1. Module 1–Windows AppFabric Cache
Introduction and Architecture

2. Outline Why caching? Introduction to Windows Server AppFabric Cache
Overview of Cache features

3. Typical Web Architecture
Need to route users to same machine (i.e. sticky sessions)
Users
Each machine round trips for data
Some data might be expensive to retrieve
Cached data is typically stored in the memory of one (each) server
Web Tier
CPU and disk can get saturated due with an increase traffic
Database

4. Why use caching? Most applications end up needing
Robustness
Speed
Scalability
This is especially true in the web world
Need to scale can sneak up quickly
Almost any web application can benefit from caching
Services can benefit as well

5. Core caching concepts Caching Distributed Caching Expiration Eviction
Storing a copy of data closer to application logic, away from its source
Distributed Caching
A set of nodes in a farm that coordinate to create a unified view of a cache to a caching client
Expiration
When an object is removed from the cache because of staleness
Eviction
When an object is removed from the cache due to low memory conditions

6. What is the AppFabric Cache?
A Distributed cache for “data”
In-memory cache
Uses an explicit API
The “Data” can be
CLR Object
XML
Binary data
Optimized for the cache-aside pattern in this version
Programming against the cache is explicit
Changing the cache doesn’t update the original data store

7. What is Windows Server AppFabric?
A set of technologies and capabilities integrated into Windows Server and IIIS
Designed to make building, managing, and scaling IIS applications easier
Ships as part of Windows Server AppFabric 2010 Wave 1
Part of Windows Server
HA requires Windows Server 2008 Enterprise or DataCenter Editions.
Requires .NET 4.0
WORKFLOW HOSTING
MONITORING
CACHING
SERVICE HOSTING
ACCESS CONTROL
High Availability
Scale Out
Multi-Tenant
Management

8. AppFabric architecture
Cache runs on N nodes
Nodes that share configuration are part of a single Cache Cluster
Runs as a Windows Service (one service per server node)
Typically not the same servers as the client (web) tier
More memory is better
In-memory copy cache is optional setting (“Local Cache”)
All the cache nodes in a cluster are part of a fabric
Share configuration
Communicate
Cooperate
Same algorithms that support Windows Azure’s fabric

9. Stripped memory across server nodes
CacheServer#1
CacheServer#2
CacheServer#3
Cache
Cluster
obj#4
obj#3
obj#2
obj#1
Objects in memory

10. AppFabric Cache Users Web Tier Caching Tier Database
No stick load balancing needed – all servers have copy of cached data
Users
Easy access to cache cluster
Web Tier
Multiple machines means scale and potential for high-availability
More machines == more memory for cache objects
Caching Tier
Reduces load on database
Database

11. AppFabric Cache advantages
Native .NET API
Scalability
Performance
Typically linear as more nodes are added
LocalCache increases raw performance
High Availability
Distributed nature safe guards against cache client and server failures

12. Installation and configuration
Installs as part of Windows Server AppFabric
Not dependent on other AppFabric features (e.g. service and workflow management)
First node creates cluster configuration
Other nodes are configured to just join existing cluster
Nodes can be added to or removed from the cluster at any time – cluster automatically reconfigures itself

13. Cache Cluster configuration
Configuration abstracted behind a provider model
OOB providers
SQL Server Database
Network share (XML file)
Configuration Provider model open

14. Cache size configuration
You must configure “projected” size
1-5 nodes (small)
6-15 nodes (medium)
> 15 nodes (large)
Cache optimizes itself based on this setting
Setting not affected by actual number of nodes added to cluster
You can change this (and other) configuration settings on a stopped cluster

15. Remember – PowerShell can also be called from .NET Code!
Administration
PowerShell cmdlets are used to administer the cache cluster
Remember – PowerShell can also be called from .NET Code!

16. Cache creation Caches are named constructs Partition design based on
Multiple items can live in one named cache
Key for each item must be unique
Partition design based on
High availability requirements
Grouping based on logic
New-Cache cmdlet
Each named Cache can have its own settings for
Availability
Expiration
Eviction

17. Cache Clients Rich API Can be .NET 4.0 or .NET 3.5 SP1
Centers around the DataCache type
Can be .NET 4.0 or .NET 3.5 SP1
WCF NetTcpBinding used under client API
Can be configured in code or config
Simple configuration – server name and port
Can configure N servers
Configuring max N is recommended
Client is “smart”
Requests routes to cache node where data lives automatically
Routing table updated on a constant basis by cache cluster
Client essentially becomes part of the cluster fabric
Local cache configured at client
Local cache automatically updated (no guaranteed speed)

18. Programming the cache //create DataCacheFactory based on config file
var dcf = new DataCacheFactory();
//get the cache named "TestCache"
var cache = dcf.GetCache("TestCache");
//Add an item called "Test" - throws if exists
cache.Add("Test", new Data { TheData = "Test" });
//Put an item called "Test2" - overwrites if exists
cache.Add("Test2", new Data {TheData = "Test2" });
//Get "Test3" - add if not in cache (cache-aside)
var test3 = cache.Get("Test3") as Data;
if (test3 == null) {
test3 = new Data {TheData = "Test3" };
cache.Add("Test3", test3); }

19. Client API Clients can specific named Regions for items
Regions are stored on one cache host only
Created implicitly if not specified
Cache Items can be tagged
Tagging enables items to be retrieved without specific knowledge of the item
Client can register for notification of cache events
Bulk API available as well

20. Security Domain Based Security Option On by default
Domain Account / Local Account based Authentication
Only authorized servers can join the cluster
Only authorized clients can connect to the cluster
Transport Level Security
Turn on/off Signing or Encryption
Can turn off Cache Security
Security always has a performance implication
Use Firewalls, IPSec, VLANs to protect cache
grant-cacheallowedclientaccount  MyDomain\Machine1$
grant-cacheallowedclientaccount  MyDomain\Jon

21. Management PowerShell enables configuration and some monitoring
Logging
Cache participates in ETW tracing
Logging available on both client and server
Rich set of performance monitor counters for monitoring both cache and host performance

22. ASP.NET Session Provider
Custom Session Provider included with Cache
Just a configuration change
enables session state to be cached across a cluster
You may want to enable HA on the cache used
<sessionState mode="Custom" customProvider="SessionStoreProvider">
<providers>
<add name="SessionStoreProvider" type="Microsoft.ApplicationServer.Caching.DataCacheSessionStoreProvider, Microsoft.ApplicationServer.Caching.Client, Version= , Culture=neutral, PublicKeyToken=31bf3856ad364e35" cacheName="BlueYonderSession"/>
</providers>
</sessionState>

23. Caching Patterns
It is extremely useful to classify the type of data you are going to cache as either:
Resource, Activity, or Reference Data
Classification will inform what features to use
Use the correct feature for the correct type of data
Will help you to design the property number of caches
i.e. No need for HA for Reference data
More on this later

24. Summary
Windows Server AppFabric Cache is a distributed application cache
Caching is important to application scalability
Explicit caching API available to .NET 3.5 SP1 and .NET 4.0
ASP.NET Session provider an implicit way to take advantage of the Cache’s capabilities

25. Questions?