http://msmvps.com/blogs/nunogodinho/archive/tags/AppFabric/.NET+Services/default.aspxTags: AppFabric, .NET ServicesenCommunityServer 2008.5 SP2 (Build: 40407.4157)http://msmvps.com/blogs/nunogodinho/archive/2009/11/19/pdc09-day-three-building-scalable-and-reliable-applications-with-windows-azure.aspxThu, 19 Nov 2009 17:22:41 GMTd67277c4-116b-43f1-b688-e9ef184ea916:1740803NunoGodinho0http://msmvps.com/blogs/nunogodinho/rsscomments.aspx?PostID=1740803http://msmvps.com/blogs/nunogodinho/commentapi.aspx?PostID=1740803http://msmvps.com/blogs/nunogodinho/archive/2009/11/19/pdc09-day-three-building-scalable-and-reliable-applications-with-windows-azure.aspx#comments<p><strong>Brad Calder – Director/Architect</strong></p> <p><strong>Data Building Blocks</strong></p> <ul> <li>Types of Storage in Windows Azure</li> <ul> <li>Volatile Storage</li> <ul> <li>Local Storage</li> <li>Caches (eg. AppFabricCache and MemCache)</li> </ul> <li>Persistent Storage</li> <ul> <li>Windows Azure Storage</li> <ul> <li>Blobs</li> <ul> <li>Provide interface for storing name files</li> </ul> <li>Tables</li> <ul> <li>structured storage.</li> </ul> <li>Queues</li> <ul> <li>reliable storage and delivery of messages</li> </ul> </ul> <li>SQL Azure</li> </ul> </ul> <li>Storage Account Performance at commercial Availability</li> <ul> <li>Capacity</li> <ul> <li>100TB</li> </ul> <li>Throughput</li> <ul> <li>Up to a few hundred megabytes</li> </ul> </ul> <li>Partitioning of Data Objects</li> <ul> <li>Load balancing is an internal concept of Windows Azure Storage</li> <ul> <li>Allows the system to automatically scale out access to your data to meet its peek traffic demands</li> </ul> <li>What matters to the application is the partitioning key used for objects</li> <ul> <li>All objects with the same partition key are always grouped</li> </ul> </ul> <li>Per Object/Partitioning Performance for Commercial Availability</li> <ul> <li>Throughput</li> <ul> <li>Single Queue and single table partition</li> <ul> <li>Up to 500 transactions per second</li> </ul> <li>Single Blob</li> <ul> <li>small reads/writes up to 30 MB/s</li> <li>large reads/writes up to 60 MB/s</li> </ul> </ul> </ul> </ul> <p><strong>Scaling Computation</strong></p> <ul> <li>Compute Service Model – What is describes?</li> <ul> <li>The topology of your service</li> <ul> <li>Types of roles and their binaries</li> <li>How the roles are connected</li> </ul> <li>Configuration of the service</li> <ul> <li>How many instances of each role type</li> <li>Application specific configuration settings</li> <li>How many update domains you need</li> </ul> </ul> <li>Best Practices</li> <ul> <li>Due to application failures, upgrades or hardware failure</li> <ul> <li>Use multiple instances</li> </ul> </ul> <li>Queue Workflow Concepts</li> <ul> <li>Windows Azure Queue Provides</li> <ul> <li>Guarantee delivery</li> <ol> <li>Worker dequeues Message and marks it as Invisible</li> <li>Worker deletes Message when finished processing</li> </ol> <ul> <li>Note: If worker role crashes, message becomes visible for another worker to process</li> </ul> <li>A message may be processed more than once</li> <li>Assume messages put into same queue can be processed in any order</li> </ul> <li>Best Practice</li> <ul> <li>Make work items idempotent</li> </ul> </ul> <li>Scaling Queue Throughput</li> <ul> <li>Batch Work Items into Blob</li> <ul> <li>Group together many work items into a blob</li> <li>Queue up pointer to the Blob for the message</li> </ul> <li>Use Multiple Queues</li> <ul> <li>Job Manager</li> <ul> <li>Responsible form managing the execution of the Queues</li> </ul> <li>Work Items</li> </ul> </ul> <li>Continuation for Long Running Work Items</li> <ul> <li>Want to continue on failover</li> <li>High level approach</li> <ul> <li>Bread work item into smaller and repeatable steps</li> <li>Record progress</li> </ul> </ul> </ul> <p><strong>Lifecycle Management</strong></p> <ul> <li>In-place Rolling updates</li> <ul> <li>Specify the number of upgrade domains in service model</li> <ul> <li>Breaks your roles evenly over the number of domains</li> </ul> <li>Rolling upgrade</li> <ul> <li>Walk each upgrade domain one at a time</li> <li>Upgrade just the roles in the current domain</li> </ul> <li>Benefits</li> <ul> <li>Minimizes availability loss</li> <ul> <li>Only one domain of roles are restarted at a time</li> </ul> <li>Allows local state to be uses across upgrades</li> <li>Caches application upgrade issues early</li> <ul> <li>Detect upgrade issues after first few domains</li> </ul> </ul> </ul> <li>Versioning with Rolling Updates</li> <ul> <li>Always assume you will have old and new running side by side in your service</li> <li>Version Everything</li> <ul> <li>Protocols, Schemas, Messages and everything else</li> </ul> </ul> <li>Windows Azure Tables Schema change</li> <ul> <li>Have a version property in each entity</li> <li>Types of Schema changes</li> <ul> <li>Adding a non-key property</li> <ul> <li>Perform similar step update process</li> <li>Update “IgnoreMissingProperties”</li> </ul> <li>Removing a non-key property</li> <ul> <li>Perform similar upgrade process</li> <li>Update “IgnoreMissingProperties”</li> </ul> <li>Changes in Partition Key or Row Key</li> </ul> <li>Most uses schema change is adding a non-key property</li> </ul> </ul><div style="clear:both;"></div><img src="http://msmvps.com/aggbug.aspx?PostID=1740803" width="1" height="1">MicrosoftArchitectureCloud ComputingWindows AzureWindows Azure Service Platform.NET ServicesAppFabric