Kent Tegels
SQL Down Under Show 39 - Guest: Kent Tegels - Published: 29 Aug 2008
In this show Kent Tegels disusses the spatial data type in SQL Server 2008.
Details About Our Guest
Kent Tegels is an developer with Developmentor.
Show Notes And Links
In August 2010, we got the sad news that our friend Kent lost his battle with illness and passed away. I cannot begin to express how sad I was to hear about this. The SQL and XML communities are weakened by this loss.
Farewell Kent and thanks for all the fish.
Show Transcript
Greg Low: Introducing Show 39 with guest Kent Tegels. Our guest today is Kent Tegels. Kent is a developer with Developmentor. I'm pleased to have Kent back. He was here on show three when we were having audio problems. I was determined Kent would come back.
Kent Tegels: I’m glad to be back. You have a great show and I’m happy to be here.
Greg Low: We were fairly detailed before but let’s start with a quick summary of your background.
Kent Tegels: I'm with Developmentor. I’m the database curriculum lead. I spend time worrying about people’s servers. Teach a lot of SQL Server 2005 classes. We are just getting the 2008 classes ready for June and July. I’m looking forward to teaching those. That keeps me busy.
Greg Low: That’s great. I’ve been doing the same, running around the country doing readiness events for SQL Server 2008. I’ve been keeping busy with the new Microsoft Certified Masters program for SQL Server. For those who’ve come along later in the year, I encourage them that it looks like a good course. I’m looking forward to presenting a week of that later in the year.
Kent Tegels: I look forward to seeing that, seeing what’s covered.
Greg Low: I’m allocating six to eight days to build each day of content. Interesting. A lot more depth than some areas I looked at before. They’ve broken out technical content from the Microsoft Certified Architect content. I think the idea is you can now do the Masters as one level of certification, the technical content, then another two weeks of architecture work. I don’t know if that’s available yet.
Kent Tegels: Someday we’ll see a data architecture program. Software architecture is one thing, but not a lot of people are good at data architecture and I’m hoping they’ll take the opportunity to go that direction.
Greg Low: Yes. That’s great. What is the main thing you’re busy with yourself?
Kent Tegels: Mostly working with what’s new in 2008. Neils Berglund, my cohort and I have divided up the changes that applied to developers between us. We each took responsibility for the things we like. I’ve been working a lot with business intelligence stuff. Spatial stuff, the things you can do, what’s offered, what isn’t. It’s been fun. It’s been frustrating. Overall, I’m impressed with the package.
Greg Low: That’s the topic for today, spatial implementation in SQL Server. I’ve come across a lot of people who think spatial isn’t something their applications would take advantage of in any way, just specific applications that would use spatial.
Kent Tegels: For the next five to six years that might be true. Well recognized topic, but specialty. Now that we’re seeing it in other products, just a matter of time before users ask where things are located in space, how to organize and understand data. Matter of time. Driver also from business intelligence state. Consumers, managers, decision makers will want to understand why one zip code has more sales. Huge amount of interest in analyzing census data, making good business decisions off of. SQL Server might be a down payment on the best is yet to come. We’ll see where Microsoft goes with it.
Greg Low: Strikes me, when I look at analysis services, created a new branch of product. Spatial is a significant turning point. Having built a lot of business applications, I think on how many applications I’ve built that could have taken advantage of spatial awareness. Nearly everyone. If I look at how quickly it is on phones. I went and bought and iPhone. Dark side. What’s interesting is that I take for granted that I can pull up a map, ask where something is, and rather than get directions to there, from here. Aware of all those things. I don’t have to tell it where it is. So routine for me now. Might come faster than we imagine.
Kent Tegels: I think spatial awareness in portable devices has created a lot of interest in it. 1998, I was trying to find specific hotel. I pulled out my laptop, the USB antenna for Map Point. Now I look at my phone and can fire that up and it would give me walking directions. Before, a lot of power to do simple thing. Now that we’ve built up server infra, and know what GPS can do, has evolved quickly. I don’t see that curve dampening. Will continue to evolve. We’ll have many devices location aware, feeding data out, making presence knowable to us. Now that Homeland Defense and national security is issue, knowing where people are, pinpointing events accurately, concern. I think there’s a lot of space to grow. Won’t take a lot of time to fill.
Greg Low: Amazing how fast it creeps in. PASS. How many people are members where there isn’t a chapter within 50 or 60 kilometers? You can now answer those questions. Before, diabolical to do.
