Performance monitoring and tuning - Revision history

JDamon: Minor clarification of MAXSPINS

2022-02-02T19:23:46Z

Minor clarification of MAXSPINS

← Older revision		Revision as of 19:23, 2 February 2022
Line 1,256:		Line 1,256:
	You can change the number of spins that can be taken when MPLOCK is not available before going into wait by adjusting the setting of the <var>MAXSPINS</var> parameter. </p>		You can change the number of spins that can be taken when MPLOCK is not available before going into wait by adjusting the setting of the <var>MAXSPINS</var> parameter. </p>
	<p>		<p>
	Rocket Software recommends that you test the setting of the MAXSPINS parameter at your site. Increase the value of <var>MAXSPINS</var>, watching for a drop in the values of the LKWAIT and LKPOST statistics. When the values of ~~the~~ statistics stop dropping, ~~set to~~ that ~~current~~ MAXSPINS ~~value~~.</p></li>		Rocket Software recommends that you test the setting of the MAXSPINS parameter at your site. Increase the value of <var>MAXSPINS</var>, watching for a drop in the values of the LKWAIT and LKPOST statistics. When the values of those statistics stop dropping, that's an indication that you've found a good setting for the MAXSPINS parameter.</p></li>

	<li>STWAIT and STPOST measure the number of WAIT and POST operations associated with MP subtask switching, and are, therefore, indicators of one component of MP overhead.</li>		<li>STWAIT and STPOST measure the number of WAIT and POST operations associated with MP subtask switching, and are, therefore, indicators of one component of MP overhead.</li>
	</ul>		</ul>

	====Fast logical page reads====		====Fast logical page reads====
	<p>		<p>

JAL: /* Loop */ add `s`

2018-07-25T18:54:14Z

Loop: add <var>s

← Older revision		Revision as of 18:54, 25 July 2018
Line 1,725:		Line 1,725:
	===Loop===		===Loop===
	<p>		<p>
	When a request has nested loops, the innermost loop might require maintask execution (it might contain a <var>DELETE FIELD</var> statement, for example). In this situation, the innermost FOR or <var>REPEAT</var> statement requires a task switch, if the outer loop does not contain any serial statements or <code>MP=NO</code> $functions.</p>		When a request has nested loops, the innermost loop might require maintask execution (it might contain a <var>DELETE FIELD</var> statement, for example). In this situation, the innermost <var>FOR</var> or <var>REPEAT</var> statement requires a task switch, if the outer loop does not contain any serial statements or <code>MP=NO</code> $functions.</p>
	<p>		<p>
	If the amount of work done for each iteration of the outer loop is small, then the decision to switch tasks each time the inner <var>FOR</var>/<var>REPEAT</var> statement is encountered is a poor one. In this case, the number of instructions executed by an offload subtask does not outweigh the cost of task switching. The programmer can override this decision by omitting the <var>LOOPS</var> keyword from the <var>MP OPTIONS</var> statement (or the equivalent bit from the <var>MPOPTS</var> parameter).</p>		If the amount of work done for each iteration of the outer loop is small, then the decision to switch tasks each time the inner <var>FOR</var> or <var>REPEAT</var> statement is encountered is a poor one. In this case, the number of instructions executed by an offload subtask does not outweigh the cost of task switching. The programmer can override this decision by omitting the <var>LOOPS</var> keyword from the <var>MP OPTIONS</var> statement (or the equivalent bit from the <var>MPOPTS</var> parameter).</p>

	===Calls===		===Calls===
	<p>		<p>

JAL: /* Optimization using MPOPTS and MP OPTIONS */ add section ID

2018-07-25T18:30:09Z

Optimization using MPOPTS and MP OPTIONS: add section ID

← Older revision		Revision as of 18:30, 25 July 2018
Line 1,637:		Line 1,637:
	All ON UNITs receive control on the maintask. The loop and CALL statements contained within ON UNITs are handled with the same logic as loop and CALL statements outside of ON UNITs.</p>		All ON UNITs receive control on the maintask. The loop and CALL statements contained within ON UNITs are handled with the same logic as loop and CALL statements outside of ON UNITs.</p>

	==Optimization using MPOPTS and MP OPTIONS==		==<b id="mpOptsStmt"></b>Optimization using MPOPTS and MP OPTIONS==
	<p>		<p>
	When the SOUL compiler analyzes the MP characteristics of a request, it must make decisions about how to handle the following situations: </p>		When the SOUL compiler analyzes the MP characteristics of a request, it must make decisions about how to handle the following situations: </p>

JAL: X'02' bit of XMEMOPT must be on to enable zIIP support

2017-02-01T21:44:28Z

X'02' bit of XMEMOPT must be on to enable zIIP support

← Older revision		Revision as of 21:44, 1 February 2017
Line 548:		Line 548:
	</ul>		</ul>
	<p>		<p>
	Additionally, values in the <var>[[SCHDOPT parameter\|SCHDOPT]]</var> ~~parameter~~ support zIIP processing, as described below.</p>		Additionally, values in the <var>[[XMEMOPT parameter\|XMEMOPT]]</var> and <var>[[SCHDOPT parameter\|SCHDOPT]]</var> parameters support zIIP processing, as described below.</p>

