This article is more than 1 year old

Cloud computing aka 'The future is trying to KILL YOU'

The brutal tech truth that links the problems of Rackspace, Dell, HP, IBM, Oracle, SAP, others

Tech tectonics reshape the landscape – and why

The reason why this is all happening is that during the past ten years there have been a series of advances within the technology landscape that make cloud computing's rise inevitable – even with the NSA revelations.

For one thing, there's been the rise of consumer web giants such as Google that, on running into issues with typical software, were forced to design software systems that could keep up with their vertiginous growth. This meant they created systems – in Google's case, GFS and MapReduce, which are the basis of Hadoop – in which it would be trivial to add another server, or ten servers, into a system and see a gain in performance. This triggered a movement away from integrating hardware with software and towards making software not care about hardware in the slightest, other than as additional capacity.

Once this had been achieved, companies like Google and, later, Facebook, were able to start designing their own data center systems out of cheap components and go to overseas commodity manufacturers such as Quanta or Wiwynn to have them made.

This not only developed the overseas ODMs' manufacturing skills, but it also increased the quality and lowered the cost of customizing systems, leading more and more companies to follow suit.

Last year, for example, the market for ODM shipments straight from the factory to big cloud providers grew 45.2 per cent to $783m for the year across 326,000 units. This is 6.5 percent of the total server revenue and 14.4 percent of market shipments, according to IDC.

"Initially [the rise of ODM servers] has affected the US more than the European market and what that really allows us to do in Europe is know it's coming, it's a threat and we need to make sure we are delivering value," HP told us last month.

As more cheap hardware has become available, companies have been compelled to focus more on software that can scale and deal with failures in its cheap underbelly, leading to less of an emphasis on hardware. And so on.

This cycle has, like everything else in the technology industry, been accelerating with software innovations leading to new hardware approaches that feed back into new software.

Google is thought, for example, to have started designing its own server and storage gear in the first half of the last decade, and then by around 2010 have moved into custom networking hardware.

This was followed in 2011 by Facebook founding the Open Compute, a cross-industry scheme which thinks up and publishes the designs of new stripped-down servers, storage, and networking gear to help other companies use these contract manufacturers.

The upshot is that not only is this way of doing things spreading, but the companies that can take advantage of this Ouroboros of technical experimentation and evolution are able to move far more quickly than their competitors – witness the rise of Amazon against typical retailers – and also depend far less on the services of the traditional IT companies.

The cycle of change spins ever faster

We can see the acceleration of this cycle in types of technology as well. VMware, for instance, was founded in 1998, but by many accounts its virtualization approach didn't have a serious effect on the industry until a decade later. Hadoop, by comparison, was created in 2005 and started to cause change in the marketplace by 2012 or so, though its revenue has lagged VMware.

NoSQL databases are on an even shorter trajectory to interesting uptake (and, later, money), with MongoDB – founded in 2007 – and other NoSQL databases already having an effect on the industry and causing incumbents such as Oracle to look closely at the new genre of databases.

The intervals between invention of new technologies and significant changes in the industry appear to be shortening, which means that if you are either a slow-moving organization or a hardware specialist, things are tricky.

For those who cannot take advantage of this accelerating situation, the future is grim. Without an understanding of how systems behave at huge scale, you can't design the inventions needed to surmount their problems – so as Google and Amazon get bigger, smaller competitors get proportionally less capable of out-inventing them.

This is why we have the troubled OpenStack project, a cross-industry open source software scheme to design software that can run at the scale of these consumer internet giants. OpenStack has been enthusiastically adopted by incumbents such as IBM, HP, Oracle, and others, to help them make tech to defeat their younger rivals. [In a sign that irony is dead, they have even referred to themselves as "The Rebel Alliance." — Ed]

It's a bad time to be either a small provider of cloud services or a big provider of on-premises software. For that reason, Rackspace's decision to investigate being acquired or partnered with should not be treated with surprise – it was inevitable.

By comparison, opportunities are rife for those distributed systems software companies that lack a traditional business to protect. It also aids the Asian manufacturing companies, as they have lower labor costs and are more comfortable with thinner profits than the established server, storage, and network-makers they are displacing.

HP, Cisco, IBM, Oracle, Dell, and other big incumbents should be extremely worried: every trend in technology points to a future that has no bias toward their current profit-generating businesses. The growth days are over and winter has arrived, forcing these companies to battle each other to maintain margins and shipments, and distracting them from the threats coming from below.

When Marc Andreessen said "software is eating the world," he neglected to mention quite how sharp its teeth would be. ®

More about


Send us news

Other stories you might like