Interview Q and A for Oracle Performance Tuning Part - 6

301. Part of the sqlhc.sql script that inserts rows into the plan_table, but the documentation states that SQLHC Leaves "no database footprint". How can this be the case?

The SQLHC script performs a rollback at the end. Any additions made to the plan table (or any other object) are removed at the end of the script and the original state restored.

302. The SQLHC Documentation states that there is no configuration required, but it does not state how to create the 'plan_table'. How can I create the plan_table?

On most databases, the plan table is pre-created centrally and referenced by a public synonym using the 'catplan.sql' script from $ORACLE_HOME/rdbms/admin. If it is not there or accessible then you should check your installation logs for issues.

303. Is the SQLHC script the same for all versions?

Yes. The script is identical

304. Can SQLHC be used on Oracle 9i or below?

No. SQLHC was designed to work on 10g and above. It will not work on 9i or below because it uses SQL_ID to identify the SQL and this is not available in 9i or below.

305. What is the licensing requirement for SQLHC?

SQLHC requires no license and is FREE. Since SQLHC can use information from AWR reports if the Diagnostic or Tuning pack are installed, it asks if these packs are licensed at your site. If they are licensed, then answer "Yes" to this question to perform additional checks. If they are not licensed then Answer "No". You should check with your site license manager if you have doubts regarding the licensing arrangements with regard to Oracle packs. SQLHC does NOT check for a valid license

306. What is missing from the report if the Tuning Pack is not installed?

SQLHC uses packs to give it access to AWR (Automatic Workload Repository) based information (AWR information is accessible via the Diagnostic pack; the Tuning pack is a superset of the Diagnostic pack so either will provide access to this information). The AWR infrastructure is installed by default (because the database uses AWR information for internal purposes), the real question is whether you are licensed to use it or not.

Report sections that use AWR information are annotated with a 'captured by AWR' key phrase.These sections currently include: "Indexes Summary", "Historical SQL Statistics (DBA_HIST_SQLSTAT)" and "Historical Execution Plans",

If AWR information is not made available by the presence of a license for one of the aforementioned packs, used then the AWR related information in these sections will not be reported.

307. When using SQLHC, is it more beneficial to have AWR than not having AWR?

Yes. AWR information is a component of the diagnostic pack and is fundamental for performance diagnostics and interpretation. Generally, if you do not have AWR then you would need to use statspack but this is of limited use on later versions.

If you have enterprise editions it is very useful to have AWR to diagnose performance issues. Specifically for SQLHC, having AWR information provides more usable diagnostic output than without, so we would recommend it but it is not a pre-requisite.

308. Are the global health checks specific to any version of Oracle?

Some of the checks (such as those introduced in a particular version e.g. automatic statistics gathering) are specific to versions. Otherwise, no, the checks are a combination of best practices and indicators of invalid information, among other things.

309. Does SQLHC work with --Insert application here--

SQLHC is a simple Health Check script that uses SQL commands run against the data dictionary to produce a report. It is a good, fast way to check your SQL for issues independent of the app you are using. This means that it can work against any SELECT generated by any application software. We would recommend that you execute the script in SQL*Plus but you can probably execute it anywhere that can run SQL Statements.

310. Are there any limitations on using this script in a Database running EBusiness Suite?

No.

311. Is it practical to use this tool for SAP?

Yes.

312. Does this SQL Health Check tool take into consideration that the query is executed on an Exadata system/database ?

No.

313. Are there any Exadata specific settings?

Not at this time.

314. Can this tool be used with Data Guard?

Yes.

315. How we can integrate this script in to GridControl?

SQLHC is a simple SQL script that you run outside of Grid Control.

316. Are there specific checks related to Peoplesoft?

Not currently.

317. Are there specific checks related to Siebel?

Yes, there is at least one health check for Siebel included in SQLHC , Compatibility with Specific Database Features

318. Can SQLHC be used on a remote database?

SQLHC assumes that you have access to the database where you want to run the health check so connecting to a remote database will work fine.

319. Can SQLHC be used on SQL that accesses a remote database?

SQLHC assumes that you have access to the database where you want to run the health check so selects like :

select * from table@remote_database

will not work. With this select and selects like:

select * from local_table, remote_table@remote_database

we suggest using SQLTXPLAIN instead of SQLHC.

320. Does SQLHC work with XML Type and XML indexes?

No.