			====XMEMOPT====
			<p>
			The X'02' bit of <var>XMEMOPT</var>, which is its default, must be on to enable zIIP support. </p>

	====SCHDOPT====		====SCHDOPT====
	<p>		<p>

JAL: /* IOS Branch Entry */ add ATB buffers

2017-02-01T20:19:40Z

IOS Branch Entry: add ATB buffers

← Older revision		Revision as of 20:19, 1 February 2017
Line 109:		Line 109:

	<li>Placement of all database I/O control blocks, hash cells, page fix lists, and buffers above the 16-megabyte line, significantly reducing the risk of control block overflow into CSA</li>		<li>Placement of all database I/O control blocks, hash cells, page fix lists, and buffers above the 16-megabyte line, significantly reducing the risk of control block overflow into CSA</li>

			<li>Placement of <var>[[NUMBUFG parameter\|NUMBUFG]]</var> buffer pool buffers above the bar (ATB). </li>
	</ul>		</ul>
	<p>		<p>

JAL: /* Activating IOS Branch Entry */ add ATB buffers

2017-02-01T20:16:55Z

Activating IOS Branch Entry: add ATB buffers

← Older revision		Revision as of 20:16, 1 February 2017
Line 117:		Line 117:
	To enable IOS Branch Entry, perform the following steps:</p>		To enable IOS Branch Entry, perform the following steps:</p>
	<ol>		<ol>
	<li>Ensure that M204XSVC, the <var class="product">Model 204</var> Cross-Memory Services ~~SVC~~, is linked into the Model 204 nucleus. This is the default as of Model 204 7.5. Otherwise, install M204XSVC as an SVC.</li>		<li>Ensure that M204XSVC, the module that provides <var class="product">Model 204</var> Cross-Memory Services, is linked into the Model 204 nucleus. This is the default as of Model 204 7.5. Otherwise, install M204XSVC as an SVC.</li>

	<li>Authorize <var class="product">Model 204</var> under APF. This is the default as of Model 204 7.5. </li>		<li>Authorize <var class="product">Model 204</var> under APF. This is the default as of Model 204 7.5. </li>

	<li>Specify the <var>[[XMEMOPT parameter\|XMEMOPT]]</var> parameter (include the X'02' bit) on the User 0 input line.		<li>Specify the <var>[[XMEMOPT parameter\|XMEMOPT]]</var> parameter (include the X'02' bit) on the User 0 input line.
	If M204XSVC ~~gets~~ installed as an SVC, also specify <var>[[XMEMSVC parameter\|XMEMSVC]]</var> on the User 0 input line. </li>		If M204XSVC is installed as an SVC, also specify <var>[[XMEMSVC parameter\|XMEMSVC]]</var> on the User 0 input line. </li>
	</ol>		</ol>
	<p>		<p>
	When IOS Branch Entry is enabled, control blocks, hash cells, and page fix lists are allocated above the 16-megabyte line.</p>		When IOS Branch Entry is enabled, control blocks, hash cells, and page fix lists are allocated above the 16-megabyte line, and <var>[[NUMBUFG parameter\|NUMBUFG]]</var> buffer pool buffers are built above the bar.</p>

	===Saving storage when using IOS Branch Entry===		===Saving storage when using IOS Branch Entry===

JAL: add `s`

2017-01-31T20:32:07Z

add <var>s

← Older revision		Revision as of 20:32, 31 January 2017
Line 407:		Line 407:
	The LFAREA is dynamic and shared by all regions. When z/OS experiences a shortage of 4K pages, it steals 1MB pages and subdivides them into 4K pages. If the <var class="product">Model 204</var> request for virtual storage with 1MB pages cannot be satisfied, then regular 4K pages will be used for this purpose, without generating a warning.</p>		The LFAREA is dynamic and shared by all regions. When z/OS experiences a shortage of 4K pages, it steals 1MB pages and subdivides them into 4K pages. If the <var class="product">Model 204</var> request for virtual storage with 1MB pages cannot be satisfied, then regular 4K pages will be used for this purpose, without generating a warning.</p>
	<p>		<p>
	When buffers are allocated with 1MB pages, <var class="product">Model 204</var> must use IOS Branch Entry (XMEMOPT=2). EXCP and EXCPVR do not support buffers allocated with 1MB pages above the bar.</p>		When buffers are allocated with 1MB pages, <var class="product">Model 204</var> must use IOS Branch Entry (<var>XMEMOPT</var> parameter X'02' bit). EXCP and EXCPVR do not support buffers allocated with 1MB pages above the bar.</p>

	===Parameters===		===Parameters===

JAL: /* Model 204 workload eligible for zIIP offload */ add section for Fast/Unload zIIP processing

