Microsoft Canadian Colleges Exec Briefing – Data Center Management

by | February 2, 2011

Rick Bakken, Senior Director, Data Center Evangelism, Microsoft Corporation

Delivering Infrastructure to the Cloud

Data Center Innovation – asked the question “Why are we building giant refrigerators for our data centers?”  In the Chicago data center, there are two floors.  Top floor is standard raised floor and the bottom floor is container based servers.  Made a decision to run the data centers hotter.  Microsoft buys 5% of the servers bought annually around the world.

Timeline – PC era, Portal era, Online App era, Web Services era, Cloud Computing era –>

  • opened first data center in 1989
  • launched in 1994-95
  • added Trustworthy Computing in 2002
  • added Security development lifecycle in 2004
  • Windows Live launched in 2008
  • Windows Azure cloud service launched in 2010

Microsoft runs on a huge global scale 24 x 7 with over 200 cloud services running today (2011).  Microsoft has more than 10 and less than 100 data centers world wide.  The Network Architecture is really the key to cloud based computing scalability.  Geo-redundancy is a feature of this style of network architecture.  This is a combination of few High Density data centers and large numbers of edge nodes.  Microsoft Global Foundation Services works from layer 0 to 4 of the OSI model.  They have some serious process and procedures to bring a low cost data center service to market.  Microsoft is investing a huge amount of capital in the Global Network and particularly Dark Fibre.

Sustainability Evolution

  • Generation 1 – 1989 – 2005 – co-location with server based architecture – $25M per MW
  • Generation 2 – 2007 – density with rack based architecture – $17M per MW
  • Generation 3 – 2008 – containers and pods architecture
  • Generation 4 – 2010+ – modular with ITPAC (pre-assebmled components) architecture – $4 to 8M per MW
  • Generation 5 – ? – remove all moving parts from the data center servers
  • Generation 6 – ? – completely green data center – all components fully recyclable

Server Hardware Design Considerations

  • the data center is the server
  • performance/watt/dollar – PUE is no longer a useful measure need a new way to measure
  • drive change in the industry through strong partnerships
  • deliver value to online service partners through customized designs at the application layer
  • think about infrastructure refreshes using OpEx instead of CapEx

CTO “Journey to the Cloud’  2008 to 2020

  1. IT Workload Analysis – what is running where?
  2. Virtualized Data Center – efficiency gains
  3. Legacy Migration- cloud rehost, rewrite ancient code, reduce power
  4. Cloud – on/off premise utility model
  5. Data Sovereignty appliance – Azure Appliance refreshed on a OpEx cycle every 3 years

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.