Random giving the same value after too many usages

Rasmus Franke

Hello I'm coding on a simulator at the moment, what it is simulating is not very important. Each simulation returns a number in the 7000-10000 range, depending on certain random events. The simulation is run several thousand times to get a good average value.

The problem: after running around 1000 to 4500 simulations (it varies), all new random numbers are zero which makes all the following simulations identical. The current implementation looks like this:

in class Simulation:

public static Random Random = new Random();
.....
public double getRandom()
{
double random = Random.NextDouble();
Console.Write(String.Format("{0:0.00} ", random));
return random;
}

The getRandom method is called around 500 times per simulation, and several thousand simulations are run in 2 threads working in parallel. This problem didn't exist before i introduced multithreading in the program and I'm frankly stumped of how it can appear since Random is supposed to be thread safe when used as a static.

Any help would be greatly appreciated.

6 Replies

Know the answer? Post it — somebody with the same question will find it here.

Rasmus Franke17y ago
Turns out it was a thread problem afterall! Locking it solved the issue. It appears that I missunderstood the quoted line, it's only Random's own static members that are thread safe, not the whole object after I declare it static :).
Thanks for the help!
Sam Hobbs17y ago
If it were me, I would try using a lock on the object anyway, for diagnostic purposes. If it solved the problem then I would proceed on that path; it might be a bug but I doubt that is what is happening here.

Assuming that a lock is not needed, the other possibility is memory corruption. I would look to see if there is anything that could be clobbering memory. I would look to see if there is debugging functions I could use to catch alteration of memory outside of specified sections of code. Simple data breakpoints would not help since the object would be properly modified during normal use of the object. Debugging memory corruption is among the most difficult bugs to diagnose and the easiest way, if it can work, is to look at the code. It might not be possible to find the bug that way but if it is possible then it is likely the easiest way.
Rasmus Franke17y ago
"Any public static (Shared in Visual Basic) members of this type are thread safe" Quoted from http://msdn.microsoft.com/en-us/library/system.random(VS.71).aspx

I did kind of solve the problem, but in an ugly way. I simple made a check if the random numbers were starting to become zero, and in that case replaced the Random with a new Random, giving it a new seed. It works, but I'm still very curious about what could make behavior such as this, I can't imagine any thread problem that could make a random suddenly start throwing out only zeros.
Sam Hobbs17y ago
Using the same name for an instance of a class as the name of the class is a big mistake. Expect to get criticized for that. I don't think it is a problem here but most developers will say don't do it, even if it seems to work.

Are you sure that the Random class has built-in protection to prevent simultaneous access from multiple threads? I doubt it, but I don't know. It is typically our responsibility to do things such as that. If you are using the same object from multiple threads then I think you must provide the serialization to ensure the object is used by only one thread at a time. Do you know how that works?
Rasmus Franke17y ago
As seen in the code, Random is static, thus only created once, then used a lot of times. It's not an issue of seeding either I think, because that would make the same series of numbers come up. My problem is some 100000 correct random numbers followed suddenly and without apparent reason an infinity of zeros.
Sam Hobbs17y ago
Yes, line breaks are a problem and I reported it. Initially the response was that I am using an old version of IE but I am not. Eventually they realized it is a problem and they say they will fix it. The only real problem is that they just don't tell us that we can and must use HTML in our messages. So if you just put a
at the end of lines you will get a line break.

To answer your question, look at the Random Constructor documentation; it says "Random objects that are created in close succession by a call to the default constructor will have identical default seed values and, therefore, will produce identical sets of random numbers". Does that explain the results you are getting?