Tom Petrocelli's take on technology. Tom is the author of the book "Data Protection and Information Lifecycle Management" and a natural technology curmudgeon. This blog represents only my own views and not those of my employer, Enterprise Strategy Group. Frankly, mine are more amusing.

Friday, August 18, 2006

Fibre Channel Switches - The Ranks Are Getting Thin!

Now that Brocade is buying McData (which bought CNT some time ago) the number of FC switch vendors is getting to be quite small. Three to be exact. Once the deal closes, all that will be left is Brocade, Cisco, and QLogic. Sure, there are some smaller players that will insist they are FC switch vendors but they really are iSCSI switch makers with a bit of Fibre Channel tossed in. None of them make the range of product that the FC Three do nor do they make director grade switches. Everyone else, including the big storage vendors, buy from these three companies.


Is this good? I always feel that more competition is better. It drives innovation and lower cost. From the point of view of the eventual consumer (mainly IT shops) this is a bad thing. With fewer vendors selling FC Switches, prices may rise and product development slow. Costs more, less filling. I can only imagine that the arrogance factor will increase a bit too.


From a product perspective this isn't a happy place either. Both companies have almost total overlap of products including directors, mid-level switches, and small switches. You could pick through the products and find small differences but nothing that will give the combined company a huge competitive advantage. Fewer overall product choices for consumers is not good and it raises the risk of technology monoculture issues. With so much of the market concentrated in one company - Brocade - the chance of a bug effecting a large number of companies is one that needs consideration.


Which brings us to the "why". Where is the value of the merger or acquisition? Truthfully, Brocade gets nothing dramatic from McData. Granted, many feel that the Brocade director is not nearly as good as the McData one. McData also has some pretty good mainframe connectivity capabilities, mostly ESCON and FICON. That's not the biggest chunk of the market, so it's not enough to buy a company the size of McData. Perhaps, the reason is that they got all excited at prospect of swinging to a profit that simply had to go out and spend that money! Woo hoo! Keep these guys away from the Home Shopping Network.


The best reasons for Brocade to do the deal are to eliminate a competitor that sometimes gives it fits in certain accounts (such as EMC) and to bulk up. That suggests that Cisco is hurting both companies. With it's breadth of products, enormous, well developed, channel, and shear scale, Cisco is pushing these guys around a bit. Or at least scaring them. That makes this look like a defensive move rather than a strategic one.


Maybe the deal makes sense financially. I'll leave that to the stock analysts. Investors seem to be voting this one with their feet. Brocade's stock plunged right after the announcement and hasn't recovered despite a good earnings announcement. That's confidence for you. While I don't think customers will like it in the long run, Cisco will likely approve. It gives them one less target to worry about. They can now focus on crushing only one competitor. Sweet for them but not for the rest of the industry or IT consumers.


Monday, August 07, 2006

Linux Breaks Apart Under the Open Source Model

I first worked with Linux back in the mid-90s. At the time there was only one real distribution, the Slackware distro, and it was a serious chore to install. It required manual configuration that was much more difficult than your average UNIX box let alone Windows or DOS.

Back then, as we experimented with Linux, one thing became abundantly clear - it was a one size fits all approach and that didn't work. The truth is, operating systems variations proliferate because different missions have different needs. A server OS is different than a desktop OS. A secure server is a totally different animal altogether. The requirements of multimedia development, programming, and games all vary widely. Linux in the old days really wasn't designed for that.

To be fair, Linux was (and technically still is) just the kernel. Diversity tended to be across operating systems then instead of within them. Desktop applications? Windows. Server applications? UNIX and maybe Windows NT. File server? Windows NT and Novell for the die hards. Multimedia and graphics? No question about it, go Mac.

This is in sharp contrast to the OS world of today. Starting with fragmentation of UNIX in the 1980's, we now experience so many variants of operating systems, that the mind boggles. The latest list of Vista variations shows six different types and that doesn't include the Microsoft server operating systems or versions of Windows for embedded and mobile applications. Windows Server 2003 also has a bunch of variations, such as the Storage Server, and Longhorn is expected to be completely modular. Just trying to figure out which future Windows to use makes my head hurt.

It should then come as no surprise that the once unified Linux now ships in more distributions (a fancy word for versions) then there are mints in a Altoids tin. The Distrowatch website lists a top 100 distributions. That implies that there are more than 100 hundred distributions! I can believe it. There actually are hundreds of distributions. While many are simply different packaging, most are specialized distributions aimed at increasingly narrow markets. Talk about slicing the baloney thin. There are special multimedia distros like Gen Too, desktop-oriented ones like Ubuntu, server versions like Red Hat Enterprise, and many more. Some are designed for lower end machines, such as Damn Small Linux, and a few, like Slackware, seem to be heirloom distributions. More to the point, there are dozens of entries in each category, even relatively narrow ones like embedded real-time operating systems.

Ultimately, this is a ying-yang situation. While being able to find an operating system variant that suits a very narrow need is attractive, supporting that many operating systems drains resources away from the core system. It is also pretty confusing. Just trying to compare this many versions of an OS can be daunting. It's no wonder that only a small number of distributions make up the majority of installations with a few additional ones staking our majority claims only in special niches.

Altogether, this is what happens with open source. Everyone wants their own flavor or feels like tinkering with it. Before you know it, variations proliferate like zebra mussels. Even fairly obscure new technology like VoIP fractures quickly under the open source model. There are lots of SIP servers out there and a bunch of variations on one implementation called Asterix. Great stuff but there's no control.

It is clear that there should be variations on the Linux theme but a limited number. One for servers, one for desktop, one for embedded, etc. The very nature of the Linux and open source world makes this unlikely. Unfortunately, the proliferation of Linux distros will weight down on Linux, hurting it in the long term.

The fracturing of the Linux world into hundreds of variations is a side effect of the open source movement. In the end one of two things will happen. Either most of these distributions will fade away as the programmers get bored with it or Linux will eventually fail altogether, paving the way for more years of the Windows hegemony. I wonder if Bill Gates and Steve Ballmer are sitting down with a glass of champagne right now toasting Linux.