Lubor Kollar has been part of the SQL Development group since the 6.5 release in 1996. As program manager he was responsible for SQL Servers relational engine, introduced fall over clustering, query parallelism, and led the design of table and index partitioning in SQL Server 2005. After shipping SQL Server 2005, Lubor joined Customer and Partner group consisting of several customer partner facing teams. Lubor is currently managing the customer advisory team, the SQL CAT team, designing the largest and most demanding implementation of SQL Server around the world. Before joining Microsoft, Lubor was developing DB2 engines for various operating system platforms in IBM Laboratories in Toronto and Santa Theresa for almost ten years. Lubor loves outdoors, skiing, hiking, gardening, mountain biking, fishing, and mushrooming, to name his most beloved outdoor work and home activities. He holds professional ski instructor license and during winter weekends he teaches skiing at a local ski resort near Redmond. Since Lubor isn’t a very frequent name, you can use your favorite search engine to find out more about Lubor, his white papers, books, blog contributions, articles, conference papers, patents, trips, and more.
Greg Low: Introducing Show 36 with guest Lubor Kollar.
Our guest today is Lubor Kollar. Lubor has been part of the SQL Development group since the 6.5 release in 1996. As program manager he was responsible for SQL Servers relational engine, introduced fall over clustering, query parallelism, and led the design of table and index partitioning in SQL Server 2005. After shipping SQL Server 2005, Lubor joined Customer and Partner group consisting of several customer partner facing teams. Lubor is currently managing the customer advisory team, the SQL CAT team, designing the largest and most demanding implementation of SQL Server around the world. Before joining Microsoft, Lubor was developing DB2 engines for various operating system platforms in IBM Laboratories in Toronto and Santa Theresa for almost ten years. Lubor loves outdoors, skiing, hiking, gardening, mountain biking, fishing, and mushrooming, to name his most beloved outdoor work and home activities. He holds professional ski instructor license and during winter weekends he teaches skiing at a local ski resort near Redmond. Since Lubor isn’t a very frequent name, you can use your favorite search engine to find out more about Lubor, his white papers, books, blog contributions, articles, conference papers, patents, trips, and more. So welcome, Lubor.
Lubor Kollar: Thank you very much. It’s great to be with you. Thanks for introduction and how I end up working on SQL Server. This happened in 1996. Started in 1995, I was development manager at IBM Canada. Shaped first DB2, now SQL E. SQL2 UNIX database server. I was pretty challenging project. I got call from Microsoft. Had friends here. Asked me to come over and look around. I did in early 1996. Spring. Did some interviews. IBM and Microsoft database, desktop extension of Excel.
Greg Low: Not serious.
Lubor Kollar: I was talking with people here and it was serious. I started thinking about. But I had good position at IBM, friends, didn’t want to leave Toronto. I said I’d think about it but didn’t make a decision. They called me and invited me and my wife and daughter and son to come. Invited all of us. We came and spent weekend here. Friday I was at Microsoft talking with more people, looking around. They took my wife somewhere, I don’t know where. We met again in the evening. She sounded excited about moving here, living here, and so on. We were halfway there. I remember it was a sunny day. I took the car and drove to near mountains to Snoqualmie Pass. Man what I saw. The mountains, the ski lifts, beautiful weather, I saw it and compared to the flatlands around Toronto and freezing winter there. I checked the rain and to my surprise overall the amount of water in Seattle is less than Toronto. Toronto we have summer storms. Here in Seattle it’s not very strong rain, but unfortunately takes several weeks.
Greg Low: I find in Seattle everyone talks that it rains all the time. I don’t find that. So much of time there’s a mist or a very light rain.
Lubor Kollar: Comes from the ocean and sits here, hits the mountains which I love so much, and city. We have unfortunately, day before yesterday and today, mostly wet, tomorrow rain, day after tomorrow again. Stretches.
Greg Low: What’s interesting is that your reasons for moving there, the things that seem to have first prompted was that Katarina was happy with the area and you did too. All things outside the job.
Lubor Kollar: Other things helped me on the work side. I talked with people here. I spoke with Jim Gray who I knew from IBM. Persuaded me that serious things were going on here and I could invest here. I saw enthusiasm, tons of smart people. Young product. That was encouraging for me. There was that part. OK, I was coming, and that happened. That’s how I came to Redmond.
Greg Low: The working environment in Microsoft compared to IBM. Did you find major differences?
Lubor Kollar: Yes. When I came, very different. Team was smaller. We had here very technical management, everybody knew technically deeply. At IBM we had a new wave, technical managers. Technical knowledge didn’t cut it. Very refreshing and I enjoy it here. Windows part was harder for me as I came from UNIX environment. Many people will know what I’m talking about. Learning curve for me. That’s how I came. I came May. Went pretty fast. If it was summer, trying to hold me, but IBM and things there, but once I decided I didn’t look back. I went first, then went back six weeks later to move my family, sell the house, then came to Microsoft.