Kent Tegels: Now you can look at postal codes and know how many people are in those codes. But some of those postal codes land areas are so large you may not be able to physically commute.
Greg Low: Services I take for granted. I use Google maps. Live Search and Earth can do same. I can take series of addresses in database and fire against web service and have it tell me where they all are is amazing.
Kent Tegels: I think where the sweet spot is is building the analytical applications. Modeling applications. There’s a whole series of talks at Redmond University on this. Every time I listen to that set of talks, I think we can do this with SQL Server now, and don’t need the huge amount of money. We have to write code. I like doing that, exploring things.
Greg Low: Suddenly spatial awareness for the masses. Microsoft does good job commoditizing complex technologies. I can see those going into mainstream use.
Kent Tegels: SQL Server has been Microsoft’s way of throwing things at the wall and seeing what sticks. We’ve seen with XML, Service Broker, Analysis services brought business intelligence to the masses. Spatial is in that tradition.
Greg Low: Maybe we should go through implementation. We have two new data types.
Kent Tegels: Geometry, good for storing 2D data. Geography, set up to model the Earth. Calculations based on not quite spherical measurements of the curvature of the Earth. Useful if you want to know realistic distances between things, but not perfect 3D, XYZ, but works well enough for terrestrial based calculations.
Greg Low: Critical point. No height component. XY, not Z. We have positions all over the globe but don’t have the height above the globe.
Kent Tegels: There’s also no 3D for geometry. There are applications I can think of where 2D is fine, but would be nice to have true 3D geometry support. When I talk on spatial data types, not just where things are on Earth, but putting things in relation to other things. A collection of books. I might want to know genre, clusters. If someone buys a book in one, what others are close in semantic dimension? If we have that we can build geometry of relationships. Atypical use of what it was intended for, but works nicely.
Greg Low: You also mentioned the calculations done on ellipsoid, squash to ellipsoid. Targeted at Earth. I was in summit last year. A guy at the back raised his hand and asked if he could model Saturn with it. No.
Kent Tegels: Not in this version. Maybe by 2011, we’re thinking of how to map things on the Moon, Mars. Maybe there will be non-scientific applications, but business or individual applications. Today, Earth only.
Greg Low: Key point is not parameterized. All calculations preunderstand the rough shape of the Earth. If you’re doing calculations on sphere, most people remember enough to work our rough calculations. But on a squashed ellipsoid, different story.
Kent Tegels: Right. Wonder sometimes if statistics not geometry. That’s what I know best about math, but when I look at what we do on projections, it hurts my head. I’m glad Microsoft wrote all that stuff and I don’t have to. Might be nice to map Mars or Venus, but at same time, I’m happy if Microsoft writes the code and not me.
Greg Low: Absolutely. Another key point is that this is a separate MSI.
Kent Tegels: Spatial data points are available in .NET. If you’re trying to build up, you don’t necessarily have to be in SQL Server context. Indexing in SQL Server usually works but not always. Persistence in transactions and things in SQL Server. Not limited to just in SQL Server.
Greg Low: That’s key point I was getting at. They ship separate installer file with assembly with clever code that does calculations. Same as the one in SQL Server. Nice that you can build in your own applications, taking advantage that someone has done the calculations for you.
Kent Tegels: I’m glad they did it that way rather than as in SQL XML where they have on server side, not client side. Nice to see SQL XML things on SQL Server but you can’t win them all. I’m happy we won the spatial one.
Greg Low: That’s good. Key point is that they’re system types.
Kent Tegels: Yes. Permissions. Use CLR types. You do great presentation called SQL Server for DBAs. Now you have CLR where you turn on or not. There to use. Interesting. Microsoft confident. There whether on or not.
Greg Low: Also implemented is hierarchy ID. Do you have thoughts on that?
Kent Tegels: If you keep an eye out in October, I have an article on that for a bills of material system.
Greg Low: Have you seen implementation? More I play with, less happy I am with how it’s implemented.
Kent Tegels: Yeah, that’s my experience too. I won’t say unhappy. I’m happy we have the data type. It could be improved. Nothing is ever perfect. Useful, you bet. Shortcomings, but let’s give chance to evolve.
Greg Low: What I see as difficult are things that should be autonomic, adding a child to a parent. That makes you find position and then add there. Concurrency, all should be single operation.
Kent Tegels: Yes. Just enough interesting stuff to make interesting topic to talk on. I think useful for where it is. Construct once, leave it alone thing. If you can do that, most effective. Same with geometry and geography data types. Construction can be fun, then read only data. True of hierarchy IDs. Leave alone after constructing.
