Tuesday, December 17, 2013

Best Practices for operational excellence in SharePoint 2010 - Part 1

  1. Network adapters
    • Gigabit network adapters for all server roles
    • Dual network adapters in production environment for WFE and App servers. One for users and other for SQL Server communication.
    • Private network adapters for inter-server communications, for tasks like backups so that this traffic does not affect the overall farm performance.
    •  Consider VLANs to reduce network traffic under heavy load conditions
  2. Keep minimum network distance between WFE, App and DB servers
    • Maintain Latency under one millisecond
  3. Web servers and app servers
    • Separate system components into logical derives and user RAID for redundancy.
    • Allocate 200 GB of disk space for operating system and temporary files
    • Allocate 150 GB of disk space for logs.
    • Consider number of users, requests and feature required to identify the WFE servers required.
    • Consider HA and DR appropriately.
  4. Database Servers
    • The following databases should be kept separate from other databases:
      • TempDB
        • Size: Medium
        • Must be on a separate spindle from all other databases
      • Secure Store
        • Size: Small
        • Host on a separate database instance and limit the access to one administrator
      • Search Crawl
        • Size: Extra Large
        • Optimize for read
        • Host on a separate server from the Search Property database
      • Search Property
        • Size: Large
        • Optimize for write
        • Host on its own server
      • Usage
        • Size: Extra Large
        • Optimize for write
        • Must be on separate spindle.
        • This DB does not need high performance and will not effect farm performance.
    • The following databases should be stored in the same location as other databases.
      • Configuration DB
        • Size: Small
      • ReportServerTempDB
        • Size: Small
  5. DB server health
    1. Pre-grow all databases and logs
    2. Limit content DB to 200 GB
    3. Don't store more than 50 databases on a single physical instance of SQL server when use SQL server mirroring. 
    4. Defragment and rebuild indices if possible
    5. Performance Counters
      1. Network Wait Queue: 0 or 1
      2. Average Disk Queue Length(Latency): <5 ms
      3. Memory used: <70%
      4. Free disk space:>25%
      5. Buffer cache hit ratio: >=90%



No comments:

Post a Comment