2017-01-27T00:55:11Z

Model 204 workload eligible for zIIP offload: add section for Fast/Unload zIIP processing

← Older revision		Revision as of 00:55, 27 January 2017
Line 519:		Line 519:
	<p>		<p>
	While a very limited version of this capability is built-in for all customers of 7.6 and higher, the fully expanded zIIP support is known as <b>M204 HPO</b>, and it requires an additional license. For more information, [[Contacting Rocket Software Technical Support\|contact Rocket Software]].</p></blockquote>		While a very limited version of this capability is built-in for all customers of 7.6 and higher, the fully expanded zIIP support is known as <b>M204 HPO</b>, and it requires an additional license. For more information, [[Contacting Rocket Software Technical Support\|contact Rocket Software]].</p></blockquote>

			===Fast/Unload zIIP processing===
			The Model 204 add-on, [[Fast/Unload overview\|Fast/Unload]], offers support for zIIP processing beginning with Model 204 version 7.6.
			Available only to the [[Fast/Unload SOUL Interface\|Fast/Unload SOUL Interface (FUSI)]] in version 7.6, [[Fast/Unload overview#fuhpo\|Fast/Unload HPO]] (and Model 204 HPO) let
			FUSI requests invoke Fast/Unload code that is contained within the Model 204 load module and capable of using a zIIP processor.

			As of <var class="product">Model 204</var> version 7.7, Fast/Unload is entirely linked with the Model 204 Online load module. If run with a reference to the <var class="product">Model 204</var> load library, and licensed for Fast/Unload HPO and Model 204 HPO, a Fast/Unload job can take full advantage of Model 204 zIIP processing.

	===Model 204 parameters for zIIP support===		===Model 204 parameters for zIIP support===

JAL: add 204 nonswappable info

2017-01-26T00:10:08Z

add 204 nonswappable info

← Older revision		Revision as of 00:10, 26 January 2017
Line 13:		Line 13:

	===z/OS===		===z/OS===
	<p>		<ul>
	Run <var class="product">Model 204</var> as a nonswappable address space (~~see~~ the ~~Rocket~~ Model 204 ~~[[Upgrading~~ to ~~Model 204 7.5 or later on IBM z~~/~~OS\|installation page]] for IBM z/OS for details~~). </p>		<li>Run <var class="product">Model 204</var> as a nonswappable address space. This requires adding the following PPT (Program Properties Table) entry to the active SCHED<i>nn</i> member of SYS1.PARMLIB:
	<p>		<p class="code">PPT PGMNAME(ONLINE) /* Model 204 name */
	Run <var class="product">Model 204</var> in a performance group with a high, fixed dispatching priority rather than mean-time-to-wait or rotate (supported only in z/OS releases prior to z/OS 2.2).</p>		NOSWAP /* nonswappable */ </p>
	<p>		<p class="note"><b>Note:</b> Do not specify <code>KEY(0)</code>. During initialization, Model 204 processing checks
	Limit the paging rate of the <var class="product">Model 204</var> address space to no more than one to two pages per second through storage isolation or working-set-size protection. These controls are provided through the System Resource Manager (SRM) component of z/OS via the parameters PWSS and PPGRT. </p>		whether it is running in <code>KEY(0)</code>. If the key is 0 and the region is running APF-authorized,
			the key is switched to <code>KEY(8)</code> (user key), and a message is
			issued to the audit trail. If the region is not running APF-authorized, a
			message is issued to the audit trail, and the run is terminated. </p></li>

			<li>Run <var class="product">Model 204</var> in a performance group with a high, fixed dispatching priority rather than mean-time-to-wait or rotate (supported only in z/OS releases prior to z/OS 2.2). </li>

			<li>Limit the paging rate of the <var class="product">Model 204</var> address space to no more than one to two pages per second through storage isolation or working-set-size protection. These controls are provided through the System Resource Manager (SRM) component of z/OS via the parameters PWSS and PPGRT. </li>
			</ul>

	===z/VM===		===z/VM===

JAL: /* z/OS */ link repair

2017-01-24T19:38:56Z

z/OS: link repair

← Older revision		Revision as of 19:38, 24 January 2017
Line 14:		Line 14:
	===z/OS===		===z/OS===
	<p>		<p>
	Run <var class="product">Model 204</var> as a nonswappable address space (see the Rocket Model 204 [[~~:Category: Installation~~\|installation page]] for IBM z/OS for details). </p>		Run <var class="product">Model 204</var> as a nonswappable address space (see the Rocket Model 204 [[Upgrading to Model 204 7.5 or later on IBM z/OS\|installation page]] for IBM z/OS for details). </p>
	<p>		<p>
	Run <var class="product">Model 204</var> in a performance group with a high, fixed dispatching priority rather than mean-time-to-wait or rotate (supported only in z/OS releases prior to z/OS 2.2).</p>		Run <var class="product">Model 204</var> in a performance group with a high, fixed dispatching priority rather than mean-time-to-wait or rotate (supported only in z/OS releases prior to z/OS 2.2).</p>