It’s been an eventful year in information technology. It depends on who you ask whether 2016 will go down in history as a year of success or bad news. Meanwhile, the shiny new year is ahead of us, filled with potential. Whether that is potential disaster or victory remains to be seen.
The biggest news was the mega-merger of Dell and EMC and why it happened. We are inevitably moving towards a world dominated by low-priced commercial off the shelf (COTS) hardware, which is severely impacting traditional OEMs. With COTS gear often priced at 33% of a traditional box and much of the need for hardware maintenance disappearing as we move to a cloud model of resilience, the new players such as SuperMicro and the PRC of the original design manufacturer (ODM) world are picking up unit volume.
The impact of COTS has shrunk the worldwide revenue base for systems over the last few quarters, as low-priced units have replaced expensive models in the mix. The new vendors avoid the vendor lock-in by not requiring “proprietary” versions of drives, too, making buys from distribution very attractive compared to locked-in units.
The pressure on RAID array vendors has been compounded by a move to hyper-converged systems. Here, the drives in a server are shared with all the other servers, creating a storage pool, but also reducing the need for stand-alone storage boxes dramatically and forcing lower revenue per sale.
2017 will see this accelerate, with COTS hyper-converged servers being matched up with the same leading-brand software that Dell, HPE and others are using today. The result will be a move towards new vendors, such as SuperMicro, in a segment currently dominated by Dell Tech, HPE and Cisco.
Hard drives have had a rough year. Declining volumes for a number of past quarters signal the replacement of hard disk drive (HDD) storage by solid state drives (SSDs), while announcements of pending deliveries of 100TB SSDs spell the writing on the wall. HDD vendors just can’t match these capacities.
The user base responded by surging SSD orders in the second half 2016, just as much of the foundry capacity was off-line transitioning to 3D NAND. The result is an allocation problem for flash die, complicated somewhat by the rush to replace the Samsung Note 7 with alternatives. This is a short term blip that will resolve sometime in the first half of 2017 when new foundry capability starts adding to the pool of available die, while the new die configuration will provide much higher total storage capacity. At that point the slide in HDD sales will really begin.
PC sales continue their slide. The arch-pundits have promised that next year will be better, as replacement cycles kick in, but I’m personally very skeptical. People are migrating to mobile solutions for general computing and connectivity. Online services such as MS Office365 are biting into sales, while Adobe has blown up the workstation market in video with a pay-as-you-go cloud-based video editing solution.
Virtualization and cloud approaches to high-performance computing are changing the landscape tremendously. Now, the barrier to entry for high-performance is so low, since no supercomputers need to be purchased, that many more users can do simulations and analysis with the result that research projects are completing much faster.
In the general commercial market, the buzz of the year has been Docker containers. This alternative to traditional hypervisors from the likes of VMWare is between three to five times denser in virtual instances per server, which makes server efficiency zoom, while drastically reducing server count needed.
An unintended consequence of containers is that more storage I/O is needed so SSDs are preferable to HDDs in container servers.
Next year will bring server CPU chips with core counts above 20. This will boost the power of a single server by around 3X, while new fast NVMe drives will provide a needed ramp in I/O performance above today’s levels. Networks will jump from 10 GbE to 25 GbE keeping balance in the systems. This will add to the pressure on server buying. All in all, the combination of containers, more cores and SSDs will push down server demand even further, especially given the fact that containers can be installed on existing gear to boost performance.
The advent of the software-defined datacenter is a major game-changer in 2017. We already are seeing control software unbundled from switch hardware and then virtualized on servers. This approach is spreading to storage in 2017. The result is inexpensive merchant hardware that does as much, if not more, than existing bundled approaches. The vendor base changes, with a host of new players involved and the result should be cheaper and faster switching and later in 2017, a burst of new alternatives in storage.
Open-source storage is driving hard too. Ceph, the leading open-source object storage platform is going from strength to strength, with collaboration between Red Hat and SanDisk in 2016 boosting SSD performance and with Mellanox boosting network performance tremendously. Ceph now delivers file and block I/O within the same object storage pool, as “unified” storage.
The sum of all the above points to a reduction in server and storage revenue in 2017, but some increase in unit count. The vendor base will migrate towards the lower-priced COTS ODMs, while this and the fact that standard drives are used in this gear will potentiate business for master distributors.