SQL Server Boss Gives a Sneak Peek

By Matthew Hicks  |  Posted 2002-09-09

SQL Server Boss Gives a Sneak Peek

With Microsoft Corp. coming off a major security push and moving headlong into its next major SQL Server release, code-named Yukon, eWEEK Senior Writer Matt Hicks recently sat down with the companys SQL Server Vice President Gordon Mangione at Microsofts Redmond, Wash., offices. Mangione offered both a birds eye view of what to expect in Yukon, due in 2003, and a deeper look at a revamped security push for SQL Server.

eWEEK: Microsoft is supposed to release a beta version this year of Yukon. Where does that stand?

Mangione: Yukon is a big release. I mean theres no ifs, ands or buts about it. The teams rolled off our whole security initiative—for SQL Server 2000 and Yukon. Well be coding for most of this year, and youll see us into beta early next year.

eWEEK: So not this year?

Mangione: No, no…the security stuff was interesting. The reality is we looked at what happened last fall, and we as a company had to do something different. What we were doing [with] knee-jerk reactions wasnt going to work. I worked for Brian Valentine [senior vice president of Microsofts Windows Division] for four years, and hes very much, "were going to go change the way were doing things." And he literally wrote an internal mail, and then Bill [Gates] backed him up with the Trustworthy Computing [initiative] within a week that said, look, were going to take everyone off writing new code and youre going to dive in and were going to code review every single line of code. Were going to redo our processes. Were gong to take those tools that we always talked about using in research to help us design better code, and were going to make them part of the standard process.

So we were kind of fortunate in SQL [Server] that we were able to build upon the work that really got started in the operating system, but the reality is that from start to finish it was three months. Every develop, every tester, every program manager…we literally code-reviewed every single line of code, we rewrote entire test plans, we built security threat analysis and really looked hard at everything we do inside the product…

It was three months of absolute dedicated time on it and that did impact the Yukon schedule. It was frankly an easy decision to make. But really whats happened more than anything is weve looked at our processes from end to end and made sure that this just has to be part of what we do—every code review, every build.

eWEEK: When it came to SQL Server and security, has it made any significant changes in how youre architecting the product?

Mangione: That exact question got asked of Windows. Had we done something architecturally flawed? …We found out there wasnt anything architecturally flawed in what we did. Things that seem like small things, like buffer overruns when you combine them with other things can be used as launch points to go and execute code off of a stack or execute code internally. Its mostly little small things…

Page Two

: SQL Server Boss Gives a Sneak Peak">

eWEEK: Oracle Corp. and others are pushing this idea of using Linux on Intel Corp. servers as a way to lower the cost of ownership in deployment. Do you take that seriously?

Mangione: We take Linux very, very seriously. Where Linux could be and where it is today are two very, very different things. Its kind of funny: Linux is driving a wedge between Oracle and Sun [Microsystems Inc.] that I dont think ever would have happened. From my perspective, I always looked at them as two sides of the same coin. I mean they would go in together, they would sell together, [and] theyd sell to scale up. In some ways Linux is probably hurting Sun more than it is hurting even Microsoft at this point or any of the vendors in thats space…

Oracle, again, realizes its in a bit of a quandary. The proprietary Unixes on the proprietary hardware, you can see that world is going to come to an end. Its going to go the same way as the Unix workstation business. So Oracle has to do something for [the PC server] platform. They think Linux probably is the thing they want to do. Linux isnt really going to scale over 4-processors. Once you get into databases, scale-up and [deploy] big boxes, it really is databases that drive that business. I dont know what theyre intentions are to really take Linux up to 8-proc or 16 proc. Were going to start seeing 32- and 64-proc PC servers coming out, and Linux just wasnt architected to scale on those platforms.

eWEEK: What aspect of Yukon is is going to be most significant to customers?