Greg Low: Most missing with hierarchy ID is things like moving sub tree, re-parent only moves me, not my children. When I look at data type, can always derive from it and add more down the track. I think there will be more. Of all new data types, this is most Version 1 type.
Kent Tegels: I go along with that. I might say geometry and geography more evolved. Things I’d like to do. We have almost all OGC methods implemented we have plenty of non-std implemented, for most part mature for Version 1. We’ll see where it goes.
Greg Low: Geography and geometry. Geometry is flat Earth; geography is sort of round Earth, geodesic type calculations. What we should talk on is things like advantage of data type versus just storing data in other form, latitudes and longitude in database.
Kent Tegels: Yes. I think that’s good question. Mental example. Let’s say you’re building application to help customer find your location. Can do since SQL Server 6.5. Have had to store, then write find functions, compliance set code. Everyone had to write from scratch or find code and implement. Now you have specific data types that know how to act with latitude and longitude giving you distance but other things, methods like collections. Imp to remember about both is they don’t just store one point; they can store multiple points connecting together. Collections of polygons. Complex shapes. Data I work with a lot is census data. Organized along tracks, sometimes postal codes. Demographics, racial background, education. Can store these demographics and have in same row, one or two actual data or realization of that polygon. Much use for this. Relating shapes to other shapes, where on territory meets another, concentrations of people in those areas. That’s where data types shines, storing data by data type.
Greg Low: I think that’s key. Store shape, and that shape can have multiple discrete shapes including things like holes in the shapes. Potentially, what are the most imp properties of the data types?
Kent Tegels: As the normal .NET type?
Greg Low: Concept of non-geometries and things like that.
Kent Tegels: When I’m dealing with collection, member of that collection can see what member looks like. Series of points in geometry collection. I want to parse to discrete point. Property called number of geometries will tell me that. I can get one geometry or sub geometry. Holes might represent a lake. I might have data that represents state and am interested only in water areas. I might be working with potable water application and want to know where my water sheds are. Might be good case where I have complex geography and want to look at the holes and say they’re potentially water shed.
Greg Low: One thing also is in some cases might be a lot of data stored in data type. Decision in every application is which precision data stored to. Lake or country border. Whatever. You could go down to nearest meter or 100 meters. Depends on perspective. Changes amount of data.
Kent Tegels: It could. Most people either already have the data and want to store or query. Not a lot of consideration I’ve seen given on scale and projections. Spatial references, yes. Requires data on SQL Server. Expect chunks to be significant in size. We can now have CLR driven objects that exceed. That was a problem with user defined types in 2005. You couldn’t store enough data. In 2008 you can. You can use more data types with CLR objects that enable things like geometry and geography to store these large chunks of data.
Greg Low: Yes. Increases number of things you can use CLR user defined types for. I talked with product manager at the time. 8K limitation was arbitrary. They just wanted to stick toe in water.
Kent Tegels: I was under impression they were limiting to one page. Conceivable. Might have said one page of data easy, multiple more hairy.
Greg Low: Another question. Different types of data. Points and polygons. Different types you can store?
Kent Tegels: You can store lines. Lines connected together to form polygons and collections of those as well. Formed out of services of line fragments. Store points and lines and composites of those. Just one point or one line or one polygon or collection.
Greg Low: Challenge in getting data in.
Kent Tegels: Absolutely. As much as I love SSIS they don’t give a lot of help getting the data in.
Greg Low: Not at all. If we look at interchange formats, what they are, what we can bring in and parse.
Kent Tegels: Major one I encourage people to use is GML. A lot of data already expressed in GML. SQL Server has added extensions to read and write GML. Not all GML is useable by SQL Server. Isaac Truman, program manager for spatial stuff, has nice summary of what’s allowed, not. Dominant data source we see in U.S. is ESRI Shapefile format. Microsoft ships nothing now that helps us get that in SQL Server. You end up looking for third party tools. Shaped SQL and variances of that. Feature Management Engine, FME. Greg, I love that product. I use daylights out of it bringing data into SQL Server and getting geometric and geographic data out of SQL Server and into another format like Live Earth. I’m a Google maps guy. I rarely use Live Earth. Tools to get from point A to B. SQL Server out of box isn’t giving us a lot of help here.
Greg Low: I think what they’ve provided is basic plumbing. There are layers we’ll need. Tools for importing data is imp. GML, Well-Known Text (WKT).
