We have a database at my work which outputs around 10k-20k log files a day and it has been running around the clock for the best part of a year. Excellent stuff! A year of no downtime and almost constant load is punishing enough for any machine.
The downside of this is that there are something in the order of 4m log files now stored in a directory on that computer. Each file only contains a timestamp and a SQL statement, so the're mere bytes in size. Maybe 1kb at the most.
Now, I'm in the process of writing an app which removes any file that is older than six months in that directory.
I wrote something fairly elegant which gets a fileinfo array of the directory using getfiles(), and this obviously results in a massive array of information for 4m files when I only need their fullname and the creationtime. Because the database will be running constantly on a 1gb machine (and the db takes up about 600mb of mem), it's not practical to create such a large array, as the memory will run out.
So... Is it possible to do this in small measures so that the memory usage would never be over, say 100mb?
Here is my current code which, when run on a test directory, took about 30 minutes to process 130,000 files, using about 45MB of RAM, and using almost all the CPU for explorer.exe for 20 minutes. Doing a calculation to scale that to 4m made me a little concered that it wasn't efficient enough:
DirectoryInfo dir = new DirectoryInfo("C:\\log\\test");
Cursor.Current = Cursors.WaitCursor;
FileInfo arrFiles = dir.GetFiles("*.txt");
//dateTimePicker1.Value is a DateTime value which is given by the user from the datetimepicker control (it's set to 'DateTime.Now.AddMonths(-6)').
if (arrFiles[i].CreationTime < dateTimePicker1.Value)