I had a bit of a false start in the blogosphere, but this time is for real. I will do my best to routinely blog about Solaris, DTrace, performance, and whatever else moves me.
To kick it off, I felt like I should address a question I routinely get from friends, co-workers, and customers - Why did I join Forsythe? Well, I was having a great time at Aeysis, but Forsythe brings some interesting new opportunities to the table. In addition, I think they have a great story to tell, and Paul Zajdel - my boss - really gets it. Solaris is the "Crown Jewels" for Sun (as he puts it), and Solaris 10 clearly proves that innovation in the OS is not dead. This is why our team is focused on not only helping customers get to Solaris 10, but to also take advantage of all the cool new features (okay, they are not all that new, but to customers running Solaris 8 - they certainly feel new. Solaris 10 dropped in 3/05 so new may be a stretch). Performance (DTrace) is obviously a huge part of what we do, but data center consolidation with the help of Containers, storage re-architecture with ZFS, SMF, FMA, etc. are all keys to improving utilization and reducing cost.
Here are a couple of interesting statistics - 85% of my work over the past two years has been to look at applications that had serious issues after an upgrade - one or more of the hardware, application, or the OS was upgraded (which meant an investment in either capital or time). That is a huge number. The other interesting thing is that 90% of the time, only the applications were changed to address the issues. Now for the best part - the vast majority of the time, it was 10 lines of code or less that accounted for the lion's share of the performance problem!
This is why those of us at Forsythe are so keen in putting the "VA" back in VAR. We firmly believe the "Value Add" in that acronym should mean something. So often today there are resellers that are config/quote/invoice. Hardly any value in that. An architecture rooted in observed performance characteristics is the only way to go. Having true systemic observability is amazing. With DTrace, I don't care what language an application is written in or, to be honest, what it does. With DTrace, it is a very quick process to uncover where time is being wasted. One application I looked at, we spent 5 minutes and _commented_ out one line of code for a 67% win. The application was 1.2 million lines of C++ code (I would hate to have had to code review that instead of using DTrace). Had this customer merely thrown hardware at the problem, there performance would not have improved much (if any) without this change. A mere 5 minutes to validate a solution. Clearly, well worth the investment. Oh, and they didn't have to upgrade the hardware as originally planned. Nice little savings there as well.
So what I hope to do in the future is detail some of these types of wins with as much as I can put out here without violating the NDA's that we have in place to allow us to look at these applications.
By the way, I still consider this the Texas Ranger's Blog (nothing to do with the baseball team). I just don't use that as the title of the blog; however, I still whole-heartedly agree with the (mis)applied phrase associated with the Texas Rangers. It is not a critical mass of people that solves the problem - it is having the right number of the right people.
