This presentation discusses performance tuning for Visualforce and Apex. It covers best practices for SOQL, SOSL, batch Apex, record locking, and debugging tools. The presentation emphasizes following SOQL best practices, tuning Apex code, and optimizing Visualforce pages to improve performance. It also provides tips for using indexes, filtering data, avoiding cross-object queries, and handling record locking and sharing calculations.

1. Performance Tuning for Visualforce
and Apex
Stuart Bernstein
ISV Technical Evangelist
Thomas Harris
LMTS, CCE Production Support
John Tan
SMTS, CCE Technical Enablement

2. Safe Harbor
Safe harbor statement under the Private Securities Litigation Reform Act of 1995:
This presentation may contain forward-looking statements that involve risks, uncertainties, and assumptions. If any such uncertainties materialize or if
any of the assumptions proves incorrect, the results of salesforce.com, inc. could differ materially from the results expressed or implied by the forward-
looking statements we make. All statements other than statements of historical fact could be deemed forward-looking, including any projections of
product or service availability, subscriber growth, earnings, revenues, or other financial items and any statements regarding strategies or plans of
management for future operations, statements of belief, any statements concerning new, planned, or upgraded services or technology developments
and customer contracts or use of our services.
The risks and uncertainties referred to above include – but are not limited to – risks associated with developing and delivering new functionality for our
service, new products and services, our new business model, our past operating losses, possible fluctuations in our operating results and rate of
growth, interruptions or delays in our Web hosting, breach of our security measures, the outcome of intellectual property and other litigation, risks
associated with possible mergers and acquisitions, the immature market in which we operate, our relatively limited operating history, our ability to
expand, retain, and motivate our employees and manage our growth, new releases of our service and successful customer deployment, our limited
history reselling non-salesforce.com products, and utilization and selling to larger enterprise customers. Further information on potential factors that
could affect the financial results of salesforce.com, inc. is included in our annual report on Form 10-Q for the most recent fiscal quarter ended July 31,
2012. This documents and others containing important disclosures are available on the SEC Filings section of the Investor Information section of our
Web site.
Any unreleased services or features referenced in this or other presentations, press releases or public statements are not currently available and may
not be delivered on time or at all. Customers who purchase our services should make the purchase decisions based upon features that are currently
available. Salesforce.com, inc. assumes no obligation and does not intend to update these forward-looking statements.

3. Ask Questions on the Chatter Feed

4. Stuart Bernstein
ISV Technical Evangelist
Thomas Harris
LMTS, CCE Production Support
John Tan
SMTS, CCE Technical Enablement

5. Introduction
This session is designed to help with common Visualforce/Apex
performance issues targeted towards intermediate/advanced
Force.com developers.

6. Agenda
 SOQL/SOSL
 Batch Apex
 Record Lock Contention
 Visualforce
 Debugging Tools
 Q&A

7. You Need to Build on a Solid Foundation
 Follow SOQL Best practices
 Then tune Apex
 Then tune Visualforce

8. Indexes Are Vital for Well-performing SOQL
 Without an index the SOQL query will not perform well
 Standard indexes already exist on Record_Type_Id, Division,
Created_Date, System_Modstamp (Last_Modified_Date),
Name, Email (for Contacts and Leads), and the unique
Salesforce record ID (Foreign Keys)

9. Administrators Can Create a Custom Index

10. Need Additional Indexes?
 Salesforce support can create single and two-column custom indexes
on most fields (with the exception of multi-picklists, formula fields that
reference other objects, and Long Text Area/Rich Text Area)
 Open a case
- Sample Query
- Bind Variables
- Org ID and user ID who would run the query
- Suggested index - OPTIONAL
 If they create a case, then SAVE where the index was created

11. Use a Standard Index
Select name from account where
Id=‘00530000003hQLJ’
Performs quickly because ID is an standard index

12. Use a Custom Index
Select name from account where
SLA_Serial_Number__c=‘1234567’
Performs quickly because SLA_Serial_Number__c is a custom
index

13. Use a Custom Index
Select name from account where
SLA_Serial_Number__c=‘’
Doesn’t work because null prevents the index from being used.

