
We think that in the long run, disaggregated but composable compute and storage are going to be the architecture of choice for future platforms – this is what the hyperscalers and cloud builders have created – and this does not fit into the IDC definitions. Disaggregation means smashing the motherboard in the system (logically speaking) and using protocols and interconnects to link CPUs, memory, memory, and storage to each other; composable means essentially creating systems from these components on the fly to run specific workloads, something a lot more physical than a logical partition.
There is no reason an integrated system or integrated platform (which adds database and other middleware to an integrated system with the basic systems software) cannot – or will not – be deployed on disaggregated and composable infrastructure. At some point, this will just be called a system, much as is the case with cloud. Even bare metal machinery has cloud attributes these days.
We live in the future. Amazing, isn’t it?