Consider, IBM has to make big inroads into the cloud computing world as a commercial necessity – to sell lots of servers (hardware) and lots of layers of management applications (software). It’s a hardware, software and services business, right?
Facebook operates one of the world’s most massive web-systems, far bigger than any bank and far different from any bank, which in fact are not web-systems but just very traditional datacentres and networks. Facebook’s front is as social platform, but one of its key competencies is massive web-systems operations. Tim o’Reilly said Operations is the new secret sauce, and Facebook is a master operator – just look at the travails of Twitter or Tumblr, both notorious for their multiple system failures.
So perhaps Facebook and IBM have some common business competencies. Both know a huge fundamental amount about hardware, servers, networks, configurations and operations – on a very big scale. Both have some of the best brains in the industry.
But how they see the cloud world is remarkably different – they are laying their bets on difference races. For example IBM just had a major cloud announcement – IBM Unveils Cloud-Computing Initiative Aimed At Large Companies. This actually means “companies with very large computing requirements” and in this context Facebook is a very large company.
The gist of the IBM announcement is that “IBM is attempting to differentiate itself from some of the main players who have already made big bets on the cloud. Competitors such as Amazon’s Web Services, Microsoft and Google”. That’s a good aim and given its assets IBM should be able to do that.
The IBM SmartCloud “will enable clients to use a Web-based interface to install applications and configure databases on a platform provided by the company“. That notion centres on configuring and managing servers as the key unit.
On the other hand Facebook has just announced its Open Compute Project, which makes public and gives away all of it’s design and construction plans and details for its new data centre – all the building works, the aircon, the electrical wiring etc – and all the server architecture and design. Facebook designed their own “vanity free” servers, in the same vein as Google using commodity servers. They don’t need brands, fancy face plates and video cards in their servers. For its efforts, Facebook’s new data centre uses 38% less energy to do the same work as it’s existing facilities, while costing 24% less. For example, the data centre uses ethernet-powered LED lighting and passive cooling infrastructure to reduce energy spent on running the facility.
Complex management interfaces, architected upwards from a server-centric model, can’t manage what Facebook needs. So they’ve architected it themselves from the top down, while at the same time redesigning hardware from the bottom-up.
IBM’s not doing anything stupid, they’re just doing what the rest of the industry is doing. The point is that Facebook isn’t perceived as a technology company, far from it. It’s often not perceived as an innovative company – when they do innovative things the social media is repleat with condemnation and those stating “that was obvious any fool could have done it”. Facebook is not perceived as an expert in data centre design and rarely as an operator of massive web-scale systems that eclipse the complexity of the banks and airlines.
But Facebook is all those things. It’s extremely technology-astute, it’s innovative, it’s ground-breaking, it’s astoundingly connected in web-services and data-interfaced world, and it has access to all the capital it needs to do anything.
And it’s Open Compute Project may just unleash a whole wave of creativity which will dramatically change the way massive cloud computing is implemented, and for sure will encourage many start-ups to do things the Facebook way as versus the “industry” a la IBM way.
Should we expect it, as the norm, that companies with web-systems like Facebook could not use what IBM offers for cloud computing? This is kind of a high level statement, but true. If they did use the industry-commercial offers then this would kill Facebook, or Google, same story. Facebook and Google aren’t building their own systems from the ground up because they really want to, but it’s because they know they have to, for economic and performance reasons, because they understand fundamentally what is needed, and they can.
The cloud is going where the likes of Facebook and Google (and Amazon and Rackspace) are taking it. Not where the IT industry might hope it might be based on their current inventory and skill set.
The cloud shift is accelerating. It’s not hype, it’s not only remodelling business and business opportunities but also the whole IT industry and it’s ecosystem, and it hasn’t even started yet!
What do you think are the implications of Facebooks own-build of cloud computing?
What do you think is the ultimate potential of the Open Compute Project – just PR or a shift in industry power?