14. Use a Custom Index
Select name from account where
SLA_Serial_Number__c IN :ListOfSerialNums
May work well
if no null variable in ListOfSerialNums
filter matches less than 10% of the total records up to 333,000 records
30% of the total records up to one million records if this was a standard index
(it’s not)

15. Stay Away From Cross Object References
Select name from contact where
Account.SLA_Serial_Number__c IN
:ListOfSerialNums
Solution
Copy the SLA_Serial_Number and External Id to the contact
OR query for the account IDs with the SLA Serial number first

16. Don’t Use NOT IN
Select name from account where SLA_Status__C NOT
IN (‘Valid’)
NOT IN also prevents the index from being used

17. One (best) Index Will Be Used
Select name from account where SLA_Status__C NOT
IN (‘Valid’) AND SLA_Serial_Number__c=‘1234567’
If one index is unavailable, then the best remaining index will be
used.

18. One (best) Index Will Be Used
Select Id from Product2 where For example in …
From Product2 where (NPA__c = :tmpVar1 and
NXX__c = :tmpVar2 and IsActive = true)
Instead use
Select Id from Product2 From where
CompoundField__c = NPA__c + NXX__c + ‘true’

19. Query Only the Necessary Fields
SELECT
AccountNumber,AccountSource,Active__c,AnnualRevenue,BillingCity,Billi
ngCountry,BillingPostalCode,BillingState,BillingStreet,CreatedById,Cr
eatedDate,CustomerPriority__c,Description,Fax,Id,Industry,IsDeleted,J
igsaw,JigsawCompanyId,LastActivityDate,LastModifiedById,LastModifiedD
ate,MasterRecordId,Name,NumberOfEmployees,NumberofLocations__c,OwnerI
d,Ownership,ParentId,Phone,Rating,ShippingCity,ShippingCountry,Shippi
ngPostalCode,ShippingState,ShippingStreet,Sic,SicDesc,Site,SLAExpirat
ionDate__c,SLA_Serial_Number__c,SLA__c,SystemModstamp,TickerSymbol,Ty
pe,UpsellOpportunity__c,Website FROM Account

20. SOSL Performance Best Practices
 SOSL queries for the data using a different technique but then
takes the data and adds the where clause using SOQL
- Eg FIND {foo} RETURNING Account (Id WHERE City='Rome') is
analogous to SELECT Id FROM Account WHERE ID IN (<list of Ids
from search>) AND City='Rome’
 Follow SOQL best practices for the where clause

21. Batch Apex
 Most common issue is performance of SOQL used in the start()
method.
 Avoid the use of non-selective filters in the start() method.
- Queries will perform significantly worse as more data is added.
 Don’t add a non-selective filter to filter out a few rows relative to
the overall size of the result set. Instead, use Apex to filter out
rows in the execute() method.
 Avoid errors with bulk saves.

22. Filter in Execute() Method
 Suppose you have 5M records in Account. You want to process the
80% of those records represented by this query:
- SELECT Id FROM Account WHERE Rating__c IN ('Excellent','Good','Moderate')
 Since there is no way to use a selective filter in a query for these
records, this query will not run quickly even if Rating__c is indexed.
 As an alternative, use this query in the start method:
- SELECT Id FROM Account
 Filter out the unwanted records in the execute method.

23. Filter in Execute() Method – Sample Code
global Database.QueryLocator<SObject> start(Database.BatchableContext bc) {
return Database.getQueryLocator('SELECT Id FROM Account');
}
global void execute(Database.BatchableContext bc, List<Account> scope) {
List<Account> actualScope = [SELECT Id, Name, Description FROM Account
WHERE Rating__c IN
('Excellent','Good','Moderate’)
AND Id IN :scope];
for(Account acc : actualScope) { /* do something */ }
}

24. Concurrent Batch Apex Jobs
 Running concurrent Batch Apex jobs on the same record set can
lead to contention.
- Concurrent jobs will likely perform DML that requires locks on the
same records.
 It’s possible but difficult to parallelize Batch Apex jobs.
 Try to improve performance by increasing the batch size and tuning
SOQL queries before attempting to run jobs in parallel.

25. Parent-Child Locking
 Insert of Contact requires locking the parent Account.
 Insert or Update of Event requires locking both the parent Contact and the
parent Account.
 Insert or Update of Task requires locking both the parent Contact and
parent Account, if the Task is marked as complete.
 Insert of Case requires locking the parent Contact and parent Account.
 In objects that are part of a master/detail relationship, updating a detail
record requires locking the parent if roll-up summary fields exist.

