Greater capacity: Some dump data sets are quite large compared with other data sets generated by a system. An extended sequential data set can hold the largest SVC dumps, as much as 128 gigabytes.
Support for striping: Striping spreads sections, or stripes, of a data set across multiple volumes and uses independent paths, if available, to those volumes. The multiple volumes and independent paths accelerate sequential reading and writing of the data set, reducing the time during which dump I/O competes with production I/O.
It is recommended that the number of stripes match the number of volumes you use. This combination will yield the best performance because MVS™ data management allows random access to any record as though it appeared on a single volume. This is particularly useful during an IPCS analysis of a dump. The savings when loading the data set are real but smaller, the result of reducing the number of times end of volume processing comes into play.
In a striped data set, when the last volume receives a stripe, the next stripes are placed on the first volume, the second volume, the third, and so on to the last volume, then back to the first volume.
If you use more than six dozen stripes, the performance benefit of each additional stripe is much less than the performance benefit of adding the earlier stripes. Keep in mind that this is talking about the original data set definition. You can not add stripes to an existing striped data set. You must plan ahead. The faster processing speeds up moving dump data from relatively expensive data space storage to less expensive DASD.
Using DSNTYPE=LARGE: In z/OS® V1R7 and later releases, sequential data sets that use DSNTYPE=LARGE are allowable for SVC dumps when the systems that are involved in processing a DSNTYPE=LARGE data set are migrated to V1R7 prior to their use. If analysis using an earlier release is required, use z/OS V1R7 to transcribe the dump into a data set supported by the earlier release.
Placing dump data sets in cylinder-managed space: In z/OS V1R11 and later releases, extended format sequential data sets can be placed in either track-managed space or cylinder-managed space. SVC dump fully supports placement of dump data sets in cylinder-managed space.