321. Does it work with encrypted columns?

Yes. When gathering statistics, DBMS_STATS does not treat these columns any differently from any others. Since SQLHC

simply reports on the health and validity of the statistics on these columns it will work fine.

322. Does SQLHC work with LOB columns?

Yes it does, but since we do not collate the same type statistics information for LOB columns as standard columns and

we don't directly access these columns with standard predicates, the same output as a standard column is not provided.

323. Does SQLHC advise for extended statistic or index reordering?

No, extended statistics and index reordering are not considered by SQLHC

324. Is SQLHC RAC aware?

Yes, it reports on information available across the entire RAC (ie. AWR/statistics) and also that which is specific to the

node (e.g. initialisation parameters etc)

325. Can I use SQLHC to analyze a call to a pl/sql procedure?

Not with this tool. We recommend that you use SQLT for more advanced capabilities.

326. Is there a way to run without having to execute the SQL (like EXPLAIN PLAN?)

For example: EXPLAIN PLAN FOR SELECT /* TARGET SQL */ * FROM dual;

No. Currently, if you run SQLHC against the SQL_ID from an 'EXPLAIN PLAN FOR' command, it runs and completes successfully. However it does not pickup the statistics for the tables in the query, nor does it produce the current explain plan it. For SQLHC to retrieve information the SQL has to have been executed at sometime on the instance and the SQL data still needs to be in memory or in the Automatic Workload repository. Essentially, SQLHC creates a report by extracting existing 'historical' data about a SQL statement. It does not execute and monitor a current statement.

327. Does SQLHC show multiple plans from the SQL Tuning Advisor?

No. Since SQLHC is designed as a lightweight health check it does not have the same functionality as a more comprehensive

tool such as SQLTXPLAIN or SQL Tuning Advisor itself. If you want to include SQL Tuning Advisor in a SQL report which includes Health Checks and much more then use SQLTXPLAIN instead (ensuring you have the appropriate license for the Advisor).

328. How does SQLHC differ from SQLT?

SQLT is far more advanced than SQLHC; while it incorporates similar checks to those done in SQLHC it is far more

extensive generally. SQLHC is simply a quick health check for a statement.

329. What is the difference between SQLHC and AWR and TKProf?

SQLHC, AWR and TKProf are completely different tools for different purposes.

SQLHC is a FREE script to check the health of the environment as applies to a single SQL statement See:

Document 1366133.1 SQL Tuning Health-Check Script

AWR is the Automatic Workload Repository and is a fundamental mechanism for the storage of general performance information in the database. Based on this stored information, reports are provided that can help diagnose and resolve issues related to the performance of the database and statements within it. 1363422.1 Automatic Workload Repository (AWR) Reports - Start Point

TKProf is a tool that formats trace files to provide collated information about a process as recorded in that trace file. See:

Document 760786.1 TKProf Interpretation (9i and above)

330. SQLHC mentions some issues with system statistics. Are system statistics supposed to be gathered?

Statistics pertaining to the actual system itself are useful so that Oracle can determine the likely load that the queries will

be running in an adjust plans accordingly. 470316.1 Using Actual System Statistics (Collected CPU and IO information)

331. Are statistics on dictionary objects supposed to be gathered?

Yes. Since the cost base optimizer relies on accurate statistics it is sensible to gather statistics on all tables and maintain

them. 457926.1 How to Gather Statistics on SYS Objects and 'Fixed' Objects?

Alternatively Dynamic sampling can be used. 336267.1 Optimizer Dynamic Sampling (OPTIMIZER_DYNAMIC_SAMPLING)

332. Are statistics required on temporary tables?

It is usually best to use dynamic sampling on Global Temporary tables, because of their volatile nature.

Alternatively you can fix their statistics to a 'general' value to preserve a particular 'acceptable' access path.

Document 130899.1 How to Set User-Defined Statistics Instead of RDBMS Statistics

See the following document for other suggestions : 356540.1 How to workaround issues with objects that have no statistics

333. Can the tool determine whether statistics were gathered using ANALYZE instead of DBMS_STATS

Although the statistics collected by ANALYZE may be different to DBMS_STATS the changes may not be sufficient to positively identify the use of ANALYZE nor is there any view that identifies ANALYZE usage or otherwise. The tool will rather identify issues with the statistics validity which is likely more important than the source for plan generation assuming the statistics are correct. the ANALYZE command should not be used in version 11 only DBMS_STATS should be used to collect statistics.