Greg Low: The features you’re involved in putting into SQL Server 2005 were the enterprise and reliability features.
Lubor Kollar: Yes. The job I was offered here was program manager in relational engine. That’s where I started here. One of the big changes between then and now is that now we’re more enterprise product. I remember when I started enterprise SQL Server, only several percent of the sales, now substantially more. We made product to be more enterprise ready and added many features. One of the first was fat cluster failover. Very challenging. It was before 7.0. We didn’t support multiple instances of SQL Server on one Windows server. Even on 7.0. If it failed over in 7.0, had to do twists and turns to support them running side by side. Bill Gates, I was working on demo and got shirt and still have it today. Relation engine. I had many responsibilities. There was much in optimization, failover clustering. That’s the same structure. Small team. Moving forward fast.
Greg Low: One thing I’d like your opinion on. One of the feelings I got on the main reason they moved to multiple instances was to deal with different collations. One feeling I get is that they’ve taken collations to the column level, I wonder if that was taking almost too far. Have you ever seen anyone use different collations on different columns on a table?
Lubor Kollar: Yes. Not only that I saw customer databases where they use different languages within one doc. One row, one column, mixed languages.
Greg Low: Mixed languages I can see. My feeling is collations are one of the problems in 7. If you needed databases with different collations, needed separate servers because couldn’t have separate instances. When they got into 2000, multiple instances fixed 99.9 percent of the problems I had with collations. That they took to database, column level, introduced things to do with collations and scripting. Irony, I see more issues with collations because of the pain around scripting where people script to database on one server, move to another, then create with wrong collations for that server.
Lubor Kollar: There are some collation challenges. Column level collation, I know some customers use. I don’t know how many do. We have quite a few features we wonder why and how. Our customer base is wide and deep. People are creative, often using features in way we didn’t anticipate. From 2008, we are just learning there is a new feature. Analysis Services. Similar to 2000 scalable shared databases.
Greg Low: We should mention that. That’s something I find few people aware of, scalable shared databases. Could you describe?
Lubor Kollar: Happy to do that. If you make volume on which you have SQL Server data like the files, make read only, and that volume accessible from multiple servers, SAN, then you can connect, attach read only volume to multiple SQL Server instances. Separate boxes. Wal-Mart, we showed queries, scalability. Only thing to watch for is IO. Going after same IO. If you have enough IO capacity you can attach to multiple servers. I’m still fishing for customers using this in production actively. If anyone listening does, shoot me an email at email@example.com, I will give a nice gift. I had potential customers. Back to what I wanted to say. This is also requested from online services and cubes. Mount to servers, connect to cube, do reporting. Not available in 2005 for cubes but is in 2008. Ability with read only cubes. They can be shared among multiple servers. We went through top programs with customers. Originally designed for scaling out. Talking to customers, looking around, we found customers are using for detaching the cube and copying it over and creating multiple copies. Exciting. It works. No need to backup and restore. One difference between relational attach and online services is relational you can change name of database, can’t in online. Otherwise, works well.
Greg Low: You’re now with Customer SQL CAT team. People might not have come across. Can you describe team and role?
Lubor Kollar: After shipping 2005 I was looking for new challenge. I have passion for working with customers, talking with people. I like hands on, things working. Broken, I fix. 2005 shipped. Usually after shipping, big move in organization. Opportunities, hate people working with last two years, want to work with someone else, fall in love with some new technological stuff. Brings different knowledge skills around organization so we do support it. I made deal that I would try for three months, customer advisory team. It had been around for five years. It grew to six or seven members. I tried for three months. I spent half time working on release follow up. I started getting feeling and liking customer advisory team. Exciting, smart people, technical knowledge. I started liking it. I started thinking what to do next and came up with idea of leading that team, assuming bigger role. We sat down. I was afraid to be manager, takes away technical fun. Determined that wouldn’t be taken away. I have to devote to management. It’s important and good. I took the job two years ago.
Greg Low: I notice when I talk with people who work in product group, one thing they yearn for is more customer facing role. Have you found that has been something you had wanted to have?
Lubor Kollar: Yes. I had that urge. I had exposure before. Development is like full-time, full-time, heads down. Now my job is to work more with customers. Not doing like side job during my off time. It’s my major role, taking most of my time. What customer advisory team is? It’s team of ten people. We are working with most advanced, demanding customers.
Greg Low: So this is the tricky implementations of SQL Server.
