Add an additional TODO comment for the global dataset directory information.

git-svn-id: https://hyracks.googlecode.com/svn/branches/fullstack_hyracks_result_distribution@2482 123451ca-8445-de46-9d55-352943316053
diff --git a/hyracks/hyracks-control/hyracks-control-cc/src/main/java/edu/uci/ics/hyracks/control/cc/dataset/DatasetDirectoryService.java b/hyracks/hyracks-control/hyracks-control-cc/src/main/java/edu/uci/ics/hyracks/control/cc/dataset/DatasetDirectoryService.java
index ba9dfab..8aab024 100644
--- a/hyracks/hyracks-control/hyracks-control-cc/src/main/java/edu/uci/ics/hyracks/control/cc/dataset/DatasetDirectoryService.java
+++ b/hyracks/hyracks-control/hyracks-control-cc/src/main/java/edu/uci/ics/hyracks/control/cc/dataset/DatasetDirectoryService.java
@@ -23,6 +23,12 @@
 import edu.uci.ics.hyracks.api.exceptions.HyracksDataException;
 import edu.uci.ics.hyracks.api.job.JobId;
 
+/**
+ * TODO: The potential perils of this global dataset directory service implementation is that, the jobs location information
+ * is never evicted from the memory and the memory useage grows as the number of jobs in the system grows. What we should
+ * possibly do is, add an API call for the client to say that it received everything it has to for the job (after it receives
+ * all the results) completely. Then we can just get rid of the location information for that job.
+ */
 public class DatasetDirectoryService implements IDatasetDirectoryService {
     private final Map<JobId, NetworkAddress[]> jobPartitionLocationsMap;