http://www.sqlservercentral.com/blogs/steve_jones/2008/11/19/pass-summit-2008-ted-kummert-keynote/

Printed 2014/08/31 02:33AM

PASS Summit 2008 - Ted Kummert Keynote

By Steve Jones, 2008/11/19

A film clip opening from MS, quotes and short sound bites from a variety of people from he SQL Server team and some shots of past conferences.

Ted is the corporate VP in charge of the Data and Storage Platform Division, which includes SQL Server 2008. Last year we were in the SS2K8 CTP process, having just completed the Nov CTP, which was a far cry from the final product, but this year we have a new version.

Ted reiterated the reason I like the Summit in Seattle. So many of the SQL Server team in Seattle and they come over. He had all of the people in the SQL Server team stand up, all of them standing up wearing the same shirts, a burgundy or purple polo. There's a lot of them.

The SQL Cat team is the exception, with bright green shirts to make them stand out. 

What's coming? 

The data platform division is looking to all kinds of data and all platforms, from mobile and desktop devices to the data center to the cloud. The include Office, BizTalk, SQL Server, Sharepoint and more. It sounds like they are trying to cover all things for all people, which isn't something I want to hear.  The more of this there is, the more there is fragmentation and teams moving off on their own as they have conflicting requirements.

Ted isn't a great speaker. As he speaks, it feels as though his talk is a bit contrived and the things he says he "likes" are a bit forced. I also feel that he doesn't necessarly understand what he's talking about, but just talking. I doubt you want your Vice President playing in code, but it might be better to have someone else giving more of the details.

The talk seems to be a marketing pitch for SQL Server, with the highlights of things added in 2008. Why you should be using 2008. It feels like the new car pitch, which is the same each year. You want to buy this year's model, it's much, much better than last year's. I don't know how you balance this out, when do you keep the older version around because it works. I wish that they would mention that the investment in 2000 or 2005 isn't lost and that these versions work well together.

Ayad Shommout - Lead DBA, Caregroup Healthcare. Welcomed on stage as a person that upgraded to SS2K8. It's a hand-picked person that is upgrading their systems to SS2K8 from 2000/2005. The guy isn't a great presence either, and the two of them together seem really forced.

Since RTM, unprecedented interest in adoption? A nice phrase, what does that mean? Some fact given

Adoption inside MS? 

They're good numbers, but so many people I talk to don't upgrade that quickly. The versions are coming more quickly, and my guess is that we are going to have more and more mixed shops. With SQL 10.5, or potentially 11, coming out in 2010 (18 months after SS2K8), that means that if you started upgrading your 2000 servers to 2008, my guess is you might not finish if you have any number of servers, say 20-50. You might end up with 2000, 2005, 2008, 2010 instance in your shop. At JD Edwards, with hundreds of server instances, we had mostly 2000 when I was there, with a few 7s still around. My guess is if I were still there we'd be in the midst of still moving 2000 to 2005 right now. And possibly have 2008 instances as well.

SQL Server will support more than 64 physical threads in Kilimanjaro, so you can get larger and larger servers. A WinHEC demo showed a 128 core system.

The acquisition of DATAllegro is being integrated into SQL Server to gain some scale out for data warehouses. This is Project Madison, releasing in 2010, and there's a demo of this being given by Jesse Fountain. Shows off some early work on the Massive Parallel Processing from DATAllegro. The DATAllegro engine used Linux and that type of platform. This has been replaced with Windows and SQL Server and uses their technology on top to give you the scale out servers.

They moved to a refererence architecture to get this working. Delal in this case, though they are working with HP as well. They generated a fact table with 1 trillion rows, and the associated dimension tables, and got a DW with over 150TB. This was then split out across multiple nodes of SQL Server. The DATAllegro technology then allows a join to accur on each node and the results compiled.

The demo shows multiple instances of SQL Server, 24 boxes, each with 8 cores. There is a control node as well that manages things. On a separate screen they show an SSRS report that is the front end. Code shows dbcc dropcleanbuffers to clean cache and then a query. The demo of the MPP system, showing bars for the loads on each core, is very cool. Not sure how applicable it is, but it looks cool. Show that to an exec and he'll think that he's getting some work done for his money.

Manageability

Looking forward to SQL 10.5, multi-server management is receiving focus to reduce TCO. I wasn't sure how big a deal multi-server management was in SS2K8, but they must see value there. It's a good idea since I've always tried to set up the same management tools and processes on each server, so that I can efficiently keep an eye on things. Ted says that the "down payment" on these things is being released in Kilimanjaro, but wasn't this done in SS2K8?  A mistake by Ted.

Dan Jones comes out to talk ab out this and corrects that right away. SS2K8 was the down payment, Kilimanjaro builds on this. We manage instance by instance? Not the good DBAs.

SQL Server Fabric - announced as the new buzzword. There is a Fabric Control Point, what most of us would call our management instance.

We get another DAC. A database application control point, great, now how do we keep this separate from the Dedicated Administrator Connection.

Demo: Looks like a fancy MDW. We see policies now kind of extended to profiles of systems. What's under or over utilized? It's a nice way for things to evolve, at least it looks good. Like with policies, I worry about the complexity for most DBAs.

dacpac - deployment package. It's going to be fun to talk with DBAs in a few years with all these new acronyms. I wonder how many will still be around?

Import / Export is a good idea, but what about undo? I can see people putting the wrong one on the wrong server, how easily is it fixed?

Deploying applications to a fabric, not a server? Let the system figure out where to put things? It's a good idea, and makes sense in terms of moving to more of a cloud structure, even an internal cloud, not necessarily the Internet. Connectivity? We provide a "friendly name" that allows us to connect through AD, not just using a  server or instance name. I assume some backwards compatability, but there is a lot of work for third party vendors to do here. At least that's my guess. They will need to plan for single instance or a fabric. I've always done this with my clients at large companies. You tell me the specs for your app and I'll decide where it goes. That allows me to reuse underutilized instances and combine apps. 

SQL Server Data Services

Azure data platform includes SSDS and eventually more services and capabilities will be added, eventually to include all aspects of SQL Server services. The public CTP of SSDS is available.

We had a fairy tale story through slides that sucked. Sorry, it did.

OK, the network died. That's great. If you see this I managed to save this and post it later.

Managed Self-Service BI

Quite an acronym. 

Ted gives some scenario, which makes sense. You build some case, but I think it's easy to get caught up in building that solution and not really consider how different the needs may be of most people. You might build something that no one needs. I see here where 37 Signals avoids these models and tries to instead get actual work done and not contrive some example.

Demo: Donald Farmer in wings. That's quite a sight. Wish I was closer so I could get a picture.

Mix and match data from a database with that on a local system. Using Excel to bring in a bunch of data on the desktop. 10mm rows in Excel, very cool!

Bringing in data from a variety of sources, each as it's own tab. Donald created a pivot table using multiple tabs and as tabs are added, Excel infers relationships between the data fields. You get a variety of visual filters as well. I can see some people getting lost in this and wasting time with "what ifs", but it's a powerful tool.

His latop is quick, I wonder what specs are there? The "pop up" of each screen is pretty cool, if it works that fast.

Ted mentioned SP3. They haven't forgotten SS2K5. A good thing, though not highlighted very much. Still, I guess they are in the software business. 

 


Copyright © 2002-2014 Simple Talk Publishing. All Rights Reserved. Privacy Policy. Terms of Use. Report Abuse.