334. Since 11g Oracle automates the statistics collection, why is the script checking for statistics health?

Just because statistics collection is automated does not mean that large data loads or different collection intervals may

not have changed them. We also cannot guarantee that the automated statistics have not been disabled.

335. Does SQLHC use just the most recent statistics for health check?

Yes. SQLHC uses the statistics at the time that it is run. For a more comprehensive look at the historic statistics on

objects use SQLT.

336. If the statistics have been locked, then will SQLHC give the desired results?

Yes. SQLHC will give the same results whether the statistics are locked or not. Locking statistics does not prevent them

from being invalid.

337. SQLHC says that Value A is greater than Value B. This is impossible. How can this have occurred?

SQLHC reports on what it finds in the dictionary and if it finds issues it outputs a warning. It does not necessarily

comment on the origin of the problem. As for potential causes, then:

different statistics may have been collected at different times

indexes and tables may have been collected separately when row counts were different, (for example if you gathered index stats, then deleted rows from the table and then just gathered on the table then you would have a mismatch.

and so it is possible for meaningless statistics to exist. The tool checks for such issues so that plans are not affected.

338. If a collective parameter such as "OPTIMIZER_FEATURES_ENABLE" has been set, does SQLHC just report on that parameter change or on all the affected underlying parameters?

SQLHC just reports on the top level parameter change. Each parameter is handled independently of each other.

Individual issues may generate more than one observation

339. Does SQLHC show all the checks done or just the results that are non-compliant?

SQLHC shows only the not compliant observations. All Statistics and plans are shown for the associated objects however.

340. Is the "Tables" and "Indexes Summary" for all tables or only for the tables affected by a particular SQL?

SQLHC reports on all the tables referenced in the query you have submitted to the tool.

341. Does SQLHC check the dictionary statistics against the actual values in the objects?

No. SQLHC just compares the dictionary statistics against each other and against known bets practices

342. Does the health check account for bind variables in any way?

Yes, but for a more detailed analysis of the binds / histograms on relevant columns etc then SQLTXPLAIN will provide

more details (SQLHC is designed to be very fast).

343. Will SQLHC tell you that the same SQL is used with different plans ?

Yes. It will show there are multiple plans but that is not the primary function of the tool. For that you would be better to use

SQLT. If you have an extremely large number of versions then refer to: 296377.1 Troubleshooting: High Version Count Issues

344. Can the health check script help me identify changes to the explain plan over a long period of time?

Yes, if AWR data is used.  With AWR, SQLHC could help in showing the plan changes from the one in memory compared to the one stored in the AWR repository. It will list all the execution plans of a query from the past (assuming they are still available in AWR)

345. Does SQLHC provide any advise to improve the performance or is it just a health check?

SQLHC does not provide any direct advice. However as a reult of implementing the suggested changes, performance may be improved. For specific SQL Tuning Advice, you should use the SQL Tuning Advisor. 262687.1 How to use the Sql Tuning Advisor.

346. Does SQLT provide SQL recommendations or does it only Provide the Execution plan?

SQLTXPLAIN provides far more than just the Execution plan. It generates a whole set of detailed information about a query and its execution and can also link to the tuning advisor to provide specific plan advice.

347. Is SQLHC able to identify disk I/O performance issues?

It is not currently designed to incorporate this kind of general checks. We welcome any additional health-checks that you may suggest. If any specific health-checks are needed and not covered by this script, then as long as the health-check can be produced with a SQL Command (leaving no database footprint ) then these can be implemented in future versions.

348. How is it best to use the information provided by SQLHC?

SQLHC does a number of check on the validity of various statistics and parameters with a brief explanation of why they may be a cause for concern. Typically you would be running SQLHC because you are concerned about the performance of a particular SQL statement. Since invalid or ill-advised setting may cause the optimizer to pick a sub-optimal access path it makes sense to rectify any potential problems found by the health check.

So, review the findings, assess their potential impact on the statement in question and implement fixes as appropriate.

349. Does SQLHC take into account session parameter changes as well as global ones?

Yes.

350. Could SQLHC not be integrated with advisor tools rather than having multiple tools?

SQLHC was deliberately created from SQLT so as to provide a very lightweight initial check. If you want a more

comprehensive report then use SQLT. See:

Document 215187.1 SQLT (SQLTXPLAIN) - Tool That Helps To Diagnose SQL Statements Performing Poorly

351. Application users are complaining about the slowness of the database, what should I do? /

If I want to quickly check for the cause of a performance issue with AWR, what are the principal things that I need to check?

