You can tune VSAM NSR to provide better sequential performance by requesting additional data buffers using BUFND. The performance difference depends on the degree of tuning in your applications. If you use the default number of buffers, the difference in performance is much less pronounced than if you have optimized buffer use to transfer entire cylinders of data in a single I/O.

For sequential processing, the way that locks are held and released is predictable. Locks are obtained and released in sequence through the data set as the job progresses.

Sequential update applications that use large buffers can have the greatest performance difference when moved to DFSMStvs, which caches data in the coupling facility. DFSMStvs does not attempt to read ahead for sequential access. Unless a record is found in the SMSVSAM buffer pool, each VSAM request results in the transfer of one control interval. This is a lower degree of buffering than you can do for sequential nonshared access to VSAM data. So, the elapsed time of sequential updates increases because the data transfer is less efficient.