As MACRO completes construction and we concentrate more on analysis we will find it desirable to continue to produce DSTs. The reasons for this are clear. In the spring of 1993 MACRO produced about 225 MB of raw event data per day. It took about 1.5 hours to read in one day's data and perform the central tracking on an IBM RS6000-320. As we enter the new run, we expect approximately 50%more streamer tube and scintillator data, and probably 2 new trigger systems. Therefore, we should expect that the amount of data and the time it will take to analyze that data will only increase.
DSTs(Data Summary Tapes) offer a nice solution to this problem. With DSTs we reconstruct some things that everyone needs (such as the tracking and perhaps the ERP reconstructions) once and then write this information out to disk. This way the user need not waste CPU time computing things like the tracking every time the analysis is run. Instead, he or she can concentrate on just doing the interesting physics calculations.
DSTs also allow us to be selective in the data that we look at. Often we don't use all of the raw data that exists on the data tape. For example, someone doing muon astronomy seldom if ever looks at the QTP information. Therefore, it is nice to ``throw this data away''. In other words, we only write the data to the disk that we think is useful. This helps us immediately because now we can store many more runs on a disk. Also, the analysis jobs run much faster, since the amount of I/O the program has to do is reduced. As an example, with a prototype DST which we produced with our FARFALLA package, one day's worth of data only occupied 10 MB, and a simple analysis on this data took only 1 minute.
Of course there is a problem in a scheme like this: no one DST can satisfy everyone. If you decide to throw away the QTP information no one who wants to look at the QTP can use the DST you produced. So in practice, what usually happens is that different people write different DSTs, each one tailored to the people who will use them for their analysis. So a very useful thing to have are tools which allow you to easily design and create a custom DST.
Even though groups and individuals may have separate DSTs(which may have even been produced on different sorts of computers) it is nice to have a file format which is transportable and standardized. This way a DST produced on a DEC ULTRIX machine can be successfully read on an IBM RS6000. Also, a DST produced by a group should be readable by any other group regardless of the information contained in it.