Well, we should differentiate whether we talk about the hard sciences or business school/econ dept. at universities. For the below I want to focus only on business school or economics majors: There is a very specific reason why Linux/Unix is so prevalent: Some of the Linux/Unix based software houses understood early on that you groom future users by training them on free machines and apps in college environments. A big sponsor is SAP, it is hilarious to see whole class curricula made up to tailor to the fact that SAP setup a Unix machine in the department complete with software install. Another is Oracle. But to state the obvious: There is not one single software application out in the business and economic real, at least not in academia, that would require linux/unix OSs. Not one single one! (Please name it and I bow down and apologize). No matter the size or complexity of time series in econometrics, there is no reason why a Linux or Unix OS should be favored over Windows. And there is no low-latency app that would require a Linux/Unix kernel. Same with R, students are nowadays so sworn into apps like R that they come out of even grad school or PhD programs and know shit about the underlying mechanics/math of algorithms. Most R packages are not even peer-reviewed, contain errors, the basic R package is very badly designed, meaning, it is not extensible beyond the loading of function code wrapped into "packages/libraries". No dynamic graphics capabilities, very little interactivity with other low-level languages (other than shipping bits and bytes through a very narrowly designed API). My point being, most students are learning the wrong tools in order to be orderly prepared for number crunching and work in the real world. Those who clamor on this site that Linux/Unix is the latest and best shit in the universe are in fact old animals who just do not know better. They still have a deep rooted belief that C++ must be faster than C#, that Apple OS must be safer and more stable than a Windows OS (though thoth claims have been refuted by as credible magazines and research tanks as Dr.Dobbs), they still think that Windows OS sockets and tcp/ip is too slow and inefficient (when it also has been proven that anything but ultra high frequency trading is well served with Windows Sockets). And the best comes next: Now that the C++ guys who always believed in their superiority over the lower .Net coders were kindly pointed the way out (aka their ass was fired because developing even business critical apps in C++ is just not worth the multiple of time it takes to develop and the costs associated), those same guys now try to re-invent themselves and try to convince the rest of the world that their Scala/Akka, Erlang, and what have you languages (that were actually designed for very narrowly defined purposes) solve all the problems in this world. Or look at the Python crowd. It seems to be cool nowadays to perform quant work using Python, one of the slowest languages (cause its interpreted) to analyze time series, chart, trade, what have you. (Python was also designed with a narrow focus in mind, like glue between Super-Computer apps rather than having to write shell script or Pearl) That is the same crowd by the way that bemoaned Windows and its socket and other implementations vs Linux. So, it seems it is suggested we get back 15 years and start all over with interpreted and scripting languages. Isn't it curious that all those scripting languages and esoteric developments such as Scala,... are all targeting non Windows OSs? Simple, its because all the same guys grew up with the "we deserve free, open-source above everything, pirating-yes, we hate Microsoft" mentality. And admittedly, Microsoft had to come along way to present an extremely stable and efficient OS (beyond the Windows Server family). It is ridiculous to read blog posts in 2014 where coders recommend coding in text based editors (I hardly could imagine a more wasteful and time inefficient way to program, let alone debug and compile than writing code in text editors). But I guess there are also people that delight in walking 2 hours to work and back because they think it is healthy, wasting 4/16th (8 hours sleep) of their day while others exercise for 30-60 minutes and get the same out of it and commute to work by car or public transportation. All fine so far, but let's not have those people convince the rest of the world that their approach is the best just because it is old-school. Some comes maybe across as a rant but I just wanted to share my view as someone who has worked for a long time at the front lines of institutional prop and quant trading (though never in IT, but I still think I know enough on the computer science side to comment on). Companies such as banks and hedge funds nowadays run very limited code on Linux machines, most of that being execution related code. Everything else runs on Windows Servers. It is true that Windows in its earlier stages was not fit to handle low-latency trading applicable work but as some high throughput exchanges nowadays prove, a Windows OS/Windows Server can very well handle tier1-revenue exchange order books.
Why would you have a Mac then in the first place if you need Windows would be my follow-up question...writing emails or browsing websites with a ONE-button mouse feels cooler?
