Linux 2.4 has arrived, and the technology is great. Its just too bad that technology alone doesnt make a success story.
OK, Linux 2.4 can do serious enterprise computing lifting. With it, you can run up to 32 Pentium-class processors using symmetric multiprocessing (SMP). Come the day Itanium is really released, Linux will be the first operating system up and running on it, and youll be able to scale Itanium SMPs up to 64 processors.
The networking layer also is much more scalable, thanks to “wake-one” scheduling. That ensures that only one processor, instead of a gaggle of them, is listening for network traffic on one socket. Linux 2.4 has much better memory support, as well. Even 32-bit Intel processors running Linux 2.4 can address up to 64GB of memory. Thats not too shabby for all but the most-demanding database applications.
Some technical goodies are still absent. A journaling file system (JFS) is a necessity for an enterprise OS, and 2.4 doesnt have one. But thats a minor detail. Come 2.4.1, which probably will be out in February, youll be able to easily deploy the ReisterFS JFS.
So whats Linux 2.4s problem? While the Linux network operating system is second only to the Windows 2000/NT family on Intel processors, according to IDC, to make the next step forward, Linux needs nontechnology goodies to convince CIOs to take the Linux plunge.
Heres what I think Linux needs to enable you to make a real run for contracts currently going to W2K/NT integrators. First, you need to educate decision makers about Linux and open source. Too many senior IT folks still think of open source as this scary thing that may come back and bite them with licensing problems.
Next, Linux value-added distributors and ISVs need to stick to open source and the Linux Standard Base. If Linux starts forking, its back to the Unix wars of the last two decades, and the only winner there was Microsoft. Lately Ive begun to see signs that Linux vendors are getting more interested in proprietary Linux product add-ons. If that continues, we could see attempts to create proprietary Linux kernel programs and then Linux will be in a world of hurt.
Linux desperately needs serious management tools. Here, progress is being made. Caldera Systems Volution looks pretty darn good for managing multiple servers, in my tests. Linuxcares Linux Managed Services also looks like an excellent service play. Finally, there are too many Linux certifications. Red Hat has one, the Linux Professional Institute has one, and several training companies are pushing out other certifications. Linux needs one solid certification path that will be as well-known as an MCSE or CCIE.
If it gets all of those pieces into place, Linux will step beyond being a big business to become a big business.