Firstly, we would suggest that you examine the findings from an ADDM report from the time period and implement the suggestions.

Then look at the AWR report.

Very quickly, get some background information: take a look at the version to get an idea of what you are dealing with, then look at the Elapsed time and DB time to determine how much load the database is under.

Then look at the top wait events that are occurring on the system. If possible compare this to another report where the load was similar. Then look for a general area where most of the waits originate. For example, if the top 5 waits are:

library cache load lock

cursor: pin S wait on X

row cache lock

latch: row cache objects

library cache: mutex X

Based on this information, decide what the majority of the problems are related to. If you look at the names of the events you can get some clues: "library cache" and "cursor" are to do with parsing. "row cache" is to do with the in-memory cache of dictionary information. From this you can deduce that most of the time is being spent parsing and manipulating dictionary objects -now you can use this to target the next stages of your investigation. Now pick one of the waits and address it : pick the worst bottleneck to start with as this will give you the biggest benefit.

Once you have made changes, examine the effect on the system. Compare it to the previous period and the baseline. Repeat until acceptable performance has been achieved. For reference,Document 1359094.1 How to Use AWR reports to Diagnose Database Performance Issues

352. What is the principal event that causes performance issues?

There is no principal event applicable to all applications. Some applications may put more pressure on I/O while others may contend for CPU or for memory structures. Instead of focusing on individual events that may not be pertinent to your system, you need to focus on whatever is consuming time on the system. You want the system to spend most of it's time on the CPU without "wasting" CPU with poorly tuned SQL statements.

353. What does the AWR DB time represent?

DB Time is the time used by the database sessions and processes during the AWR report period.

354. Why does the DB time exceed elapsed time at the top of the report?

On modern, multiprocessor systems more than one database call may take place simultaneously. For example, if there are four, heavily loaded CPU's, then in one hour of elapsed time, there is an upper limit of 240 minutes of DB time possible (4 x 60 mins.).

355. Why does the sum of the percentages of DB time under "Time Model Statistics" equal more than 100?

In the time model statistics hierarchy, a child statistic may be counted under more than one parent.

356. How do you account for multiple CPUs in AWR reports

Near the top of the report, "CPUs" reflects the number of threads. "Cores" is the number of physical CPUs.

357. In the Load Profile, how do you interpret DB Time(s) of 63.8 (64) seconds VS DB CPU(s) of 1.8 seconds?

Near the top of the report, the AWR reports a number of statisitics and among these are:

Elapsed time (this is the Wall clock time)

DB Time (this is the total time used within the database within the Elapsed Time Period)

As an example, you might see:

Elapsed : 60.68 (mins)

DB Time : 3,856.09 (mins)

This means that for every minute of Elapsed time there is 63.6 minutes of work in done in the database (3,856.09 / 60.68). This figure is reported (in seconds) as DB Time(s) in the Load

Profile section alongside the amount of CPU used per second (DB CPU (s)):

DB Time(s): 63.6 (per second)

DB CPU (s): 1.8 (per second)

On a Multi CPU system, the total CPU Time available to you (assuming 100% CPU usage) is going to be : Number of CPUs / second.

In other words, if you have 8 CPUs then you could potentially use 8 seconds of CPU time per second. In this case DB CPU (s) : 1.8 (per second) is reporting that the system is using 1.8

seconds of CPU of the potential 8 seconds that it could be using (ie 1/4 of the CPU capacity). This means that 61.8 seconds/second of the DB time is not on the CPU and therefore is something else. As correlation, this should be reflected in the O/S CPU statistics from the period.

In terms of interpretation "DB Time(s): 63.6 (per second)" means that the system is heavily loaded, "DB CPU (s): 1.8 (per second)" means that most of the time is not CPU but something else.

358. How to obtain CPU consumption by user for each hour from AWR?

It is not possible to report an individual user's CPU usage using an AWR report.

359. What is the difference between the instance CPU and host CPU sections of the AWR report?

Host CPU information is coming directly from the operating system and is reflecting the CPU usage on the whole machine.

Instance CPU is reflecting the CPU as used by the instance itself.

360. What do waits for cpu+wait for cpu indicate?

These waits show the time spent in the CPU run queue

361. How do support interpret AWR? Does Oracle support have tools to interpret the results from an AWR?

No. We just interpret the output manually. It is good practice to start with an ADDM report then progress to AWR analysis as necessary. Document 1359094.1 FAQ: How to Use AWR reports to Diagnose Database Performance Issues

There is no substitute for experience however.

