Peter Sutton
Harry Yudenfriend

zHyperLinks substantially improves VSAM application availability, including all of the system functions that depend on VSAM (e.g. catalog and HSM). This blog entry focuses on the availability improvements provided by zHyperLinks.

Background

In prior blog entries we introduced the concept of a new I/O technology for the DS8880 in support of the mainframe, called zHyperLinks:

On January 10, 2017, IBM issued a Statement of Direction for zHyperLink, the first new mainframe input/output (I/O) channel link technology since FICON. zHyperLinks is a new short-distance mainframe attach link designed for 10x lower latency than High Performance FICON (zHPF). This technology will be installable on all existing IBM DS8880 storage systems, and the hardware capability is already included in the DS8880 R8.2 delivered in December (model numbers 984, 985, 986 and 988). zHyperLink is intended to speed up DB2 for z/OS transaction processing and improve active log throughput.

Applications and Middleware that use VSAM Restart Faster with zHyperLink
VSAM applications that need to be restarted after a system failure can expect to see up to a 66% reduction in the time it takes to restart, without any application changes or additional CPU cost. This is possible because of the way VSAM recovers the dataset after a disaster. If a VSAM dataset does not go through normal z/OS CLOSE processing the metadata for the dataset will be discovered the next time a dataset OPEN is issued by an application. The VSAM access method needs to make sure that all the records that have been added to the dataset are included in the VSAM catalog. VSAM does this with its VERIFY function, https://www.ibm.com/support/knowledgecenter/en/SSLTBW_2.2.0/com.ibm.zos.v2r2.idai200/da6i2ver.htm ) by reading all the records that have been added since the last time the high water mark has been updated through the end of the dataset. This could be tens of millions of I/O operations. zHyperLink technology literally reduces this elapsed time by an order of magnitude allowing the applications to begin production work that much sooner.

Solving Scaling and Working off Backlogs
The lower I/O latencies provided by zHyperLinks help improve availability in a few additional ways. The faster I/O helps move work through the system leading to fewer tasks, shorter lock hold time and less latch contention. Middleware, such as DB2, is limited in transaction rates by its log throughput. The faster I/O provided by zHyperLinks improves log throughput, helping scalability.

Application and middleware failures can cause short term delays as the system reacts to recover from the problem, creating workload backlogs. The faster I/O that zHyperLinks provide allows the system to work off backlogs more quickly, thus improving availability.

Finally, the faster I/O gives clients more headroom for growth. Unpredictable workload spikes, possibly driven by mobile applications, can push the limits of what the system can do. Faster I/O allows the system to meet the challenge of work load spikes.

Conclusion
By far, the most critical attribute of enterprise class storage is availability. IBM mainframe clients choose z Systems for their most business critical and demanding workloads in order to achieve the best possible availability characteristics in the industry today. Such high availability is achieved through a complete stack design and integration from applications and middleware through the operating system, clustering technologies, channels, storage area network (SAN) and storage technologies. Despite the layers and layers of technologies that help achieve continuous availability some failures can still occur.

zHyperLinks dramatically improves VSAM application, including all of the system functions that depend on VSAM (e.g. catalog and HSM) by providing a 10x speed up in I/O.

Learn more about IBM zHyperLink and the IBM DS8880 family today.

Join The Discussion

Your email address will not be published. Required fields are marked *