MiYahoo's future rests with open source and courage
Out Googling Google on bespoke scale
Comment Should Microsoft's bid for Yahoo! go through, the combined company would face one very major infrastructure question - how far is it willing to go in the war against Google?
According to some, Google enjoys a major cost savings advantage over its rivals through a series of bespoke data centers. The ad broker crafts its own servers, using cheap memory, cheap disk and cheap low-power chips. Such systems, destined for failure, cause little damage when they go down because Google's software spreads well across hundreds and thousands of machines. Google treats its clusters of machines as a single entity rather than worrying all much about individual boxes. Along the way, the company saves on energy and infrastructure costs by relying on components that many major companies would consider below their standards.
It's unclear to what extent either Microsoft or Yahoo! mimic this approach. We know that Yahoo! through the Hadoop software has placed a great deal of emphasis on replicating the MadReduce software used by Google. This code splits tasks across numerous machines and then cobbles together end results for users and is a major reason that Google's data centers work well. In addition, Microsoft has been matching Google of late by setting up $500m+ data centers in cities that boast cheap power.
No one, however, ever talks much about Microsoft or Yahoo! constructing their own servers. In fact, the two companies are well-known as the largest buyers, along with Amazon.com, of energy-efficient gear from Rackable Systems.
Most companies would consider building custom gear a waste of time, effort and money. But such a strategy may well make sense for the largest service providers on the planet. In a war to serve the most ads, searches and web pages at the cheapest price, a company would want to be damn sure that it's keeping up with rivals click-for-click.
A combined Microsoft and Yahoo could well decide to maintain their existing operations and move along with different, separate systems. Yahoo goes the open source software, Hadoop route, while Microsoft relies on its homegrown software. Although, neither company appears to have matched Google's economics using either approach so far. Why keep banging two heads against the wall?
Instead, MiYahoo could opt to steal one of Google's major points of leverage. It could start from scratch just like the youngster in Mountain View, replacing old computing systems with new gear based on a fresh scheme.
This shift would require a great deal of courage from Microsoft because, we believe, the vendor would need to move far away from its own proprietary software. The software/service provider would have to tap into the open source work done by Yahoo! and embrace the vibrant open source world where, frankly, the really interesting web infrastructure technology is being built.
In so doing, Microsoft would send a clear message to many of its largest customers. "A Microsoft-based web infrastructure is probably not the most economical or exciting path for you to take."
The superstars of today's internet already know this. They're running on software such as Apache, Linux and MySQL instead of Microsoft's family of Server software. Many more traditional large businesses, however, remain Microsoft shops and might begin to wonder if that's a good long-term bet when even Redmond has turned to the dark side.
Should MiYahoo be as drastic as Google on hardware and get into the server and switch building game? Well, that would be the boldest step of all.
If the combined company isn't willing to go that far, one would think it could demand an awful sweet deal from, say, Sun Microsystems or Dell for a massive amount of low-power systems. Both server makers have programs in place to do such work and could meet the necessary scale.
Regardless of which path it takes, the new company will need to change and focus on consistency and efficiency.
MiYahoo, as we see it, would consume even more hardware than Google and have a shot at taking the economies of scale edge. All that's required is some solid management and a pair of planet-sized balls.
Ultimately, it's these types of decisions and moves that will determine whether or not MiYahoo was a good idea. ®
Sponsored: Hyper-scale data management