362. I see high number of waits for 'Event XYZ' in my AWR report. How should I proceed?

Start by looking at the Troubleshooting Assistant and selecting the "Troubleshooting Database Resource Contention" section:

Document 1543445.2 Troubleshooting Assistant: Troubleshooting Performance Issues

This allows you to identify content based on symptoms you are encountering.

You can also look at the Troubleshooting guide to identify potential problem causes:

Document 1377446.1 Troubleshooting Performance Issues

If these suggestions do not help you then you probably will need to ask in the Community or file a Service Request.

363. In the Top 5 Timed Foreground Events 'event XYZ' has high "time(s)" and "avg wait(ms)". What does that mean?

The Top 5 Timed Foreground Events is ordered by percentage of the total database time used. Something with a high percentage and a high time is likely to be using the most database resource. The "avg wait(ms)" is the average time for all the waits in the period. The average could be made up with everything close to the average, a few peaks and lots of lower values or a mixture. Whether an average is a high value or not is dependent on the operation. Without significant contention, you would expect latches to be waited on for less than 1 or 2 milliseconds.

If they are being held for long periods then that is likely to be causing significant problems. For their waits it may be possible to have longer waits and still have acceptable overall performance. Remember to keep any waits in context of the overall time spent and proportion of the total database time. A single 50 millisecond wait for a latch in an hour is unlikely to even register in the top 5. Don't focus on that at the expense of the 500 million 2 millisecond waits for the different latch that is taking all your resources!

Note that some waits for wait events will be expected based upon your activity. For example if you are running a large number of SQL statements heavily using Parallel Execution, then you will start to see Parallel Query events appearing in your Top 5 events simply because the management of the parallel query infrastructure takes up resource. It does not necessarily mean that there is a problem. However, if your users start to report performance degradation at this time then you might want to look at how efficiently these queries are running and whether it is unrealistic to expect the system to cope with the load being imposed on it at that time.

364. Where do we find the "event XYZ" in the AWR? It is not listed in the top waits.

If an event did not account for any "meaningful" amount of time then it might not show up at all. Remember AWR shows only what is significant to your system.

365. Sometimes, in the Instance Efficiency Percentages section of an AWR report, the "Soft Parse %" value is low but the database response is good. Is this something we should be concerned about and need to investigate ?

If you do not have a performance issue, then there is little to be gained from focusing on individual measures. The "Top 5 Timed Foreground Events" section is the area that is making the most difference to your performance. If you make a change to something that is not affecting the top waits significantly then you are unlikely to notice a large benefit. There will always be waits. If your users do not have an issue with the performance of the system then there is no problem.

366. What is meant "User calls" and "Parses" ?

Parse calls are requests to parse the SQL statements that have been presented to the database.

User calls are calls to the database by user sessions. Parse calls are a subset of the user calls.

367. What is SQL ordered by Parse Calls?

This lists SQL which has had the most parse calls (ie soft parses or hard parses)

368. How do I interpret the "SQL ordered by Physical Reads (UnOptimized)" Section?

1466035.1 How to Interpret the "SQL ordered by Physical Reads (UnOptimized)" Section in AWR Reports (11.2 onwards) For Smart Flash Cache Database

369. Why does the report not concentrate on cursor waits?

AWR looks at the whole system rather than at the cursor level. If you want to look at individual cursor waits you can use ASH reports or 10046 trace

370. If the number of executions is very small (for example one execution), does this mean that the sql needs to be tuned?

No, it just means the SQL statement was only executed once in the snapshot period. You should select SQL to be tuned based upon the amount of overall resource it is taking up on the system. For example, SQL with high buffer gets, CPU or I/O usage may be a candidate for tuning.

371. How do I interpret the "Buffer Cache Advisory" Section?

Document:754639.1 How to Read Buffer Cache Advisory Section in AWR and Statspack Reports

372. Is the "Av Rd (ms)" from disk or from a buffer?

"Av Rd(ms)" indicates the average time to read data from disk into a buffer.

373. Does Invalidations or Reloads in the SQL Area indicate the Cache is insufficient?

In most cases invalidations or reloads are an indication that sharing is not occurring efficiently within the application rather than there is a need for tuning of the shared pool memory areas.

