Hi Anja,
Can you provide the details of your parallel setup? (i.e. the tc~bpem~arch~ear application properties). Also, how many Java instances are in your cluster and how many total server processes? The parallel archiving shouldn't exceed the number of server nodes.
How big is a single process instance in the XML archive (just manually archive a single instance and check the file system to get this information). In general, the performance is improved if you run smaller jobs since the size of each job is also smaller. True you will need to run more jobs in this case, but your throughput should actually be better if your jobs are currently too big.
regards, Nick