Are You Cashing In on Caching

1
Are You Cashing In on Caching?

• An in-depth look at the caching methods
• available within ColdFusion applications
• Tyson Vanek
• Senior Technical Consultant
• Webapper Services, LLC

2
Presentation Overview

• Caching overview and basic concepts
• Identifying candidates for caching
• ColdFusions template caching options
• ColdFusions native query caching methods
• ColdFusions native page/content caching methods
• Alternatives to ColdFusions native caching
  methods

3
About Tyson Vanek

• Independent Consultant
• Currently engaged full-time with Webapper
  Services, LLC
• 12 years of ColdFusion experience
• Began working with ColdFusion v1.0
• Using ColdFusion on every project since July,
  1995
• Former employee of Allaire/Macromedia Consulting
  Services Division
• Noteworthy Accomplishments
• Certified ColdFusion Developer
• Contributing author - Inside ColdFusion MX
• Contributing author - Adobes ColdFusion
  Developer Center
• 2 year author of BrainBench ColdFusion
  certification exams (5 MX6)
• Noteworthy Clients
• Footlocker
• Eastbay
• Lockheed Martin
• Abbott Laboratories
• University of Virginia
• Adelphia Communications
• Williams-Sonoma

4

Caching Basics
5
What is Caching?

• A way of saving data after it has been originally
  computed so that subsequent requests for the same
  data are executed more quickly than the original
  call by retrieving the saved data instead of
  computing real-time

6
What Can Be Cached?

• CFML Templates at the Server level
• Trusted Cache
• Class files
• Data Objects
• Query record sets
• Complex variables (e.g. structures, arrays,
  objects)
• Simple variables (e.g. strings, text blocks)
• Web Service Invocation Responses
• CFCs
• Rendered HTML Content
• Entire pages
• HTML blocks

7
Why Should I Be Caching?

• Alleviates load from the server(s)
• Makes the server faster and healthier
• Allows the server to support more load
• Decreases page load time for the end user
• Improves the users experience
• Allows the user to accomplish tasks more quickly

8
Defining Good Cache Candidates

• Items that can be described by any/all of
• the following characteristics
• Global to all users
• Used throughout the application and accessed
  frequently
• Not subject to frequent change
• Not subject to a high number of variants
• Limited number of input arguments and distinct
  values
• Takes a significant amount of time to render in
  real-time when not cached (application
  performance bottlenecks)
• Complex nested CFML logic
• Database I/O
• CFC invocation
• CFHTTP calls
• Web Service calls

9
Methods for Identifying Application Performance
Bottlenecks

• The Old Fashioned Way
• Examination of ColdFusion log files
• Pages logged when ColdFusion Administrator Log
  slow pages taking longer than X seconds option
  is enabled.
• Examination of ColdFusion debugging output
• Displays execution times of templates, includes,
  custom tags, and CFC methods
• Displayed in detail when ColdFusion Admin Report
  Execution Times option is enabled with summary
  mode selected
• Query execution times
• Cumulative template execution time
• CFML Language
• getTickCount()
• ltcfloggt
• ltcftracegt
• Using New Server Analytics Tools
• SeeFusion
• FusionReactor
• CFMX 8 Administrator

10
Important Questions to Ask Yourself When
Evaluating a Caching Candidate

• How often do changes take place that will affect
  and/or invalidate the cached data?
• Is up-to-the-second data absolutely critical?
• How many cache variants are possible based on the
  variable input arguments?
• What is the average count/size/length of each
  cache variant?

11
What is a Cache Variant?

• A cache variant is a single set of data or
  content derived from the application code and
  based on a single combination of values for any
  variable input arguments
• Examples
• Variant 1ltcfset lastNameLetter Agtltcfquery
  nameqUsers datasourcemyDSNgt SELECT FROM
  tblUser WHERE left(lastName,1)
  lastNameLetterlt/cfquerygt
• Variant 2ltcfset lastNameLetter Zgtltcfquery
  nameqUsers datasourcemyDSNgt SELECT FROM
  tblUse WHERE left(lastName,1)
  lastNameLetterlt/cfquerygt

12
Calculating Cache Variants

• Query Example 1
• SQL StatementSELECT FROM tblUserWHERE userid
  useridAND active active
• Variable Input Arguments
• userid (1,000 distinct values) active (2
  distinct values)
• Variants
• 1,000 x 2 2,0002,000 potential query
  variants exist

