Solaris 10 Success Story

A client called my firm up to help them improve the performance of their business modeling applications.

They were running Monte Carlo simulations over huge data sets as well as linear programming to model their business landscape. When I got to my client's datacenter, I immediatly became concerned. I noticed they were running the latest Linux kernel, the 2.6.11.10, on SMP machines.

I sat by the servers and could actually hear the drives thrashing about as the kernel inefficiently paged the data. In my head, I began thinking this will be a tough case as Linux lacks a tool comparable to DTrace. I ran the "top" program and indeed the kernel was thrashing the drives. I checked the amount of memory on the system viaproc, and it had 16 Gigs of RAM running Opteron Processors.

Unfortunately, I had to inform the client that their sys Admins were incompetent. First off, the data was connected to the internet, albeit through a firewall and NAT. I informed them that there is a security vulnerablity in SMP Linux kernels discovered by a PhD researcher. Linus Torvalds BRUSHED OFF THE RESULTS, because he couldn't understand the problem! Linus, when he seems to not understand the scope of a problem, often ignores engineers with 25 years of UNIX kernel experience. It happened with Joerg Schilling and now it is happening to Colin Percival.

This is what Colin says about linus: "it is at times like this that Linux really suffers from having a single dictator in charge; when Linus doesn't understand a problem, he won't fix it, even if all the cryptographers in the world are standing against him."

Joerg Schilling had the SAME IMPRESSION! The achillies heel of Linux is Linus Torvalds! When one person says something, it may be a case of a grudge, but 2 people, it is becoming a problem! That and the BitkKeeper fiasco really is souring people.

Back to my client. I immediately held a conference with the President and CEO of my firm's client. I demanded complete authority over their datacenter and that I not be disturbed by their incompetent Linux IT staff.

After being granted permission to take whatever necessary measures were needed, I marched into the datacenter. I backed up the user data. I then downloaded Solaris 10 ISOs. and burned them to DVD. I then COMPLETELY WIPED THE LINUX PARbreastIONS!

I plopped in the Solaris 10 DVD and in less than an hour I was up with a base install. I installed their custom POSIX compliant business modeling code, which was basically a bunch of Scilab scripts. I ran a test set of problems at full capacity. Some very minor disk thrashing occured...5 % of what was happening on Linux. Wow. Talk about a professionally designed OS! Linux had 95 percent more thrashing!

I quickly wrote up a DTrace program with the appropriate sensors. Turns out these Linux clowns wrote very inefficient SciLab scripts! I re-wrote some of the code to have better O(N^2) performance with the help of DTrace to maximize the architectural bandwidth.

My client's boss was impressed. Here were the recommendations I gave him:

1. Get rid of the INTEL boxes because of the SMP vulnerablity. I recommended him some nice SUN boxes matched with Solaris 10. The upcoming ZFS filesystem will be a god send with their datasets.

2, Fire your linux staff. They can all be replaced by one competent Solaris admin. Linux staff normally grew up on PC bucket hardware and don't know what Uptime and reliablity is! Most of them have never seen how a proper datacenter is supposed to be run!!