I'll spell it out: traders don't care about or reward anything in unix-land, regardless of whether that real-time code makes the core of their P/L. It only matters if they can see it (GUI / Excel), and they talk to the developer day to day. No UI, and no spreadsheet, then it's just another cost, maintained by some faceless group of "ops people", probably somewhere offshore, far far away from customers & the prop business. It's not about how good you are, it's about where you are.
I do not need to gather opinions, I worked as prop trader for several tier 1 firms and 2 large hedge funds before I started my own venture. Thus, I am well aware of what is going on in large investment banks and hedge funds in terms of system architecture. When I mentioned DMA execution platforms and pricing engines that encompasses core cash equity and derivatives "mission-critical" IT systems, respectively. I extrapolated from the experience and knowledge I gathered and there is no reasons to assume the firms I and friends work(ed) at should be outliers.
I do not know of a single bank that would in 2012 or 2013 start new major projects off the back of Java.
I do and I think your sample is off-base which is why I am asking for others' experience. But Unix plus (Java or C++) is what I am interested in, not just Unix plus Java.
true about what traders care. But not entirely true when it comes to the true cost of developing and maintaining high throughput/low latency system architecture. Unix systems and in particular C++ libraries are an order of magnitude more expensive to manage and maintain. That is why most investment banks nowadays contemplate (if they have not already started) to re-write their entire (or most of it) libraries using C#. I am not talking about the 10% of systems allocated to the hft operations that every major player nowadays deems an essential part of the business. Some are mistaken in that they think hft contributes to overall revenue more than max 10%. Thus suggestions that C++ and Linux/Unix systems are such an essential part of the business are factually incorrect.
Amazing no one has mentioned the dev environment (i.e. Visual Studio). I think that makes a huge difference from a programming standpoint. Java IDEs were never that good.
I doubt that. Yes, in VS area - but add TFS and all the other integratable platform parts, and InelliJ falls behind. The integrated testing and project planning in TFS / VIsual Studio alone are worth a LOT.
I use both Eclipse and Visual Studio and I find Visual Studio vastly inferior. Just to start with, it lacks the Local History of Eclipse (simple file copies of all past source changes without requiring the overhead of a source repository, commits etc.)