I am going to build a trading system and use with the broker "Interactive Broker" API. The API supports C++ Socket, Java Socket, DDE, Active X APIs in Windows. The API supports Java Socket API, Posix C++ Socket API in Unix kind like Ubuntu. My trading system has some real time long calculations to do and a lot of maths for backtest. I am using a retail trading program Amibroker which is written in C++ and I run it in windows xp 32 bits, it will take me days to do one serious backtest with my G620 Sandy Bridge CPU and 3GB of ram. So for my trading system, I need to have 1. speed, 2. stability, 3. fast development I have done some research, C++ is fastest but I am not good at it and it takes much longer time to develop. Other than C++, Java in linux has the best speed. I also did some research for database and look like mysql has the best speed. Netbeans should be the most popular Java IDE now and seem like its GUI design can be as easy as Virtual Studio now. I am not sure if made by Netbeans would affect the speed and if its GUI design is really that good and easy to use. Ubuntu 64 bits LTS has good long term support, good community support, and stable. I will buy a new computer if I can create a good trading system for live trading and backtest. Very likely I will buy a I7 or I5 depends on if I7 can really have better speed for my case. Actually I mainly deal with C# in my jobs and I just knew java but not good at java. What would you guys recommend? Any better solution? This will be a big project and very likely life long project for me so I seriously do research including asking you guys before I start and focus on what I should, thanks!
You totally look at the wrong point. The langauge is quite irrelevant if you need a factor of 100 faster. Java, C++, C# - they are all in the same ballpark. Your problem is a crappy computer to start with. you need a system that can handle backtests and optimizations in a large set of computers and put the budget in to actually do that. The language will not help. And to learn to optimize your code. Our own backtest platform now uses a total of 54 cores (some are down) and I am getting a new system in 2 weeks with another 128 cores. Adding additional computers to further distribute the load. And with lots of optimizations in the infrastructure we got the backtest speed up by a factor of around 100 over the last months. Not the language, the code was the issue. Obviously "I do not really know how to program" will not help you doing right decisions.
Why do you keep posting crap like ... "I wanna build a trading system..." "I'm gonna a build a trading system ..." "I have the parts to build a trading system, but my system is slow as shit and I don't really know how to program..." "Which OS should I use ..." "Which language should I write in..." "Which broker's API should I use, and which language ..." Who cares.... It's getting pretty nauseating to see "j2ee" write a new post about the same old shit over and over and over... It really sounds like you're procrastinating. Rather than attacking the elephant in your room - actually doing the research and gaining the trading experience - you're wasting your time planning your "magic bullet" - the amazing tool that will do all your research and trading for you. Quit stalling! Go buy/lease some retail software package, explore trading ideas using that package, and manually trade signals from the ideas with merit. Then come back once you've figured out how to trade and/or are making money. In a few years, you'll actually be ready to build that magical trading system - because by then you'll know what you want and how to build it. Or you may "mature" and realize this pipe dream is really just a waste of time. This is a prime example of what happens when a techie attempts to start a business - they focus on everything except making money. Which, at the end of the day, is the only reason for starting/owning a business... Which are you? A wannabe trader or a wannabe tool builder?
Damn, totally overlooked that the original post was from this guy. Just did not read the name. He can not be helped. Totally incapable of learning and processing the information he gets here, it seems. Well, I am off - let the child play in his little sandbox.
How would you argue about this? http://reverseblade.blogspot.hk/2009/02/c-versus-c-versus-java-performance.html
Fastest is direct to exchange using FIX. Even faster if the software sending the messages is residing inside a machine inside the exchange.
Nope, FIX is dead slow. Fix + FAST is better, but to be fast you must use a FPGA to decode the feed. Alterantively you realize you are not going to win against HFT and so you are ok and design strats for a 5ms-10ms delay.