Dump Channel Program Changes
Abstract
In PTF UM35132 for APAR VM65989 z/VM 6.4 is changed to use a more efficient channel program during dumps. The change applies to PSW restart dumps and to snap dumps. In our measurements the change improved the speed of dump writing by about 260%.
Introduction
In PTF UM35132 for APAR VM65989 z/VM 6.4 is changed to use a more efficient channel program during dumps. Using a more efficient channel program improves the speed of dumps.
The previous channel program does this sequence of CCWs:
- Define extent, Locate record, TIC, Write record 1.
- Locate record, TIC, Write record 2.
- Locate record, TIC, Write record 3.
- Etc. up to 90 records per channel program.
The new channel program does this sequence of CCWs:
- Define extent, Locate record, TIC, Write record 1.
- Write record 2.
- Write record 3.
- Etc. up to 255 records per channel program.
IBM was able to eliminate the Locate Record CCW because as of z/VM 6.3, z/VM no longer supports CKD devices for CP-owned space, which includes the SPOOL space, which is where hard abend dumps are written. Also, CP-owned space can no longer contain filler records. These facts let IBM eliminate the Locate Record CCW before each Write CCW. The TICs were eliminated by moving the IDAWs used by the Write CCW to later in storage.
Method
A workload was devised to populate storage. A snap dump was then taken. After the snap dump was taken, the dump was loaded from spool onto minidisk. The loaded file was then analyzed to calculate how many 4 KB records were written during the dumping and how much time the dump took. From these two numbers the dump rate was calculated.
Dumps were done using z/VM 6.4 both without and with the PTF. LPARs of sizes 512 GB, 1024 GB, and 2048 GB were used.
Results and Discussion
| Table 1. Dump speeds, z/VM 6.4, without or with PTF UM35132. | ||||
| Base | Comparison | Delta | Percent Difference (%) | |
| 512 GB LPAR | ||||
| Run ID | SPE500R7 | SPE500R8 | ||
| Number of 4 KB records dumped | 6105107 | 6100811 | ||
| Elapsed time (sec) | 845.472 | 235.572 | ||
| Dump rate (records/sec) | 7221 | 25898 | 18677 | 259 |
| 1024 GB LPAR | ||||
| Run ID | SPE500R9 | SPE500RA | ||
| Number of 4 KB records dumped | 12181457 | 12181891 | ||
| Elapsed time (sec) | 1665.362 | 466.570 | ||
| Dump rate (records/sec) | 7315 | 26109 | 18794 | 257 |
| 2048 GB LPAR | ||||
| Run ID | SPE500RB | SPE500R6 | ||
| Number of 4 KB records dumped | 24347981 | 24247810 | ||
| Elapsed time (sec) | 3398.069 | 937.022 | ||
| Dump rate (records/sec) | 7165 | 25984 | 18819 | 263 |
| Notes: 2964-NC9, dedicated LPAR. DS8800 2107-941, four FICON Express chpids (one 16S at 8 Gb, three 8S at 4 Gb), via 2499 switch. z/VM 6.4 without or with UM35132. | ||||
Summary and Conclusions
In our measurements the PTF improved the dump rate by about 260%.