Welcome to EMC Consulting Blogs Sign in | Join | Help

SSIS Junkie

SSDS developments

There have been some snippets of information trickling out of the Microsoft PDC this evening about SQL Server Data Services (SSDS). Namely:

  • That’s not its name anymore. Its now called SQL Data Services (SDS)
  • SDS is the data layer in Windows Azure
  • You can now do joins between different entities. Just by the very nature of cloud services you shouldn’t assume that this is the same as a regular join in a relational database but nonetheless this is a huge step forward.
  • Metrics such as:
    • how many containers/entities exist
    • what is total storage used by an authority/container
    • how much BLOB storage space
    • how many requests against a resource

will now be available and, best of all, they' will be available via the same RESTful interface that you use to get the data itself

  • User limits in the beta period have been increased

 

More interesting than all of that though is the tantalising information presented in a keynote slide:

SQL Services - Bob Muglia PDC SLide

[Matthew Roche has already commented about this by the way in his blog post Teaser from PDC: SQL Services and in fact I stole the picture above from his blog post but I know Matthew quite well so hopefully he won’t mind :) ]

Here we see a number of other planned offering from SQL Data Services including ETL and Data Mining. Bob Muglia (he who presented this slide) mentioned Analysis Services in his speech too. That at least answers the question posed in my blog post from 5th June earlier this year “Will we get a cloud-based aggregation engine?”. The answer, quite clearly, is YES. Excellent.

But what of ETL? Muglia brushed that one aside. He mentioned ETL in the cloud but SSIS was not mentioned specifically – I’m intrigued to see what develops here. Indeed I am wondering if there is even a requirement for ETL in the cloud; if these services are massively scalable as they are reputed to be then why do you need to move data between them? Time will tell.

And what of reference data? Muglia suggested that Microsoft would be making available pre-prepared datasets for our use; more intrigue indeed. Will this be information that is in the public domain or will we have to pay Microsoft for its usage? Will they be hosting information on behalf of third parties for public consumption? Again, all remains to be seen.

The reporting slide refers to hosted Reporting Services. Hmmm…that one doesn’t quite float my boat.

 

I’m rather excited my these developments. What are your opinions? I’m very interested to know.

-Jamie

P.S. Incidentally, look out on my blog later this week where I will (hopefully) be talking about some interesting work that I and a couple of colleagues have been doing using SQL Data Services.

Comments

 

Douglas Laudenschlager said:

Lots of comments today about PDC and the Microsoft executives discussing Windows Azure, the new platform

October 27, 2008 22:06
 

Santiago Cepas said:

Nice post Jamie, and great blog. The one stop reference for SSIS for me.

I believe ETL will always be necessary. Any enterprise BI solution worth its salt needs clean data that reflects the business processes of the organization, and follows the semantic rules the end users understand. How are you gonna do that without a proper ETL implementation?

Another issue is how willing will be your average organization to share its most valuable data with a third party.  I expect a very slow use of SDS, at least for the first few years.

October 28, 2008 11:29
 

Adam Tybor said:

Aggregation in the cloud is key.  I suspect we are going to see some kind of map reduce service that can easily be scaled out across the cluster to make aggregation really efficient.

I am hoping for something like http://research.google.com/archive/sawzall.html

@Santiago

ETL has always been the worst enterprise integration pattern in my book.  It is slow, brittle, complicated and costly.  Now we get services, storage, messages, and aggregation with unbounded resources to put BI where it belongs.  BI should be part of the enterprise system not an afterthought that hangs off the side of the system.

October 28, 2008 15:17
 

jamie.thomson said:

"BI should be part of the enterprise system not an afterthought that hangs off the side of the system."

Amen to that Adam.

October 28, 2008 15:28
 

Santiago Cepas said:

""BI should be part of the enterprise system not an afterthought that hangs off the side of the system."

I certainly agree as well. But I think is a bit utopic to expect most CIOs to share that thought.

It would be different if you have the luxury to start a enterprise IT system from scratch. But does that really happen in the messy real world?

October 28, 2008 15:51
New Comments to this post are disabled

This Blog

Syndication

Powered by Community Server (Personal Edition), by Telligent Systems