OracleAS Web Cache 10g 9'0'4 OW 2003

1
(No Transcript)
2
Howard Cohen Vice President Jesse
Anton Principal Product Manager Oracle
Corporation
3
Offload Your DatabaseA Web Caching Primer for
DBAs and Architects
4
IT Today
Isolated Applications Deployed on Islands of
Infrastructure

• Expensive hardware, software, and labor
• Inflexible architecture limits business
  responsiveness

5
Oracle Strategy for IT

• Enterprise
• Grid Computing
• Responsive
• Software Architecture

• Islands of Infrastructure
• Isolated Applications

6
Introducing Oracle 10g

• Responsive software infrastructure for grid
  computing

7
Oracle 10gLower Costs, Increase Flexibility

• Enterprise Grid Computing
• Standardize on low cost components
• Consolidate on shared resources
• Automate management operations
• Responsive Software Architecture
• Deliver applications as reusable services
• Integrate applications to leverage existing
  assets
• Optimize business processes

8
Oracle Application Server 10gIntegrated
Middleware Platform
Clustering Caching
Management Security
9
Oracle Application Server 10g Lower Costs,
Increase Flexibility
10
Performance Challenges Remain

• Dynamic applications are compute-intensive
• Repetitive content generation strains
  infrastructure
• IT planners asked to do more with less
• Traffic surges can cause delays and outages
• Planning for peak loads is cost-prohibitive
• Denial-of-service attacks are frequent
• Visibility into end-user service levels is poor
• Users want sub-second response times
• Existing measurement tools and services are
  insufficient

11
Web CacheA Virtualized Grid Resource

• What is Web Cache?
• Reverse proxy cache and compression engine
• Deployed between browser and HTTP server

• Make more effective use of low-cost, existing
  hardware
• Enables high hit-rate caching (which improves
  performance and scalability)
• Maintain quality of service with workload
  management
• Improves system reliability
• Manage the end-user experience
• Supports end-user performance monitoring and
  flexible deployment options

12
How Web Cache Works
Web Cache
ClientBrowser
Database
Web App Server
1
3
2
4
5
7
6

• Client sends HTTP request
• Web Cache responds immediately if cached object
  is available
• If object is not in cache, Web Cache requests
  object from Application Server
• Application Server generates response (may
  include Database queries)
• Application Server responds to Web Cache
• If response is cacheable, Web Cache retains a
  copy for subsequent requests
• Web Cache compresses page and responds to Client

13
Leveraged Acrossthe Oracle Platform
Oracle HTTP Server
Oracle Enterprise Manager
Portal
Web
Oracle Applications
Business Intelligence
Collaboration Suite
Cache
Wireless
Security Identity Management
JDeveloper
The Best Cache for Oracle Application Server
14
Key Featuresin Web Cache 10g
10g
15
Efficient Useof Low-Cost Hardware
IT Problem
Web Cache Solution

• Efficient use of low-cost, existing hardware
• Automatic compression
• Static content caching
• Dynamic content caching
• Invalidation
• Partial-page caching

• Dynamic, Web-based applications are
  compute-intensive, yet IT planners are asked to
  cut costs and do more with less

16
Automatic Compression

• Problem poor response times for end-users in
  remote locations or with low-bandwidth clients

Slow or Remote Clients

• Solution automatic GZIP compression
• All standard Web browsers support GZIP expansion
• HTML typically compresses down to 1/10 original
  size
• Compression level is tunable
• No application changes required
• Benefits
• Reduced network latency for dial-up, WAN and
  Wireless clients
• Bandwidth savings

17
Static Content Caching

• Problem repetitive requests for voluminous
  static content
• Answer static content caching
• Benefit get immediate boost by caching static
  files
• GIF, JPEG
• PDF, HTML
• CSS, JavaScript
• Applets, WAV, Shockwave, etc.
• Static content caching is easy
• No application changes required
• Most file types cached out of the box using
  default policies

18
Dynamic Content Caching

• Problem dynamically generated content strains
  infrastructure
• JSP, Servlet, ASP, PERL, PHP, etc. put added
  strain on mid-tier and database
• Solution policy-based caching with Web Cache
• Caching policies specified by administrator
  (rules) or by application (response headers)
• Benefits higher cache hit rates
• Offload content generation infrastructure
• Challenge dynamic content caching is harder
• How do you keep the cache consistent with the
  application?