Not to sound offensive, but I am not sure that problems in business or economics qualify as "serious number crunching work", which is the point of my prior statement and what I was discussing. I was thinking more along the lines of problems that necessitate the use of a supercomputer, such as modeling supernovas or hurricanes, and why I brought up the hard sciences. In any case, the economics departments that I have familiarity with do in fact place emphasis placed on using Linux/Unix environments for developing and running software for complex problems. Though, this is not to say that they emphasize neglecting Windows altogether. For example, here at U. Chicago, even the more advanced economics courses pertaining to numerical methods in economics involve using C++, parallel computing, etc. to develop and run software in a Linux/Unix environment. Two consulting companies here in Chicago that I am familiar with that do data mining and provide analytic services to businesses, etc. also use Linux/Unix primarily. Though, this is not an exhaustive survey. And at least some hedge funds or related do still place emphasis on Linux/Unix. For example, Renaissance Technologies. Although, most of the information I have about this is either from recruiters or job postings. You may be right though that the majority of banks and hedge funds are nowadays running Windows OS/Windows servers for their applications. You may (probably) know better than I.
Oh goodness, so what? StackOverflow, a IT and programming website with 95 million page views per month runs on Windows Servers (mostly, their linux boxes are for backup and mostly for Redis Key/Value lookups), and so do tens of thousands of sites, small and large. The London stock exchange ran until 2010 on Windows Server, Nasdaq still does (The OS itself was determined to not have been the reason for its last outage). What is your point exactly?
I think the first paragraph of my prior post came out wrong. Some business or economics problems of course involve serious number crunching work, just perhaps not to the scale of multi petaFLOPS (which is what I meant). Since I see no way to edit my post, please disregard that statement.
No offense taken, (and I see your point re super computing) but I would disagree with your suggestion that there may be any benefit of Linux/Unix itself over Windows when it comes to such hard-core number crunching exercises. The reason Linux is used in a lot of such niche environments is, I believe, because a) as long as you just perform calculations you can run on any machine you could imagine (even on an XBox, to just make the point), so naturally the cheapest alternative is chosen given the real computational power does not come at all from the OS, nor does it even come from raw parallelization, but from massive outsourcing of tasks to grids, and b) as mentioned most such researchers are accustomed to code on Linux boxes, though I would again claim there is no inherent and perceivable reason why that would have to be the case. Imho, unfortunately, it even sometimes boils down to perception, you could have wiz kid that codes VB and a mediocre coder who codes in C++ and comparable code would run faster over the VB code (just an example, but its not far fetched for most general code routines, especially on .Net4.5) yet nobody would take a VB coder too serious and no way such programmer would fetch a higher salary as your average C++ coder. Same story with Linux/Unix. It makes for a more prestigious computer lab if Uni ABC can claim they run a Unix supercomputer grid. Lastest what I heard re banks and hedge fund is that most tier 1 firms nowadays run their entire pricing algorithms on Windows machines and most such pricing algos are nowadays originating from library conversion from C++ -> C#, Java. There is a huge demand now on the banking side for capable WPF programmers that are able to transfer the new technology to the screen via GUIs. Maintaining .Net libraries is a fraction of the cost of comparable C++ stacks. Summary: How I see it is that, unless there are true and perceivable reasons why to run Windows or why to run Linux, there is no reason to switch for most corporations, government agencies, academic institutions. Its the same as why countries do not change the way traffic flows (I mean left-side or right-side driving direction), both works, but the benefits to changing either would never justify its cost. There is a linux/unix community and despite them (hopefully) seeing the advantage to program in a full fledged IDE they stick to their text based editors (or minimalist IDEs) because thats what exists on Linux and changing it would not be worth it, they hire and fire their own, they develop for their own code base, and so does the Windows community. The real turf wars are waged by Microsoft and Oracle and a number other big Unix/Linux proponents which is to promote their server software because everything else flows from there.
We like the Mac mini base model and are running Windows 8.1 on it. Simple reason is that the Mac mini is small, next to inaudible, is a power miser (around 15 watts) and does everything we need it to do. CPU usage runs around 10 - 20% so no need to have more power. If we want to travel it is easy enough to take the Mac mini in hand luggage and hook it up at the other location to either a monitor or worst case a TV with HDMI connection. Total cost of ownership is less than that of a laptop. Plus no braindead nutters on a plane squashing in their luggage in the overhead compartment while at the same time damaging your laptop (unless it is in a hard case like a Pelican). OS X is used on a second SSD for imaging of the Windows partition and same second SSD is used for file history. Works like a charm. Would have loved to use some OS X trading software but to-date unable to find something that floats our boat but as someone else pointed out MS Office is outdated for OS X not to speak of the outlook implementation on OS X leaves a lot to be desired.