19.06.2013 Views

DB2 UDB for z/OS Version 8 Performance Topics - IBM Redbooks

DB2 UDB for z/OS Version 8 Performance Topics - IBM Redbooks

DB2 UDB for z/OS Version 8 Performance Topics - IBM Redbooks

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

2.5.5 Fast log apply<br />

Once you migrate to <strong>DB2</strong> V8 CM, the new default <strong>for</strong> fast log apply (FLA) is 100 MB. This<br />

should have a minimal effect on DBM1 storage. FLA is used by the RECOVER utility, during<br />

restart of a failed <strong>DB2</strong>, or when RESTORE SYSTEM is used. The storage is allocated when<br />

needed and released when the FLA phase is complete.<br />

When multiple RECOVER utility jobs are run in parallel, <strong>DB2</strong> can take advantage of FLA <strong>for</strong><br />

up to 10 concurrent jobs, each utilizing 10 MB.<br />

The default value <strong>for</strong> FLA at restart is always 100 MB, even if you explicitly specify 0 in the<br />

DSNTIPL panel or in the DSNZPARM LOGAPSTG.<br />

During RESTORE SYSTEM, a new utility with <strong>DB2</strong> V8, the default value <strong>for</strong> FLA (if<br />

LOGAPSTG is not zero) is 500 MB.<br />

For more in<strong>for</strong>mation on FLA and its per<strong>for</strong>mance implications, refer to Chapter 4, “<strong>DB2</strong><br />

subsystem per<strong>for</strong>mance” on page 127 and Chapter 5, “Availability and capacity<br />

enhancements” on page 217.<br />

2.5.6 Multi-row operations with DRDA<br />

2.5.7 IRLM V2.2<br />

<strong>DB2</strong> V8 CM automatically exploits multi-row fetch when you use a remote client that uses<br />

block-fetching in a distributed application. The DDF address space uses multi-row fetch to<br />

build the blocks of data to be sent to the client. This implies you do not have to change the<br />

application to take advantage of this tremendous per<strong>for</strong>mance enhancement. We have<br />

experienced per<strong>for</strong>mance improvements in these client applications in <strong>DB2</strong> V8 CM.<br />

For more in<strong>for</strong>mation on the multi-row fetch operations with DRDA, refer to Chapter 7,<br />

“Networking and e-business” on page 281.<br />

The IRLM address space also changes with <strong>DB2</strong> V8 and supports 64-bit virtual addressing.<br />

The IRLM is able to support more concurrent locks. PC=YES is now en<strong>for</strong>ced when you<br />

migrate to <strong>DB2</strong> V8. The ECSA usage (PC=NO) capability can still be specified, but is no<br />

longer used by V8. There<strong>for</strong>e, the ECSA previously used <strong>for</strong> IRLM, if you used PC=NO, could<br />

now be available <strong>for</strong> use by other applications.<br />

Important: If you used PC=NO, review your ECSA usage, since potentially there can be<br />

less of a need <strong>for</strong> ECSA with IRLM V2.2.<br />

Real storage <strong>for</strong> IRLM locks increases significantly, since they have doubled in size. For more<br />

in<strong>for</strong>mation on the IRLM in 64-bit mode, refer to Chapter 4, “<strong>DB2</strong> subsystem per<strong>for</strong>mance” on<br />

page 127.<br />

2.5.8 Page-fixed buffer pools<br />

<strong>DB2</strong> V8 introduces a new buffer pool parameter called PGFIX which allows you to choose to<br />

fix buffer pools in real storage. Be<strong>for</strong>e reading or writing a page, <strong>DB2</strong> must fix the page in real<br />

storage in order to per<strong>for</strong>m the I/O. In <strong>DB2</strong> V8, with this new buffer pool option, you can<br />

decide to permanently page fix an entire buffer pool. The option to page fix a buffer pool is<br />

available in CM, but is not automatically enabled.<br />

Chapter 2. Per<strong>for</strong>mance overview 21

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!