19
Staying Freshwith Invalidation
Trigger / Programmatic
Programmatic
Web Cache
Database
Web App Server
ClientBrowsers
Manual or Scripted
Admin Console

• Consistency Management Mechanisms
• Expiration Policies (for predictable changes)
• Invalidation Messages (for unpredictable changes)
• Scope fine-grained or broad support for
  application-specified Search Keys
• Effect remove immediately or specify a refresh
  grace period
• Format XML over HTTP (Java and PL/SQL APIs
  provided)

20
Caching the Uncacheable

• Problem how to scale highly personalized
  content?
• Solution partial-page caching
• Key do this in a standard, easy-to-use way

21
Develop forPartial-Page Caching

• Define whole pages in terms of templates and
  fragments
• Specify caching policies for templates and
  fragments
• Web Cache assembles full pages from templates and
  fragments

• Industry standard markup language is Edge Side
  Includes (ESI and JESI)
• Web Cache is an ESI processor you do not need
  an edge network like Akamai to use ESI

22
Benefits of Partial-Page Caching

• Increase Content Sharing
• Reduce Space Requirement
• Reduce Cache Update Frequency
• Update when the template changes
• Update when a fragment changes
• Do NOT update when shared fragments have already
  been updated
• Result Increase Cache-Hit Ratio

23
Key Featuresin Web Cache 10g
10g
24
Workload Managementand Reliability
IT Problem
Web Cache Solution

• Policy-based workload management
• Load balancing and failover
• Throttling and surge protection
• QoS assurance
• Cache clustering

• Unexpected traffic surges can create unacceptable
  performance delays and even system outages

25
Load Balancing, Failover,and Connection Pooling
Web App Servers

• Problem
• Server capacities vary
• Servers fail

Database
ClientBrowser
OracleAS Web Cache

• Answer Load balancing and failover
• Policy-based (weighted) round-robin algorithm
• Cookies and session-IDs used to maintain server
  affinity for stateful applications
• Layer 7 status checking for failure/recovery
  detection
• Connection pooling for TCP connection reuse
• Benefits enable policy-based resource usage and
  mask server failures

26
Throttling andSurge Protection
Issue unpredictable traffic spikes that generate
cache misses
1
2

• If Web server capacity limit is reached
• Web Cache throttles requests until capacity frees
  up
• When queue is full, Web Cache starts sending
  apology pages
• Benefits
• buffer layer against denial-of-service attacks
  and other traffic surges
• maintain availability of content generation
  infrastructure

27
Assured Quality of Service

• Problem heavy invalidation overwhelms the
  application during refresh period
• Solution policy-based refresh heuristics
  determine refresh order
• Highest priority documents refreshed first
• Lowest priority refreshed last may be served
  stale
• Administrator and/or developer can specify grace
  period
• Benefit QoS maintained even after broad
  invalidation
• It is better to serve some stale content than no
  content at all

28
Cache Clustering

• Issues
• Cache capacity limited to individual machine
  resources
• Hot content lost if a cache fails or is taken
  offline

• Solution clustering on the Grid
• Virtualizes cache instances into a single logical
  cache spreads out load
• Automatic content provisioning
• partitions content space for increased cache
  capacity
• replicates popular content for performance and
  high availability
• Benefits improved scalability, availability and
  capacity of cache tier

29
Key Featuresin Web Cache 10g
10g
30
End-User Experience Management
IT Problem
Web Cache Solution

• End-user experience management
• Measure click-to-render response times for all
  users
• Analyze and report on end-user response time data
• Optimize end-user experience with caching and
  compression

• Visibility into end-user service levels is poor
• True end-user response times still a mystery

31
Existing Monitoring Tools Services are
Insufficient

• Data is misleading
• Your users dont sit in Keynotes data centers
• Client software, hardware, connection speed, and
  physical location vary from user to user
• Data is incomplete
• Only monitors some pages some of the time
• Does not track individual, named users
• Scope is limited
• Intranet applications often ignored