Mangione: Our customer set gets spread over three main areas—its developers, the IT staff and what were really doing around business intelligence to drive more value. I think in the end its going to be what our key tenets have always been: Were going to help you drive costs out of your business and were going to help you get better value out of the systems you already have today. When I say systems, its not just computer systems its your employees, your trading partners, your customers and really better understanding what is going in that environment.

Theres tons of work were doing in scalability and manageability. Weve always been leaders in ease of use, and youll see us just take that to a whole new level in the next release. Youll see systems that really help you alert before things are really going on. It fundamentally is one of the key differentiators I think were always going to have over Oracle and IBM.

I mean we sell to small and medium business, whereas I would claim that neither [IBM or Oracle] is really in that space. As a result, we have to have breadth-based programs that collect information, that update servers, that help customers work through problems and make their systems run. One of the knocks of SQL Server is by the time it gets into the IT center there wasnt anyone who was a DBA that was keeping the system up and going…

We can talk about what were doing with .Net and how its going to change the server and how people program. But I just think about that as making your developers more productive. Let them use the tools they already know today. Dont force them to learn TSQL or PLSQL or whatever else. Let them use [Visual Basic] or COBOL or Java or Perl even if they want the database to go write their applications.

eWEEK: And thats the move to support the Common Language Runtime?

Page Three

: SQL Server Boss Gives a Sneak Peak">

Mangione: Absolutely. Its probably from a technology perspective one of the most exciting thing were doing for developers. XML, well really up our game again in that space. We did a lot of great work in XML in SQL Server 2000, but theyll be lots of new technologies that come in with XML in that time frame. We really want to make it so that SQL Server can be that one repository for all of your information because theres such a win in doing that.

If your developers can use all the tools they have today, and you can put all your data inside the database then the type of value you can drive out of that and the way you can develop applications is just so much more compelling than what you have to do today. Deal with your file servers, deal with your Web servers, deal with your database line-of-business apps – its really, really hard to build apps together that way.

eWEEK: So youre not talking here about SQL Server necessarily being the common store for all this information?

Mangione: We want to make it so that SQL Server is an applicable place to put all that information. So all your documents, all of your XML and all of your regular line-of-business apps can be put in there. And thats really a technology statement in how were going to bring all that information together.

eWEEK: What has to happen for that to occur? People could do that today, I guess, if they wanted to.

Mangione: But not really with the performance that theyd need. Yeah, you can put XML into the database but it all gets shredded on the way in. Yeah, you can store your documents in the database. Were talking about pretty deep integration between technologies like NTSF [New Technology File System] and SQL Server that allow you to have files as columns inside the database and not give up the performance benefits – really get all the performance of streaming file systems with all the programming and isolation and transactions of databases.

The final area were doing a lot of interesting stuff is…business intelligence. Building a BI system today is tough. Youve got multiple vendors involved. You probably got a system integrator involved who has to bring together all the technologies. Its hard to map the data flows. Its hard to deal with all of your sources. You have to have people who understand your industries, ETL and OLAP, reporting front ends and how to analyze that [and] data mining technologies. What we really want to do in Yukon is really use things like data mining to figure out what should be mined and automate some of the processes of building transformations and building OLAP front ends and really taking a lot of the complexity out of that system.

eWEEK: It seems like a common approach among all the database vendors is to drive more and more algorithms into the engine so that more of the data mining can be done specifically there. Is this more of that or are you talking a more significant change of how data mining and warehousing is done?

Mangione: I really think its going to be more significant than that because what we really want to do is build the framework that allows you to build those applications without having to be an expert in all those areas. Yeah, Oracles put in a few things like pivoting cubes into the database, but that doesnt make it any easier to build a data warehouse. Youve still got to move your data, you still have to cleanse your data, you still have to deal with the security systems on the back end. Its really about, How do we use data mining to go and look at your existing sources and come up with recommendations that automate and build those systems for you in a very easy-to-use way?

eWEEK: Theres been a lot of talk about scalability from other vendors—weve seen Oracle make its whole Real Application Clusters [RAC] push. Will Microsoft with Yukon do something in clustering to enhance scalability?