Kent Tegels: I want to discourage people from using WKT. It’s easy to use but has performance issues. SQL Server has to parse, then load and turn into shapes. If performance matters, avoid WKT as much as you can. Same is true with Well-Known Binary (WKB) format. Open geographic folks have accepted this in binary but SQL Server has to do a lot of shaping and parsing to get this into native format. Best way to go is use native format. I blogged an example the other day of building a shape with .NET and storing in native format. Performance is as good as it gets.
Greg Low: How does size of data compare in GML and WKT?
Kent Tegels: I’ve not spent a lot of time looking at that, except where GML was much larger. Magnitude.
Greg Low: When you put in XML, worried might bloat out size of data.
Kent Tegels: Not just serializing the text. There’s a lot of metadata GML requires. I remember doing something simple and wound up with 600 or 700 bytes of data in GML, 600 or 700 characters. Inflation of data. We talked internally about spec'ing or machines for new class. First time we were talking about megabyte size data sets for demos and labs. Not hard to be in gigabyte range with spatial data, even for trivial things.
Greg Low: At present time, is there a life outside SQL Server? Something you’re passionate about?
Kent Tegels: This summer I decided to go back to grad school. I’m going at night. With heavy travel schedule, interesting. Decided to work on my masters in management. Technical path. Interesting. Primarily, did so I could see how we can be pervasive tool. What are mgrs and decision makers looking for in BI tools, learn the other side of that.
Greg Low: Good. You’re enjoying the challenge of study?
Kent Tegels: Yes. Strange being on the other side of the desk. Instruction and mentoring, then back to both, changes your world view. I’m lucky to have great professors.
Greg Low: I don’t miss having to do that at the moment. I had long period, about 21 years. I had six months off in 1986. I remember feeling guilty watching TV.
Kent Tegels: Not a lot of TV at my house these days. A bit of the Olympics. I’ve been doing a lot of road work, so that’s great to get me to stop thinking. There’s not a lot of free time in my life these days.
Greg Low: We watched a bit of the Olympics. Key passion in Australia. We sit about fourth or fifth, out of proportion for the people in the country. Crazy. This year we have about half the medals of China, but they have 60 times our population. In 2000, they had 59, we had 58. Completely bizarre. I was saying on regional director list that I wished the same passion existed for other areas in the country.
Kent Tegels: Sports are easy for people. Amazing to me is how well Silverlight has done. So much broadcasting in Silverlight. There again, we’re looking at new technology for that, yet one of the biggest websites this year run that at its best and hasn’t fallen apart. Sure, bugs, things that make hard to develop, but has held up good for the Olympics. Kudos to Microsoft for that.
Greg Low: We take for granted all those things where you can hit sites and stream.
Kent Tegels: How different from China and Athens games. Not a lot of media streamed. Now you can get whole screen view of things. That’s impressive technology. I can’t stream to save my life and I’m glad that someone else does that technology. I’d be no good at it. Shows Microsoft can put out quality product on fly.
Greg Low: Other thing I remember you being passionate about is various types of music.
Kent Tegels: Yes. That hasn’t changed, nor has it grown. Pretty much same sorts of music. I’ve recently started getting interested in new genre of rap music. It’s called Nerdcore. Its nerds rapping. The artist I’ve been listening to is MC Frontalot. Very interesting lyrics. Do you remember computer game Zork? He’s got a rap about people playing Zork. It’s a funny video, great music. Funny but interesting to see that generation combining music and technology. Maybe by SQL Server 2011 we’ll all be listening to nerdcore.
Greg Low: Intrigue me when you mention the computer games. Many people have passion about playing old games. Intriguing the young kids playing things like Pong.
Kent Tegels: Grand Theft Auto becomes repetitive after so long. Sometimes you plug in the 2600 and play Tank. Ugly graphics but different.
Greg Low: Some run quite well on hardware. Some have ported across. I have old software versions including SQL Server 1.1. Fascinating that you can read books on line in 15 minutes.
Kent Tegels: UPC?
Greg Low: Yeah.
Kent Tegels: I want to get a copy of that. It’s been so long since I saw SQL Server 1.1. I remember installing and using.
Greg Low: Ran on OS2. Had trouble booting at first, giving it too much memory.
Kent Tegels: SQL Server with too much memory.
Greg Low: OS2. Give 128 MB and won’t boot. I’ve got a copy of VisiCalc 1. It still runs well. It’s fast.
Kent Tegels: Can’t hardly keep up with it.
Greg Low: Entire program is about 122KB.