32
Click-to-RenderResponse Time Monitoring

• Measure complete end-user response times, from
  the time a user clicks on a link to the time the
  page fully renders in the users browser
• Measure individual URLs or sets of URLs,
  regardless of whether the URLs are cached
• Monitor experience of particular users (key
  customers or executives)
• Raw measurements are recorded in Web Cache access
  logs, then analyzed by Oracle Enterprise Manager
  using its Application Performance Management
  (APM) functionality

33
End-User Performance Monitoring

• Benefits
• Capture real end-user response times for entire
  application
• Pinpoint performance problems
• Improve response times using Web Cache

34
D E M O N S T R A T I O N
End-User Performance Monitoring
35
Key Featuresin Web Cache 10g
10g
36
Flexible Deployment Options
IT Problem
Web Cache Solution

• Flexible deployment options
• Departmental
• Enterprise data center Grid
• Branch office hierarchy
• Heterogeneous environments

• Performance pain points exist at the
  departmental, central and branch office levels
• Pain points span across vendor offerings

37
Flexible DeploymentOptions

• Departmental
• Deploy on same box as your Web server(s)
• Make use of existing hardware
• Enterprise Data Center Grid
• Dedicated boxes or same boxes
• Provision cache nodes as needed

38
Branch Office Hierarchy
EuropeanCentralOffice
Web Cache
Web Cache
JapaneseBranch
U.S. Branch
WebCache

• Distributed Cache Hierarchy
• Automatic Invalidation Propagation

Web Cache
WebCache
AustralianBranch
BrazilianBranch
Move cached content closer to end users
39
HeterogeneousEnvironments
Database
ApplicationServer
OracleAS Web Cache
ClientBrowser
CDNProvider
ContentManagement Systems Tools
Multi-Vendor Interoperability via HTTP, XML and
ESI
40
New in Release 9.0.4

• Security
• SSL Client Certs
• nCipher SSL acceleration
• Logging and Diagnostics
• Improved access logging and event logging
• Debugging enhancements
• Manageability
• OPMN support
• Dynamic config changes
• Workload Management
• Support for MSFT network load balancing
• Session binding in a cluster
• Invalidation
• Search keys
• Inline invalidation

41
Case Studies
42
Digital RiverTurning Cache into Cash

• Background
• Commerce Service Provider (CSP) hosting over
  32,000 sites
• Over 1.5M page views/day -- doubles in holiday
  season
• Built on Oracle Database (Solaris), Oracle
  Application Server (Linux)
• Using OracleAS Web Cache
• Saved over 3 million in hardware in first year
• Response times reduced from 3-15 seconds to 0.5-3
  seconds
• Quotes from Marty Boos, CIO
• Quote 1, Quote 2, Quote 3

43
Easy as One,Two, Three

• Install and begin caching static image content
• 1 hour, saw x results, result
• Add rules to cache dynamically generated pages
• 1 week (including testing), result 50 cache
  hit rates
• Minor app changes to cache dynamically generated
  pages
• 1 month, result 80 cache hit rates

• Now working on v4 of application (Java-based)
• Writing application to take advantage of
  partial-page caching

44
Pervasive Caching
45
Faster Response Timesfor Oracle.com
Deployed OracleAS Web Cache
Response timein seconds
Time
Data Provided by Keynote Systems
46
Higher Throughput for OracleStore
Pages Served per Second
1200
75
47
Summary

• IT Performance Problems
• Dynamic applications are compute-intensive
• Traffic surges can cause delays and outages
• Visibility into end-user service levels is poor
• Enterprise Grid Solutions from Oracle
• Web caching
• Workload management
• End-user experience management
• Benefits
• Cost savings make efficient use of low-cost
  hardware
• Reliability avoid infrastructure overload
• Intelligence gain insight into application
  performance

48
Roadmap

• Security
• Access control
• Caching on Disk
• Durability
• Integration with OracleAS Portal
• Partial-page caching
• Integration with Oracle Applications
• Advances in page compression
• Standards
• Updated JSP tag library

49
More Info
LearnTestimonials at www.oracle.comTechnical
Papers atotn.oracle.com/products/ias/web_cache/
DiscussWeb Cache and ESI Discussion Forum at
www.oracle.com/forums EvaluateSoftware and
Sample Code atotn.oracle.com/products/ias/web_cac
he/
50
(No Transcript)