Lubor Kollar: Yes. Largest, biggest, most demanding, 100TB-plus things, tens of thousands of transactions per second. Most challenging part. Team is part of the development team. How it’s happening and how it works, what we are working on. People ask about our team. I say we work with customers and make projects successful. We don’t always succeed, but do in most cases. Helping the customer. Second is bringing to the group and feed into planning, testing process, development, everywhere. I’m the face of customer on the development team. We are paid with development dollars so we work closely with development team. Showing how people using features, helping test team bring customer scenarios and design, testing things used. This is the value for the development team.
Greg Low: Do you also work on smaller, high profile projects?
Lubor Kollar: Yes. But usually those are big. Myspace.com is running on SQL Server all the production. Today someone told me that they wanted to design something for two million customers. I told them that MySpace ran for 50 million customers. So yes. As I mentioned, the third part, sharing learning’s with community. We have many channels. Conferences, TechEd, SQL PASS. Last year my team goes on track. One track throughout the conference. We took half the PASS tracks last year. Extremely popular. Problem was with fire marshal as too many people in room.
Greg Low: Yes. That was in Denver. I remember all of your team running around in the orange shirts. Obvious who was in team. Really well received sessions.
Lubor Kollar: Now in November SQL PASS will be here in Seattle. We’ll be attending all 12 sessions, my team. SQL Server 2008, all the newest things we’re finding with our customers, what works, doesn’t work, sweetest points, challenging areas, surprises. That’s how our team works for the organization. Other presentations about features, things like that. My team is talking customer scenarios. Specific customers. In Denver we had customers there. Real customers able to answer real questions. This is a trademark for the customer advisory team, around customer scenarios, real life. Popular. We also have our website, SQLCAT.com.
Greg Low: SQLCAT.com has things like papers providing guidance. One thing I like about it is that Microsoft does great job of bldg product, but then not in area of providing user guidance.
Lubor Kollar: There are now other teams and other products at Microsoft who are approaching us and starting doing CAT teams of their own. We are working with BizTalk, SP, Windows. Showing what and how we’re doing, communicating with users, succeeding. Exciting that we have good recognition not only within SQL Server world, but Microsoft.
Greg Low: I noted in Denver, I got impression not all the guys were based in Redmond.
Lubor Kollar: From my team, my ten people, only two are in Redmond. Need to be close to customers. One team is international, ICAT team. Two people in Asia, two in Europe, also covering India from U.S., rest of team across U.S. Close to the customers.
Greg Low: You were saying there were places where you haven’t succeeded. Limitation of product or unreasonable customer expectations?
Lubor Kollar: I’m thinking about specific project. We are working with customers through local Microsoft team. Often complex corporation. Local Microsoft team, often another MCS teams, sometimes consultant, customers, and often customers not directly not the customer, but that the customer is providing service for. Complex scenarios. I have one specific case where I know the requirements were never fully understood. Never got to us fully, couldn’t get answers though we tried and tried and tried. Got requirements only through rejections of things not good enough. Not up front. We would think we understood and we didn’t. Project management. Sometimes sales team does excellent job, customers who was lost. Encouraging. We lost that customer because we didn’t have resources. They know so much, we’re doing as much as much as possible. Warehouses with SQL Server. So expensive. Let’s talk again. Good example.
Greg Low: Your team. Deep technical skills are an obvious requirement, but as soon as you move to customer facing, I wonder about the other skill set, political.
Lubor Kollar: Great question. We need lots of those skills. Be sensitive, handle customers. Big portion of my team was in field or consultants. Already experienced communicators and handlers. Dealing with all levels of management, CEOs to DBAs. Have to talk differently to each people. Important. Also facing development team. Need to be sensitive and understand how development team works so we don’t accuse or have unrealistic requirements. Understand what it takes to implement things. Need to be sensitive on that. Yet should be able to produce good arguments to explain and get right features into the product. Every team member must be willing to travel quite a lot, even around the world. In general, 50 percent travel.
Greg Low: Yes, I know that life only too well. That’s a good point for a break. When we come back we’ll talk about the things you’re looking forward to in 2008 that will help you do your job.
Welcome back. So Lubor, outside SQL Server, what is life for you?
Lubor Kollar: You read it in my bio. I’m a good fisherman. Sometimes I catch more than I can eat. I have colleague, Boris Baryshnikov. Very nice guy. My hobby is fishing. His is making bread. He has PhD in physics and is good, but doing this specialty breads. Specialty Christmas breads and such. He likes fish. I bring fish and he brings bread. Everyone is happy. Colleagues look at us.
Greg Low: Is this fly fishing?
Lubor Kollar: I should be doing fly fishing, but I’m warm, float. Favorite technique. Three or four beer.
Greg Low: Fishing more river and ocean or mountain?
