PDC 2008: Mesh Services Architecture and Concepts

Okay…I couldn’t live blog the whole thing…there’s a lot to take in…but here’s the gist. Mesh takes care of a LOT of complex stuff so you don’t have to be concerned with it.

3:00 Introduction:
       
Mesh is a concept whose time has come
        It takes the Windows platform to the next level
        Allows windows to take advantage of the cloud and the many devices a user has
        This talk mostly about Mesh Architecture
3:04 Agenda
       User Experiences
       Concepts
       Architecture
       Planning and Deployment
       Implementation
       Operations, Experiences, Lessons learned
       Future
3:05 Quick Demo Using Mesh
3:10 Key Capabilities
        Provisioned State (Accounts)
        User-data storage (including replication and synchronization)
        Communications and presence (P2P when possible…fall back to relay when not)
        Notifications (Single channel for asynchronous push)
3:13 Provisioning in the Mesh
        Identity is provisioned during web signup
        Device provisioned during software installation
        Device claim – relationship between user and device established
3:14 REST in the Mesh
        Common metadata for all resources (Uri, LastUpdateTime, Etc.)
        Resource lives within collection (scope for grouping resources and tracking changes)
        Both resources and collections are represented as feeds and items
        Links allow navigation from one resource to another.
3:17 Provisioned State Services
        Account Service is a “Facade” for accessing accounts, devices, and apps.
        The client provides a local cache.
3:18 Folder Synchronization in the Mesh
        Items in data feed represent files (files are the resource)
        A data feed mapping tells Mesh that a given device wants to synchronize with a given feed
        Permission scope defined on mesh object for sharing feeds
3:21 Understanding FeedSync in the Mesh
       Bi-directional replication provided through schema and semantics.
3:23 Getting really deep into how Mesh client is implemented using the Live Framework
3:27 Architecting the Mesh Services for Scalability
       Primary scenarios for scalability concerns
       Startup
       Resource change (initiator)
       Resource change notification (recipient)
       Content replication
          This can take advantage of P2P scenario
3:31 Scale-out strategy
       Partition by user, device, and mesh object
       Use soft state to minimize i/o load
       leverage http semantics for caching, change notification, and incremental state transfer
3:35 How to Partition resources
       Every resource has a stable canonical URI based on its resource identifier
       Mapping a resource id to its home DC requires a lookup in a replicated database
3:37 General Design Principles
       Use commodity hardware
       Partition for scaling out
       Loose coupling across roles
3:38 Deployment
        2 Dual Core sockets
        8 GB RAM
        2TB HDD
        Windows Server
        New machine come on line…fabric is copied to it and it joins the cloud
3:43 Scale Unit Layout

0 Comments

Leave a Reply

Your email address will not be published.