What college-level math do you (as a sysadmin) use?

I tend to think this depends on how much programming you're doing - I find that that CS and programming specific stuff, like knowing the program you wrote will run in O(n) time can be useful to keep in the back of your head. Similarly, some parts of discreet mathematics, such as inductive proofs, set theory, and FSM's can come in handy at times.

But honestly much of what is taught is aimed at the physical and life sciences - things like geometry and calculus I haven't touched in years.

More than any of the specific mathematical techniques, I use the mathematical and scientific rigors that I learned in University.

I actually went to school for a BSc in computer network and information systems. We took basic college math (pre-algebra and algebra with some trig), basic physics, financial accounting, finite, and statistics; i've actually never taken full-blown calc.

This may not sit well with some in the CS world (not taking calc), but I don't feel limited by it. It certainly hasn't affected my day to day life as a sysadmin.

+1 on DTK's comment.

One bit of maths that I keep re-visiting is statistics. Beyond the usual 95th %ile for network traffic, coming up with decent models for traffic growth (linear regression, arima, holt-winters, etc.), determining if a change in system performance is statistically significant or not, and other such things have all proven useful.

Most of it makes me wish I'd concentrated more on that, rather the more matrix-related skills I'd focused on for my main interest at the time, which was computer graphics. A book on "Statistics for System Administration" would be a wonderful thing.

I got my degree in meteorology, so I had a lot of advanced math to take (2 semesters of differential equations -- yuck!). As a sysadmin, I think it's fair to say that I use absolutely nothing I learned in college math classes. What I use somewhat regularly is my high school statistics class.

I've used basic statistical analysis to show that HDFS placed a minimal load on a host system, that one server is more responsive than another, etc.

The math that I still use the most is statistics (and only the most basic) and some analytical geometry to think about performance, capacity and other "curves".

No calc, no differential equations, and entirely too much "commercial math" for budgets and ROI.

I'm going to semi-echo most people's comments and say that the only "higher" math that comes up is statistics, and I wish I had taken more of it.

When it comes to programming, I write mostly functional scripts which require zero math beyond basic binary logic.

I use so little math that I'm really considering taking a college course to refresh myself. It's been forever since I've done any, and I kind of like it, although I honestly wasn't much good at it (aside from Geometry, which just kind of clicked with me).

I use SPC (Statistical Process Control), specifically the EWMA algorithm, to trigger alarms without setting hardcoded limits (often very difficult to do). I also use boxplots to summarize daily measurements (response times, utilizations, etc). Recently I have been thinking about monitoring daily boxplot data (median or Q3) with SPC so that I don't have to look at a lot of data unless a statistically significant deviation occurs.

For comparing WAN circuits, I have plotted RTT boxplots for typical days as a function of the great circle distance between endpoints. These plots include a "speed of light in fiber" line as the "Physics limit". An "unreasonable" circuit will appear as an outlier.

As with DTK et al, higher mathematics helped mostly by making me think hard thoughts. I've used bits of probability, statistics, discrete math, combonitorics, and formal logic. Graph theory too.

CS doesn't count as math, right? I use everything from automata and complexity theory to programming concepts.

The thing is, many of the problems that I solve with math I could just as easily approximate an answer to or ignore.

Example: I once used simple probability to determine that file transfers from a particular vendor were failing because they were trying to send an encrypted binary data over an ASCII FTP connection. Encrypted binary data appears to be uniformly random, so the chances of the byte sequence \r\n appearing randomly are (2^8)^2. Files sized as these were would have that sequence once every three days or so which was how frequent the jobs were failing! od -c | grep further verified that the corrupted files we'd received had no \r\n sequences. Now I could've called the vendor and asked, but I decided to use math to firm up my understanding beforehand.

So Tom, if you're trying to gauge the usefulness of mathematical techniques, please be mindful of this built in bias: your survey subjects can not tell you about those times that they do not apply a technique which might be useful and lost out for it. The Rumsfeldian unknown unknowns.

My degree is Mechanical Engineering. I had Calculus, Differential Equations. I don't remember statistics but I wish I did.

In my sysadmin work, I do shell scripting mostly. Algebra and basic spreadsheet type maths.

I've done admin for Fluid Dynamics groups. My Engineering training was useful for understanding what my engineers were doing, but I never use my math directly.