Kent Tegels: Interesting to me is we’re talking software only 22 years old. In our lifetimes, we’re in the middle of where software will be by when we retire. In 20 years, think what software will look like. We’ll be sitting in retirement home talking how great spatial data and SQL Server was when we first got.
Greg Low: Back to spatial. We did touch on tools support lacking at the moment, getting data in. where is most common places you find data of interest?
Kent Tegels: For us, easy. In U.S., data owned, managed, distributed by Federal Government. They make it fairly easy to get to. I’m not saying no barrier entry, but lower than other places. Easy to get census data, go to municipalities and get plates and property lines. In other parts of world, not as easy. Getting U.K. postal codes, costs to get out of U.K. Government.
Greg Low: Early on, Australia licensed this data, thinking fairly specialized. Licensed individual companies to collect and resell on their behalf. Many cases, expensive.
Kent Tegels: Here in U.S., lower barrier to entry. We’ll probably see more use of these built in other places. Other governments might centralize this data. But there are still a lot of data in U.S. that needs to be centralized. Disaster planning. Not a consistent standard for emergency data. Fire stations, transportation connecting. Laws saying we need to have that, but not for consuming public yet.
Greg Low: We had similar things with basic streets. Slow getting GPS off ground here. Biggest hassle was who owned data. I went to Japan and for $2 had DVD of all the roads in the country. Japanese Government decided wide availability most useful.
Kent Tegels: Absolutely. Very much the case. Governments will do what they do. As developers, one of the things we deal with, doing business.
Greg Low: Locally, if we wanted to buy electronic phone books, one monopoly company charged huge amount for that. Even when they mass produced, still to buy for main city, still $500 to $600. Another company fixed that by sending phone books to India and telling them to just type them in.
Kent Tegels: Here in the U.S., there’s so much advantage in phonebooks, they’re free.
Greg Low: Free here, but no electronic versions. When they came back that data quality of what had been keyed wasn’t high, dealt with by sending to another company to type again.
Kent Tegels: Nothing like manual data strokes. All the fancy stuff you want, but if data not right, hard to fix.
Greg Low: We find universities, governments, census folk, make data available. I find most in Shapefile format. Biggest limitation is no tool out of box to pick up and drop in appropriate data types. Non trivial. Not ridiculously complicated. I have colleague building freeware porter. We’ll be dependent on those for awhile.
Kent Tegels: Microsoft realized business opportunity for ISVs in this space. Purchasable SSIS components. I thought there would be more. Not a lot in market. Data source for Shapefiles. AXL files. Some people will get around to that. Safe is king of hill, as they’ve got out of box. In market early. I don’t want to see Microsoft become free, but encourage ISV market, but now like pulling teeth.
Greg Low: Tooling wise, missing, way of visualizing data in Management Studio.
Kent Tegels: In RTM they did include Data Visualizer. In RTM, when you run spatial query, there will be new tab in results that will do basic rendering, pan, look at data. Helps quite a bit. Not perfect. Better than what we had before.
Greg Low: Good. You’ve mentioned OGC. Let’s define who they are.
Kent Tegels: Open Geospatial Consortium. GIS practitioners that have joined to create standards. A lot like World Wide Web (W3C), for geographic information systems. Their decision making process faster than W3C. Not as big.
Greg Low: Many methods, things called data types, are versions defined by OGC. One’s with ST prefix?
Kent Tegels: Spatial temporal methods. Those are defined in specifications. Inconsistencies. There’s a well known geographic database called PostGIS. They have a lot of those methods in. Oracle has some; SQL Server has most if not all of them. Nice to have standards that describe this work. Easy for Microsoft to do implementations. Not trivial for Microsoft, a lot of work for the implementations, but nice to have standard to work with.
Greg Low: There are a lot without the prefix, so they’re the Microsoft extension.
Kent Tegels: Right. Safest to assume. If no ST, likely Microsoft. Other databases might have similar functionality, but will cal something else. Not part of OGC specs.
Greg Low: Let’s talk what you think are most imp data types.
Kent Tegels: Hard. So many things you can do. Early on and often we’ll see distance used. Distance between points. Imp as people want to know if objects share common point our boundary. Example we use is in U.S., no way to know what area covered by zip code. Where it is relative to higher level of geography like state. Sometimes you want to know adjacent zip codes. Ability to do unions, intersections, disjoints. Parts that connect, overrides and don’t override each other. That’s what people will focus on.