Page Three

: SQL Server Boss Gives a Sneak Peak">

Mangione: Its funny. In the database industry scalability will always be a tenet of every release, just like availability will always be a tenet of every release…Theres a bunch of work really to scale up on that whole new class of PC server hardware thats coming out…Youll see a ton of stuff in scalability on what I call a scale-up scenario.

In scale out, youll see us do things around business intelligence that allow you to scale out on different machines. Youll see some more work in the relational engine. The reality is the manageability of these cluster servers is just not where it needs to be and as a result customers arent seeing the savings in their total cost of ownership in doing these scale-out boxes. RAC is just like Oracle Parallel Server was before it, and the complexity of managing that system is very, very difficult. Its really focused on the very high end with a large number of consultants that are necessary to keep those systems up and going.

Weve ironically collected some of the best shared-nothing minds in the business, as part of SQL Server…We will be making investments in shared-nothing because we just fundamentally believe that thats the way were going to get very, very large scalability. But were not going to have that stuff until we think the manageability of it is just nailed, and you can actually repartition and redeploy systems in a very easy to use way. Thats really the place that were going to focus on.

Well do a ton of work in scalability around data warehouses in Yukon for partitioning your data, but its all going to be focused on a single machine. It will take longer than that for us to really get the shared-nothing approach.

eWEEK: The specific new hardware youre looking to target, is this 64-bit [processors] youre talking about?

Mangione: The 64-bit will be the area where youll see phenomenal performance gains, and the number of OEMs that are investing in it is just truly amazing.

The reason it took Intel so long is they really built a true 64-bit processor with all of the performance gains, where as a lot of these other 64-bit implementations have really been 32-bit cores with extra memory. [But, Intels Itanium 2] is a wicked hot chip.

eWEEK: On the 64-bit front, I know youre in the beta version for the current SQL Server. I guess youre still looking for this year to release that?

Mangione: Well ship with Windows [.Net Server]. Were basically set to go out the day that they ship.

eWEEK: Is that allowing you to take SQL Server to some new areas or new kinds of companies?

Mangione: The query processor can work so much differently when it knows it has that much more address space. Weve taken SQL Server 2000, run it and gone through all the QA processes and got it up on 64-bit. What was more interesting was that once we got it up and running we realized we should tweak the [query processor] to do this, or the way we do parallel scans should change now because of how much memory is in the system. So weve really started to optimize the engine on top of it. I think the workloads will go up pretty dramatically. The big win is just the amount of memory were going to have access to. And the flat address space is going to make it so much more compelling.

eWEEK: Theres been a lot of talk and speculation around--some people have called it Storage+--how Yukon is this big storage push for Microsoft. What is the broader implication for Yukon throughout the product line?

Mangione: First, Yukon is the code-name for the next release of SQL Server…Theres a lot of great underlying technology thats going into the Yukon technology to actually do things like transact across files, and the system itself to be a great repository for XML technology. But thats really the SQL Server business.

Now theres another business at Microsoft which is the file server business, the Windows business and things like that, where well use a bunch of this technology to actually automate some of those processes and build a richer mechanism into those products. But its not SQL Server itself. Its not that were taking SQL Server and lumping it in. Its the components and the technology thats really a part of that to build a richer environment.

eWEEK: These components that could be utilized from Yukon in these other areas, is that what is going to be allowing for easier storage of things like documents directly into SQL Server?

Mangione: Youll certainly be able to get your documents in and out of SQL Server and store them very efficiently using either the Web or the other mechanisms that are part of SQL. I really think about the next generation of storage inside the operating system [is] not just about how do I "file save as" into this mechanism. Its really about how do I find information inside my system. So much of that is going to be [that] the apps are going to have to be taken and the shell is going to have to be taken to the next level in order to store this information.

Rocket Fuel