If cursors are not shared then the memory will fill up with cursors and space will need to be reclaimed to fit in new cursors. Existing cursors may then be removed that then have to be reloaded the next time they are executed. Adding more space just means that more unshared cursors can be stored and does not relieve the pressure. Investigate how well things are being shared before looking to add more memory.

Document 62143.1 Understanding and Tuning the Shared Pool

Document 296377.1 Troubleshooting: High Version Count Issues

374. What are the meanings of the "Library Cache Activity" Namespaces

The library cache is divided into a number of different areas called namespaces that contain library cache objects dependent on the application profile you are using.

Taking a step back, assuming you have a performance issue, you should only be concerned with "Library Cache Activity" if you are seeing a high proportion of library cache type waits. If that is the case then you are primarily looking for reloads and invalidations.

Reloads are where an object is re-executed but is found to have been removed from the library cache for some reason (for example been aged out due to space pressure and lack of use). This means that it has to be re-loaded. This is inefficient as the cursor will need to be re-parsed when potentially the information could have already been available.

Invalidations are where a cursor has been made obsolete (or invalidated) because one of the underlying pieces of information that it was created from has been changed. For example, if statistics have been re-gathered then a plan based upon those old statistics needs to be re-generated because it cannot be relied upon.

The namespace these occur on can help you narrow down the sort of area in the code that is causing the problem. in the majority of cases issues will be in the "SQL AREA" Namespace. A problem with standard application SQL cursors is most likely to be in the "SQL AREA" Namespace. Problems with high reloads and invalidations in other namespaces are rare and may require input from support to diagnose the cause.

375. How should I deal with high waits for 'Buffer Busy Waits' on RAC?

Identify hot blocks from "GC Buffer Busy" waits section of the AWR reports

376. How do I interpret the "PGA Memory Advisory" Section?

Document:786554.1 How to Read PGA Memory Advisory Section in AWR and Statspack Reports

377. How do I interpret the "OS Statistics" Section?

Document:762526.1 How to Interpret the OS stats section of an AWR report

378. Why does my storage administrator tell me that the response time measurements reported by the storage device indicate that average read time for a file ("Av Rd(ms)") is much lower than what is reported in the AWR?

As with any other application, Oracle Database is not cognizant of what goes on within operating system code, once it submits the I/O request. All it can do is to measure the starting and ending times of the request. Whether, in addition to the actual seek, rotational delay, and transfer time, one or more of the following events consumes time is unknown to Oracle Database, as the OS does not report back to the application that level of detail:

1. Swap time needed to restore the swapped out process to memory.

2. Time during which the thread issuing the read request waits in the run queue for a CPU, after an OS interrupt has detected that the read request has completed and marks the thread as runnable.

3. I/O queuing. The output from "iostat" will display average queue length under the column labeled "avgqu-sz".

379. Can you use AWR reports to identify DB performance issue due to physical limitation of hardware or some problem with configuration /SQL ?

Start with the Top 5 waits as usual. You would expect I/O waits such as 'db file sequential read' and/or 'db file scattered read' to be high for an I/O problem. Note that I/O can be across the whole system or limited to single disks so you might see slow performance against some waits and not others depending on the activity.

If you see high numbers for I/O related wait events and high average read/write times in the following section:

IO Stats > Tablespace IO Stats

IO Stats > File IO Stats

then you might suspect some issue with the I/O subsystem. This could be because it is performing slowly or perhaps is being over stressed by forces external to the database or possibly by

excessive I/O from the database. If the I/O is coming from the database then you would expect to see SQL responsible for this in the "Top SQL" sections. Look at:

SQL Statistics > SQL ordered by Reads

and look for queries with excessive Physical Reads. This could be caused by a single massive execution or a huge number of smaller executions (or a combination).

380. In AWR Efficiency Percentages, "Parse CPU to Parse Elapsd %" we are consistently getting very low % (i.e. 0.3%) but every other Efficiency Percentages is >95%

Looking at efficiency percentages alone can lead to dangerous assumptions. Always start with the Top 5 waits and work from there. If there is no performance problem and you are running within your target service level, then does it matter that some ratio is not 'optimal'?

Parse elapsed time should be mostly CPU but, with things like dynamic sampling, it is possible that other waits will account for the time. If you have identified this as causing a problem then you should trace the session (use 10046 level 8) to see where the other 99.7% of parse time is being spent: Document 376442.1 Recommended Method for Obtaining 10046 trace for Tuning.