Greg Low: I notice capabilities got better as product evolved. Things like distance. Between things, at first you could do between two points, then to arbitrary shapes, then two arbitrary shapes. Computations involved aren’t trivial.
Kent Tegels: Distance between nearest points is difficult for polygons. You could have multiple points, some equally close. Which to pick? Nice to not worry about. We don’t have 3D capability, height dimension. Would be nice to have. Now, not supported anyway. Maybe we’ll get there eventually. When Schema supports, SQL Server will support.
Greg Low: That’s good. Demographics. My mother, with last name Low, local government has decided she must be Chinese. Rule based thing. Annoys her that every Chinese New Year, local politicians send her a New Years card in Chinese. She’s not Chinese.
Kent Tegels: Can she read it?
Greg Low: Not in the slightest. Effect politicians aiming for has exact opposite of it.
Kent Tegels: If she were here in U.S., there would probably be a tax increase. She should be happy she only gets the card there.
Greg Low: That’s good. With special data types at moment, there are many differences between geometry and geography. Less on geography?
Kent Tegels: Yes. Largely because OGC said appropriate for 2D. 3D would be hard, not practical. For union, makes sense. Minimum bounding rectangle for points on a convex hall? Below surface? Above? Not every method available on both types. What’s appropriate is there. Microsoft and minimum bounding rectangle. There’s no ST envelope for geography. Microsoft came up with the envelope that will let you come up with semantic, meaningful envelope over geographies. That’s a shortcoming we’ll address with extension methods.
Greg Low: Other key things. Order you declare points in polygon.
Kent Tegels: Polygon’s for geography matter. Ring orientation tells you if inner polygon, donut in ring, or another surface exposed. That’s how SQL Server learns if interior or outer ring.
Greg Low: Common example, the shape defining postal code defines what’s in and isn’t in that area.
Kent Tegels: Right. In census data here in U.S., large bodies of water, roadways, don’t have zip codes. No mail. U.S. Census treats as XX areas. When you go look at zip code, in my case 57110, if there’s a lake, might see that area 571XX for its zip code calculation area.
Greg Low: Order of points matters. You seem to go anti-clockwise to contain something. Worth mentioning change made late in game, latitude and longitude.
Kent Tegels: Originally, user opinion was latitude longitude. No. GIS professionals and people who deal with spatial data think longitude first, then latitude. Made change late. Change they needed to make. Makes working with data sources easier. I was happy to see.
Greg Low: Means that many samples, things already out there, may have problems. Many samples done when other way around.
Kent Tegels: Our job as MVPs and regional directors to get that message out today. Issue to be aware of. Look at version. Was it RTM? When you see that come up in news groups, CPU or RTM?
Greg Low: Other area I wanted to talk, indexing. Aiming to get performance out of. Indexes try to do that. What indexes, what we can set, how effective?
Kent Tegels: My thoughts boil down to this. Be extremely careful with them. We always assume an index will help performance. Not always case with spatial data. In spatial index can specify level of detail. Four levels of detail. At each can set granularity from high to low. People assume if all four levels high granularity, will get great performance. But these objects are so big, you generate so many loads. Hurts performance. My rule of thumb is when you work with query, run without index. Use baseline. Then create geometry index, data type, with lowest granularity. Look at test query results. Increase granularity level four up. I’ll set all to low, get level, run many queries, take average, slowly from bottom up increase granularity. I find loads not as much as I need. Shrinks size of index tables and makes unlikely to be used by SQL Server. Imp to understand about spatial indexes is SQL Server will not use at all. Too expensive. Cost optimizer. Looks at as way more expensive. Probably true. But it increases input.
Greg Low: What parameters do you get to set when you build an index?
Kent Tegels: You can set number of objects in cell for four grid levels, four grid levels, granularity for grids. Smallest grid 4:4, largest 16:16 at each level.
Greg Low: Well, that brings us to time. Where will we see you? What’s coming up?
Kent Tegels: I’ll be in Los Angeles and Boston the rest of the year. I won’t make it to PASS. Bummed about that. I’ll be doing local user group talks. You’ll find me at the Heartland Developers Conference in Omaha, Minneapolis. Best place is to watch my blog on where I’ll pop up next.
Greg Low: Thank you for your time today, Kent.
Kent Tegels: Thanks for having me back. It’s been fun.
Phone: 1300 SQL SQL (1300 775 775) l International +61 1300 775 775 l Fax: +61 3 8676-4913
Copyright 2017 by SQL Down Under | Terms Of Use | Privacy Statement