• Query Example 2
• SQL StatementSELECT FROM tblUserWHERE
  department departmentAND age ageAND
  gender genderAND active active
• Variable Input Arguments
• department (50 distinct values)age (100
  distinct values)gender (2 distinct
  values)active (2 distinct values)
• Variants
• 50 x 100 x 2 x 2 20,00020,000 potential
  query variants exist

But how many likely variants exist?
13
Why Calculating the Cache Variants is Important

• Each variant represents an instance of
  data/content that could potentially be cached
• The more items in the cache, the larger the cache
  becomes in size and complexity
• Size may affect storage decisions
• Complexity may affect access/seek time to cached
  items
• With larger caches, attention must be given to
  server resources (e.g. disk space, memory
  allocation, etc)
• The size of the cache may play a role in
  determining the best location to store the cache

14
A Simple Measure of the Effects of Caching

• Example 1 No Caching
• Request execution time 500ms
• Average requests per minute 60
• Calculation500ms x 60 30s30 seconds of
  server timeallocated per minute30s x 60m
  1,800s1,800 seconds of server time allocated
  per hour

• Example 2 - Caching
• First request execution time 500ms
• Cache TTL 60s
• Cached request execution time 20ms
• Average requests per minute 60
• Calculation(500ms x 1) (20ms x 59)
  1.7s1.7 seconds of server timeallocated per
  minute1.7s x 60m 102s102 seconds of server
  timeallocated per hour
• Result 94.3 Improvement!

15

ColdFusion ServerTemplate Caching Techniques
16
Understanding ColdFusions Template Request
Workflow
17
Template Caching with Trusted Cached

• ColdFusion Administrator Trusted Cache setting
• Bypasses disk checks for more recent changes to
  the CFML template, caching the CFML template into
  RAM
• Impacts the STARTUP, PARSING, COMPILING,
  LOADING, SHUTDOWN timer located in ColdFusion
  debugging output
• Typically only enabled on production servers
  where templates change infrequently or with
  scheduled releases
• Server-wide setting with no application-specific
  control
• Templates are cached based on file date/time
  stamps
• May cause unexpected behavior when templates are
  deployed via FTP or source control promotion of a
  previous version of the template

18
Template Caching with Trusted Cached (cont.)

• Clearing cached items
• Individually
• Execute cached template while Trusted Cache
  setting is disabled in CF Administrator
• Re-enable the Trusted Cache setting once the
  template has been executed
• Entire Cache
• Restart ColdFusion Application Server (YIKES!)
• ColdFusion Administrator Clear Template Cache
  Now button (CFMX 7)
• Admin API call (may be programmatically invoked)
• Event Gateway/Admin API Tool developed by Brian
  Szoszorek (CFMX 7)
• http//www.adobe.com/devnet/coldfusion/articles/ca
  checlear.html

19
The Save Class Files Option

• New option added with ColdFusion MX 6.1
• Found in the CF Admin Caching settings
• When enabled, generates and saves a .class file
  for each CFML template executed
• Files are stored in WEB-INF/cfclasses/ directory
• File-based storage means that cache persists even
  if ColdFusion is restarted
• Server-wide setting with no application-specific
  control

20
Why You Should Consider NOT Enabling It

• Feature was automatically enabled (behind the
  scenes) in CFMX 6.0 due to performance issues
  with the compiler (javac/jikes) that was being
  used at the time
• Compiler in CFMX 6.1 was replaced and improved
• CFMX 6.1 compiler is much faster, converting CFML
  files into bytecode
• Improved performance of the 6.1 compiler has
  more or less eliminated the need for .class file
  generation/caching
• Creates additional overhead
• Potential for thousands of files in the
  /cfclasses/ directory
• ColdFusion will continue to check the date/time
  stamp of the .class file with every call to the
  corresponding .cfm template to ensure that the
  cached .class file is still valid (unless Trusted
  Cache is enabled or the template is not present
  in the Template Cache)
• Significant file/directory IO (500 ms or more)
• Most customers report little or no difference
  with this option enabled or disabled on CFMX 6.1
  or higher
• If theres little or no performance to be gained,
  why bother with the option at all?
• Enabling this option adds more decision points to
  the ColdFusion Template Request Workflow,
  inherently slowing the request process.
• Does it even work?

