Wikibon invokes the old gods to make six tech predictions for 2017
Consumer volume driving ARM and OpenCAPI forward, and its lack hobbling XPoint
Analysis Wikibon's chief technology guy, David Floyer, has put out his top six disruptive tech tips for 2017, seeing that ARM and Open CAPI will separately grow at Intel's Xeon expense, the cloud at on-premises' Server SAN expense, and flash at disk drive's expense, while XPoint will be a 2017 dud.
Here's his list:
- AWS will introduce server SAN software storage system for enterprises
- Introduction of volume ARM-based servers
- Big Data and Map-reduce functionality inside flash drives announced
- Open Power with CAPI and NVIDIA GPUs will radically outperform x86 for big data and AI computing
- Steep decline of HDDs shipped
- XPoint will stall
Amazon introducing a Server SAN offering? Why? He thinks it's growing in acceptance because "the cost of moving data, especially large amounts of IoT data, is far more expensive than processing the data locally". Wikibon supports the idea of Server SAN replacing SAN. He says that with large amounts of IoT data both in the cloud and on-premises (local clouds) then "having the same storage software and APIs in both the central and local clouds is a giant step towards cost-effective and performant hybrid clouds".
Floyer believes consumer acceptance of and hence volume for a technology is vital for its development and adoption by business. The phenomenal rise in smartphones using ARM processors is displacing low-end PCs and eroding the PC market, leading to declining x86 processor volumes. This, together with Xeon's upmarket functions, feature set and cost is proving the impetus for developing ARM servers.
Indeed, he sees a role for ARM CPUs to be embedded in flash controllers to run Big Data applications at a drive level, such as Map Reduce. This is a different take on Seagate's smart drive idea, which they applied to disk drives with the Kinetic technology of Ethernet-accessed, object-style, key:value store disk drives. As far as market acceptance goes, Kinetic disk drives appear to have fallen short, and Floyer evidently prefers smart flash drives.
The Wikibon CTO sees OpenCAPI, with its CPU-memory-IO adapter interconnect scheme, as a game-changer, not least because NVIDIA GPUs have consumer volume support.
The progressive flash takeover of disk drive storage responsibilities will continue in 2017. Performance disk drives (15,000rpm) are being given the heave-ho by flash, with 10,000rpm spinners facing an increasing onslaught as 3D flash gains layers and so capacity, and loses per-GB expense. The proportion of PCs using SSDs instead of disk drives is steadily increasing. Flash foundries are increasingly talking of QLC (4bits/cell) flash as a fast-access, cold data store, raising the prospect of disk archiving getting squeezed out between cheap and fast flash on the one hand, and cheaper-still tape on the other.
The lack of any consumer acceptance for 3D XPoint systems or volume application demand for XPoint leads him to believe XPoint will not be a significant aspect of the 2017 tech scene. This is at least partly due to Intel over-hyping the technology and indications that products using it are coming in way under early specifications. He doesn't think that XPoint ship volumes will be too low to fund future investments and manufacturing volume increases, which might drive down prices.
There is a prospect here that XPoint could fail.
Alternative post-RAM non-volatile memory technologies such as Resistive RAM are being developed and the bulk of the non-volatile memory foundry suppliers have absolutely no interest in XPoint succeeding, meaning Samsung, SK Hynix, WD-Toshiba, and others.
With Diablo and Xitore pushing the idea of DRAM caches front-ending flash or ReRAM or even XPoint and so enjoying DRAM speed rather than slower NAND, ReRAM or XPoint then the relative speed on NAND vs ReRAM vs XPoint becomes less important. XPoint isn't the only filler for the DRAM-flash gap and Intel, with its marketing function's effectiveness devalued because of the seemingly gross over-hyping of XPoint, has a tremendously difficult job in establishing a generally perceived role that only XPoint can meet. XPoint could be coming closer to a crossroads. ®