Today is 3/14

In two years we'll discover the fourth digit, but the fifth will baffle us for another five years after.

No it doesn't. But The Database Column would have you believe it.

A more fitting title to this may have been What Are These Guys Smoking?

Read the article here: MapReduce: A major step backwards.

Then take a deep breath and let it out the big "Huh?"

My first reaction to this is simple. Your typical database is based on perfectly structured data. Where the data doesn't fit neatly into the structure (schema) you transform it, often times using an ETL tool of sorts (thats Extract, Transform, Load).

Search data is anything but perfectly structured. Google indexes a whole lot of different document formats: HTML, PDF, word docs, excel files, and a whole lot more. This stuff doesn't exactly map to a neatly defined database schema does it?

MapReduce is a time tested proven and infinitely (well...) scalable method for building a resultset from a very large dataset. The neatest part perhaps is that the data that is mined can be in any number of different format stored in any media available. What does this translate into?

1. Cheap grunt servers
2. Cheap storage
3. Manageability
4. Scaling to the moon

Let's talk about that for a second. A low cost Linux server can be put into commission and last an enormous amount of time without every being upgraded. The data will be proliferated through many different systems providing an awesome set of redundant and low cost storage devices. Managing the servers is easy (in relative terms), seriously, they are grunts designed to do one task and do it very well. Upgrades should be minimal since they only have to change if the physical layout of the data changes, and it doesn't have to very often I bet. Scalability, no problem, plug in another thousand servers that are just begging for datasets to reduce and off they go.

Allow me to elaborate on the manageability front here. Say you have a linux server designed to dig through a crap-load of this data that you've been collecting through various means (spidering, general data loads, whatever). We're talking about a finely tuned $600 server (retail) that will kick ass at this job. More likely, we're talking about several thousand of these wicked machines. But, the only time you ever have to alter the innards of these beasts is if they fail or the shape of the data changes. With MapReduce you don't have to change the shape, just add a new generation of finely tuned $600 servers (retail) that that can see and process a new shape, and the new data. The old servers are just the aging rock stars, still jamming to the same tune, eventually they'll retire and the data will evolve. Ultimately the users question gets answered by everyone listening and reduced to the most relevant resultset.

Try that on any modern vendor provided database system and you'll probably find that it just can't be done that way. You have to have all of your data neatly ordered and ready to go. Show me a product that can evolve with your data, without having to migrate or transform that data's physical shape. I'll bet you can't. The reason you can't is because you have to funnel everything through their engine and it can't possibly be capable of knowing how all of the underlying data in a system like Google's. Google can add any data, in any shape it wants to their ginomrous clusters and the very nature of MapReduce screams "I don't care!" because the workers are doing the work.

So what were these guys who wrote this article writing about? Beats me. I think they were just trying to tell IT managers that Google is wrong and they know better. I think they are full of it if that's the case. Or perhaps they really just don't get it. I dunno.

Quite frankly I'm tired of hearing that we aren't in a recession, but we may be headed towards one. I don't buy the old definition of two consecutive quarters of declining GDP. That just doesn't sit well with me.

I want to know this instead. For every US dollar I spend, how much of that stays in the US economy?

I fill up the tank on my Saturn Aura, costs about $36 these days or more depending. How much of that $36 stays here?

When I buy three 2 liters of Diet Mt. Dew for $3.63, how much of that stays here?

Is there a measurement for this? I don't know. If there is, I'd like to see it. I'd like even more to hear about it in the news if that's possible.

It was cool to see Slava Kozlov back in Detroit finally. We've waited all season for this game.

My wife was pleased that both the Thrashers won (5-1 in an odd Wings loss) and that Kozzy scored a goal against Dominick Hasek whom he was traded to Buffalo for in 2001. Attaboy!

It was quite an upset for the Wings, #1 in the NHL, to lose so badly to the Thrashers who rank 21 overall. Or rather, they did until this evening. I'm not sure where this game puts them.

Even worse though is that the Wings have now lost three games in a row. Each game they were trailing by 2 points at some point. It's their biggest losing streak of the season and all of the superstar players are dressed and on the ice. Odd that.

In the end it doesn't matter. Detroit will surely make the playoffs so it was great to see Slava skating at Joe Louis Arena again tonight and leave with what is certain to be a smile on his face.

We miss the old #13, even though the new #13 is definite keeper himself.

Game on!

Is there any wonder why young people are flocking to this guy? He had it right in '88 and he still has it right today.

Bottom line: Keep it simple you morons. Stop complicating our bloody lives.

Watch and see:

P.S. I just have nothing interesting to talk about in the realm of growing software. So enjoy!