21

ColdFusion Query Caching
22
Native Query Caching with ColdFusion

• Adding the cachedWithin or cachedAfter argument
  to the CFQUERY tag
• Places the query result set into ColdFusions
  application server memory for the defined length
  of time
• Queries cached using cachedWithin or cachedAfter
  are placed in the same cache, so theres a
  possibility of re-use
• ColdFusion Administrator Maximum Number of
  Cached Queries setting
• Controls the physical number of queries that may
  be stored in the cache at any given time
• Once this threshold is reached, least recently
  used items in the cache will be dropped to make
  room for new items
• Setting value to 0
• In earlier versions of ColdFusion, setting this
  value to 0 resulted in unlimited query
  caching (extremely dangerous)
• In CFMX 6.1 (6,1,0,83762) and CFMX 7, setting
  this value to 0 results in disabling query
  caching
• Technically, allows a single cached query
• Important point about query caching variants
• Every cached query is driven by a combination of
  the following items
• Query Name
• SQL Statement (including formatting, tabs,
  spaces, etc.)
• Datasource
• Username/Password
• Any input arguments to the SQL statement

23
Query Caching Examples

• Example 1 - Using cachedWithin
• Query without caching
• ltcfquery
• nameqTop10NewsStories
• datasourcemyDSN
• gt
• select top 10
• from tblNews
• order by storyDate desc
• lt/cfquerygt
• Query with caching (15 minute cache)
• ltcfquery
• nameqTop10NewsStories
• datasourcemyDSN
• cachedWithincreateTimeSpan(0,0,15,0)
• gt
• select top 10

• Example 2 - Using cachedAfter
• Query without caching
• ltcfquery
• nameqTop10Videos_January2007
• datasourcemyDSN
• gt
• select top 10 sum(countView), videoName
• from tblVideoView
• where viewDate gt 01/01/2007
• and viewDate lt 02/01/2007
• order by sum(countView) desc
• lt/cfquerygt
• Query with caching (after 2/1/2007)
• ltcfquery
• nameqTop10Videos_January2007
• datasourcemyDSN
• cachedAftercreateDateTime(2007,2,1,0,0,0)
  

24
Clearing the Query Cache

• In Part
• Execute query with cachedWithincreateTimeSpan(0
  ,0,0,0) or cachedAftercreateDateTime(2100,1,1
  ,0,0,0)
• Only clears the single cache variant being
  executed
• Does not clear all variants of the same cached
  query
• In Whole
• ltcfobjectcache actioncleargt
• Clears all cached queries for the entire
  server/instance
• Not application specific
• ColdFusions ServiceFactory purgeQueryCache()
  method
• ltcfset factory createObject(java,
  coldfusion.server.ServiceFactory)gt
• ltcfset datasourceService factory.datasourceServi
  cegt
• ltcfset datasourceService.purgeQueryCache()gt

25
Pros and Cons of ColdFusions Native Query Caching

• Pros
• Its built-in to ColdFusion and ready to use
• Cached queries are placed in memory for fastest
  possible access time
• ColdFusions debug output will automatically
  display whether a query was executed real-time or
  if the results were pulled from the query cache
• Cons
• No control over where the cached queries are
  stored or how they are organized
• Maximum Number of Cached Queries setting is
  server-wide, not application-specific
• Difficult to clear items from the cache
• Difficult to clear multiple items from the cache
  without clearing the entire cache
• Clearing items from the cache forces them to
  execute again
• Queries cannot be cached if they make use of
  ltcfqueryparamgt
• Query caching cannot be used with ltcfstoredprocgt
• Stored procedure calls can, however, be cached if
  called within a ltcfquerygt tag (using exec, call,
  etc.)
• No way of viewing/dumping the cache in whole or
  in part
• No way to control how ColdFusion manages the
  cache and cache pops
• No way to track the size/usage of the cache
• Query Cache is server/instance wide, not
  application specific
• Code from other applications may be using
  ltcfobjectcachegt or the purgeQueryCache(), thereby
  clearing the cache for the entire server/instance

26

Alternative Data Caching Methods
27
Alternative Methods for Caching Data

• Placing data in shared scopes instead of local
  variables scope
• ltcfquery namescopeName.queryNamegt
• Server, Application, or Session scope
• Requires that you programmatically manage/flush
  cached items
• Items will automatically timeout if shared scope
  variable times out
• Can be applied to queries as well as any other
  ColdFusion data type
• Structures
• Arrays
• Objects
• Simple Values
• Must beware of application code that might be
  performing a duplicate() of shared scopes
• Placing these additional items in the shared
  scope for caching could negatively impact the
  performance of duplicating the scope
• No native support for clustered cache
  synchronization
• Currently designing and developing a cross-node
  in-memory cache synchronization system for
  clustered environments