26. Sharing Calculations
 “Implicit” Sharing
- Record access built into Salesforce applications
 Parent Implicit Sharing (Account only, private sharing model)
- Access to Contact/Opportunity/Case = Read access to parent
account
- not used when sharing on the child is controlled by parent
- when user loses access to a child, need to check all other children
to see if we can delete the parent implicit share

27. Can’t Code Around Data Skew
Universal
Containers
 Sharing calculations
impacted by Data Skew
X
 300K unassigned contacts
under account ?
 Private sharing model
 Jane has access to a single
Bob
Smith
Fred
Green
Betty
Brown … 300K
Contact
contact
 Jane losses access to contact
 Should we delete Jane’s X
parent implicit share?

28. Effects of Data Skew
 Sharing calculations run longer
 Concurrent Request Limit error
- Synchronous Apex request starts counting against the limit waiting
> 5 sec for lock.
 UNABLE_TO_LOCK_ROW error
- Timed out waiting 10 seconds to acquire lock.
- Even if no errors occur, waiting for locks can significantly slow DML
operations.

29. Recommendations
 Consider a public read/write sharing model
- Parent will still be locked. No sharing calculations.
 10K children or less
 Prevent parent-child data skews
- Detect whether existing accounts already have "reached their limit"
of child records, and create new accounts.

30. Visualforce – Best Practices
 http://developer.force.com/dreamforce/11/session/Blazing-Fast-Visu
 View State
- Filter and Paginate
- Transient keyword
- Simplify component hierarchy
- JavaScript Remoting

31. Analyzing Performance
 Measure performance in the browser first to see if the
problem is client-side or server-side.
- Chrome Developer Tools and the Firebug plugin for Firefox will both
work for this.
 After confirming the problem is on the server, use the
Salesforce Developer Console to find areas you can tune.

32. Analyzing Performance
Example Visualforce Page: DebugPage

33. Analyzing Performance
 In Chrome, press F12 to open Developer Tools.
 Use the Network tab to measure performance.
 In this case it takes 8.74 seconds to get a response from the server when pressing the “Do BOTH Slow
Query and Callout” button.

34. Analyzing Performance – Developer Console
 Use the Developer Console to analyze server-side performance.
 Apex debug logs will be generated for every server request you perform while the
window is open.
 Several tools are available in the Developer Console to find performance hotspots.

35. Analyzing Performance – Developer Console

36. 1) Open the log for the slow
request.

37. 2) Open the Timeline tab.

38. 3) Identify the slow portion
and click on the bar. This
loads the relevant log lines
into the Execution Log.

39. 4) Click on a line in
the Execution Log to
show the stack trace
for the slow query on
the left.

40. Analyzing Performance – Log Levels
 The default log levels generate a huge amount of output, most of which
is unnecessary for debugging performance issues.
 This also makes the console slower.
 Try lowering the log levels so you can look for slow queries and callouts
first.

41. Analyzing Performance – Heap Dumps
 The Developer Console can also be used to create Apex
heaps that help in debugging.
 For example, you can use it look at the bind variables for a
query.

42. Analyzing Performance – Heap Dumps

43. Analyzing Performance – Heap Dumps
 The bind variables won’t
show up in the debug
console and may not be
easy to find.
 You can set a heap
dump breakpoint to get
the variable values at
the time the query is
run.

44. Analyzing Performance – Heap Dumps
Set a breakpoint on the
line where the query is
run.

45. Analyzing Performance – Heap Dumps
 The runtime values
of the variables can
be found in the
Symbols tab of the
heap dump viewer.

46. Analyzing Performance – Useful Tools
 Developer Console
 Firebug (http://getfirebug.com)
- Web development plugin for Firefox.
 Chrome Developer Tools
- Included with Chrome.
 Workbench (https://workbench.developerforce.com)
- Good for testing SOQL queries.
 Fiddler (http://www.fiddler2.com/fiddler2)
- Windows tool for debugging HTTP traffic, integrates well with IE and Firefox.

47. Working with Support
 SOQL performance
- Open a case with org Id, user Id, SOQL and bind variables.
- Can add custom single and two column indexes.
 Record locks
- Provide support the org ID and time frame where the issues
occurred.
- Provide the IDs of affected records if possible.

48. Useful Links
Architect Core Resources
 http://wiki.developerforce.com/page/Architect_Core_Resources
- Guide to Sharing Architecture
- Record Access: Under the Hood
- Best Practices for Deployments with Large Data Volumes
- Asynchronous Processing in Force.com
- VF Performance Apex Performance (in progress)

49. Stuart Bernstein Thomas Harris John Tan
ISV Technical Evangelist LMTS, CCE Production SMTS, CCE Technical
Support Enablement