Christian Kleinerman: Introduction to SQL Server Project Madison

The Discussion

• 

  Good Information, thanks a lot; great START video. Few questions/comments:

   

  1. Did you say Compute 1, 2...n will run its own SQL Server? If so, does that mean there are multiple databases copies?

   

  2. You mentioned, Madison == Parallel Data Processing of Data Warehouse SQL Queries. This is true for READ or READ + WRITE?

   

  3. Can you please share h/w configuration & the scalability matrix information you found during your testing, for us to see what numbers are we seeing in terms of performance (out of curiousity).

   

  4. Was looking for more example on "How Parallel Processing Works" in Madison, what you spoke was Datawarehouse (not sure why ). Was expecting more pictorial examples in this Demo on "Parallel Processing stuff", looking forward to see future videos

   

  5. You mentioned the server(s) will be in Bldg#35. Do outsiders (non-microSofties) will get a chance to see what those BIG hardware looks like (I doubt we will, still asking) on which you will show the Demo. If not, can Channel-9 please record those on video, so that we can see what it looks like.

   

  6. How easy its going to be, to DEBUG complicated query on "Parallel Processing Nodes". Are you providing any TOOLS for debug & radiators screens, which displys the thread(s) for the query on which the SQL will run, to see which thread is getting resource bogged

   

  7. Do we need only 64-Bit Windows OS & 64-bit SQL Server?

• 

  Very cool work you are doing there

  One question: can you have multiple controllers? I image that there can be a situation when a controller is being overloaded, so can there be multiple controllers dividing queries into the same set of computing units?

• 

  Could I use joins in this subset sql? Or is this something like bigtable, couchdb and amazon simpledb? It would be interesting to hear how the two different approaches to solving the parallel data problem works if it's not the same. 

• 

  Thanks Charles & Christian, this is a very helpful video. It answered a great deal of my questions.

   

  I can't wait to see this working, so I'm very much looking forward to the next video.

   

   

• 

  Right to my questions ...

   

  • How does Madison work? - Can I really 'scale-out' a database easily? - The video explains that very well thanks Christian!
  • Will I need to design my database schema's and write my queries to work sympathetically with Madison? - Will Madison mean changes to the syntax in LINQ and SQL? - It does mean that some SQL statement will need to evolve?
  • Does Madison come with it's own toolset ... can I still use the existing BI toolset that comes with SQL Server 2008 EE?
  • How does the 'Landing Zone' work? - Will I need to reengineer my ETL/SSIS queries to work with it?
  • How can I avoid 'Fork-Lift' upgrades?
  • Will Madison work in a heterogeneous BI stack? Will it play nice with SAP, Oracle & Cognos?
  • Madison software and hardware are linked together, so what does this mean to the support life-cycle? i.e. Microsoft typically support software for 8 years, HP warrantee their hardware for 5 years?
  • How quickly can Madison recover in DR scenario's such as Data-Centre outage or one of the servers fails?
  • SQL Server 2008 isn't support clustering across different subnets (however Windows 2008 is) does Madison still have the same issue?
  • Does Madison work with all the new SQL Server 2008 features such as Resource Govenor and Policy Manager?
  • How can I backup Madison? Will it work with DPM (System Centre Data Protection Manager)?

     

    So I've got afew still to answer before I can design an Architecture ... but I have Ian Giles and Mark Anderson from UK SQL Server/Madison team coming to see me tomorrow ! C9/Microsoft listens! - I'm stoked! Thanks Guys

• 

  Good introductiory video. Not too many technical details though. 

  Are compute node SQL Server instances running the same code as the coordinator? Doesn't sound like they need to.

  Is data auto partitioning going to be supported?

  How Madison compares to now Oracle's Exadata?

  What kind of storage (row oriented, column oriented) is used for compute nodes?

  Coordinator still seems like a potential bottleneck, if 150 compute nodes start streaming back to the corrdinator, on a poorly scoped query there is still a good chance to food it with data. Are there any provisions for scaling out the coordinator, or it's vertical scaling for now?

   

  Really looking forward to more videos on Madison (with a bt more details on internals ) .

   

• 

  We will go deeper. As the title suggests, this is an introduction.

  C