28
Alternative Methods for Caching Data (cont.)

• Various available custom solutions
• Tyson Vaneks ltcf_extremeQueryCachegt custom tag
• http//www.adobe.com/devnet/coldfusion/articles/qu
  ery_cache.html
• Only supports caching of query record sets
• Supports caching to memory or disk
• Tyson Vaneks cacheItAll.cfc component
• http//www.coldfusionmaster.com/cacheItAll/
• Tyson Vaneks ltcf_accelerate2gt custom tag
• http//www.coldfusionmaster.com/accelerate2/
• Modified version of Brandon Purcells
  ltcf_accelerategt custom tag
• Supports caching of both simple and complex
  variables
• Supports caching to memory or to disk

29

ColdFusion Content Caching
30
Content Caching with ColdFusions ltCFCACHEgt Tag

• ltCFCACHEgt Tag
• Place tag at beginning of any page that should be
  cached
• Options include server-side or client-side
  caching (action argument)
• Creates rendered HTML output and saves file to
  disk on the server or client depending on the
  action indicated
• Cache files have .tmp extension
• Cache files are simply an HTML representation of
  the generated content
• First line of the .tmp cache file is an HTML
  comment that identifies the template and URL
  parameters used to generate the cached file
• This comment might cause browser validation
  issues
• Caches content based on the timeSpan argument,
  much like CFQUERYs cachedWithin argument.
• If no value is specified for timeSpan, the
  content will be cached indefinitely
• Cache would have to be cleared manually
• Cached files (.tmp) can be stored in a directory
  of your specification (using the cacheDirectory
  argument)
• Default location is cfroot\cache
• Automatically creates unique caches based on URL
  variables
• Support for unique caching by session (does it
  work?)

31
Pros and Cons of Caching with ltCFCACHEgt

• Pros
• Its built-in to ColdFusion and ready to use
• CFCACHE tag evaluates each unique URL parameter
  to ensure that different combinations result in
  different cached pages
• Supports both server-side and client-side caching
  for maximum efficiency
• Supports unique caching by session
• Cached entry is automatically expired if the CFML
  template is changed
• Cache can be cleared programmatically
• actionflush
• expireURL/view.cfm?id
• Cons
• Can only be used to cache an entire page, not
  smaller sections of HTML
• Limited control over how the cached pages are
  stored or how they are organized
• Difficult to view/dump the cache in whole or in
  part
• Can view/delete HTML .tmp file written to the
  cacheDirectory
• Files are named using an MD5 hash of the
  CGI.SCRIPT_NAME and CGI.QUERY_STRING
• Might make it difficult to identify the .tmp
  file(s) you are looking for
• Programmatic clearing of cached items can be
  challenging
• Variants are not created with consideration to
  CGI.PATH_INFO (SES URLs)
• No way to control how ColdFusion manages the
  cache and cache pops

32
Alternative Methods for Caching Content

• HTML Block caching with ltcfsavecontent
  variableyourVarNamegt
• Captures all content generated within ltcfoutputgt
  tags and saves it to the specified variable name
• Variable can also be placed in shared scope for
  caching purposes
• Management options similar to those discussed
  earlier regarding shared scope caching of complex
  variables
• Various available custom solutions
• ltcf_superCachegt
• http//www.cfprimer.com/download.cfm?ffFilesuperc
  ache.cfm
• ltcf_accelerategt
• http//www.bpurcell.org/blog/index.cfm?modeentry
  entry963
• ltcf_cacheOmaticgt
• http//www.devx.com/webdev/Article/27618
• ltcf_turboCachegt
• http//www.hotfusion.co.uk/TurboCache/index.htm
• ltcf_hyperCachegt
• http//www.pixl8.co.uk/index.cfm/pcms/site.product
  s.CF_Hypercache/
• Ray Camdens ScopeCache
• http//ray.camdenfamily.com/downloads/scopecache.z
  ip
• Andy Powells JohnnyCache
• http//johnnycache.riaforge.org/

33
Summary

• Discussed basic caching concepts and
  considerations
• Outlined methods for finding and qualifying good
  cache candidates
• Outlined methods for identifying application
  bottlenecks
• Reviewed ColdFusions built-in template caching
  settings and effects
• Reviewed ColdFusions built-in query caching
  mechanisms
• Presented alternative methods for caching queries
  and complex data types
• Reviewed ColdFusions built-in page caching
  mechanisms
• Presented alternative methods for caching entire
  pages or smaller sections of content

34

Questions?