Virtualization and Cloud executives share their predictions for 2017. Read them in this 9th annual VMblog.com series exclusive.
Contributed by Leo Reiter, CTO, Nimbix
2017: "Accelerated Compute" becomes known simply as "Compute"
I
predict 2017 will be the year when "accelerated compute" becomes known
just simply as "compute". This is a direct response to the use cases
driving up utilization the most, and the explosion of accelerator
availability in both the data center and the public cloud. (GPUs and
FPGAs are readily available in the Nimbix Cloud
to speed up compute intensive workflows like simulation, machine
learning, and genomics). As these use cases continue to ramp up in the
Enterprise (particularly machine learning), we'll see even more demand
for computational accelerators.
CPUs have been king for decades, and
serve the general purpose quite well. But what we're seeing now is an
emphasis on deriving insight from data, versus just indexing it, and
this requires orders of magnitude faster (and more specialized) resource
in order to deliver feasible economics. It's not that computational
accelerators are necessarily "faster" than CPUs, but rather, they can be
deployed as coprocessors and therefore take on very specialized
identities. Because of this specialization, they can be programmed to
do certain very discrete computations much quicker and at lower
aggregate power consumption. Application developers and ISVs are
pouncing on these capabilities (and their increasing availability) to
create amazing new products and services.
A good example of a red-hot technology in this space are GPU-accelerated databases, such as GPUdb from Kinetica
(available as a turnkey workflow on the Nimbix Cloud). Rather than
focusing on indexing massive amounts of information like a traditional
RDBMS, it's used to ingest fragments into memory for tremendously fast
queries. In fact the queries are so fast that it blurs the line between
analytics and machine learning (after all, machine learning involves
processing massive data sets very quickly in order to create "models"
that operate somewhat like human brains). Despite the advanced
computing underneath, these tools serve traditional enterprise markets,
not just "research labs". Not only does its product name imply it, but
the use case simply would be impossible without GPUs. This is a very
real example of mainstream technology that demands computational
accelerators.
I
talk to customers and business partners from all walks of life every
day. The one common thread they all seek is more accelerated
computational power (at reasonable economics) to do even more advanced
things. I don't see this trend slowing down anytime soon, which is why
I'm predicting that we'll drop the "accelerated" in front of "compute"
as it will become a given.
##
About the Author
Leo Reiter is a virtualization and cloud computing pioneer with over 20 years
of experience in software development and technology strategy. Prior to
Nimbix, Leo was co-founder and CTO of Virtual Bridges. Leo is an
entrepreneur with a strong background in LeanStartup and Agile
methodologies.