Lubor Kollar: No, lakes. Local lakes well stocked. I have summer house on one lake close by. So close I can drive to work every day. Good weather in fishing season and I fish. I ski and like hiking. It’s great here. I have website, sqlhike.com. You might know some people there. We went between two passes, about 100km. One week hike. Out of touch with civilization, no roads. Pacific Crest. Go to the website and you’ll see maps and pictures. Life outside work.
Greg Low: Then you don’t mind cold weather.
Lubor Kollar: Sunny weather, not cold at all. You’ll see in pictures, nice until last phase when we had to cross at 5,000 for 1,500 meters. Started snowing. Elevation. Other than that, most of hike was warm. We swam in mountain lakes.
Greg Low: You have professional ski instructor license as well.
Lubor Kollar: That’s for winter. Compared to east coast U.S., weather mild here, especially on west side of mountains. We have rarely sub-zero. This year we had 600 inches, like 30 meters, of snowfall. All the moist air hits the hills and dumps the snow. Good snow conditions.
Greg Low: That’s great. Just coming into colder weather here. Last time I saw you we had just come out of summer at event in Orlandia and Copenhagen. That’s the most interesting SQL Server event because it’s held in a water park a distance from Copenhagen. Water park, cold weather, but great in bldg. everyone lives in little huts. Our hut, Lubor and Katarina, my wife May and I. about five in the hut. Interesting little conference. Good fun. Interestingly cold weather compared to when we left home.
Lubor Kollar: It was there two years in a row.
Greg Low: You mentioned mushrooming. That’s one that hasn’t come up on the show before.
Lubor Kollar: Mushrooming. I grew up in Czechoslovakia. I had neighbor there who was big mushroomer. Loved mushrooms. He didn’t have children. He taught me to recognize several kinds of mushrooms. I liked it. I like to go through woods, bike, walking. Here in Northwest, great opportunities and all kinds of mushrooms. I love it. Chanterelles, I make omelets’ with those. Really good.
Greg Low: I’d be nervous of picking the wrong ones.
Lubor Kollar: You have to be careful. I don’t pick randomly, only those I know. We have here, club next to horticultural branch of local university. Meeting every month. Big mushroom show every fall. Learn there, as well as bring species for identification. Quite popular. I was surprised. Many people from Eastern Europe. Russians, Czech’s, Polish, also Italians. I’ve got a secret spot for mushrooms. In the Fall I can bring a lot in a short time.
Greg Low: That’s great. With SQL Server 2008 coming, clearly are features that will help in work you’re doing. One thing I’m intrigued by is a lot of new features seem to be enterprise related.
Lubor Kollar: I can talk on that. Marketing deciding which feature goes where. Most of the specialty features, new ones like compression, is enterprise only. Getting us on par with other enterprises, other competition. Frankly, I don’t know what is happening. Seems to me we are giving huge break to customers by licensing per socket. This is fighting back. We want to show the value, about $5,000 more for enterprise. Many licensing models, complex, must have PhD to understand. But per processor more expensive. We need to show customer value for the money. Some customers were riding wave, everything what they wanted. Not easy to answer. It’s out of my expertise.
Greg Low: With the CAT team, are you working with current shipping product or also TAP, early adopters, people like that?
Lubor Kollar: We work with both. Emphasizing more and more on 2008 and TAP. Now working on several projects known to customers which will go into production after we ship. After RTM, end of July. We have leftover work from 2005. To go back to development team with feedback on SQL Server 2005 doesn’t cut it. They’re working on next. They want to hear about 2008. Maybe the community. We do make exceptions when we do some presentations. Parts of the world aren’t as pushing forward on new technology. There we kind of cave into local marketing teams. May do exceptions, but don’t like doing that.
Greg Low: If you’re working at early stages in significant installations and projects, that also means you have good feedback to provide when being used as released product.
Lubor Kollar: Yes. My team is pushing hard to have content about SQL Server 2008. How we help the customers to adopt it, to take advantage of newest features, avoid surprises and costly mistakes. We tell customers what we know as wrong approaches. How we help with early adoption. Resource. Very popular. You need to work to get. I have one customer who is in Seattle. One feature more popular because easier to adopt and easier to show the value right away, and that’s what CIOs and management love. That’s compression. It’s application transparent. Compress databases don’t even know that.
Greg Low: Except it hopefully runs faster, takes less space.
Lubor Kollar: We knew it would have less space. But runs faster and in some cases, amazingly faster. That was news for me. I expected half. What is in play? IOs, memory, CPU. You're saving in all. We tested many times. As recent as last November we had customers here in Redmond testing. They saw improved performance. Now I have explanation. Less memory, IOs, and unless you’re 56 percent CPU.
Greg Low: Didn’t surprise me. I expected performance gains. In my work I find few systems CPU bound. Tend to be those that constantly recompile. Pretty much all the rest are disbound and/or memory constrained. Anything that reduced IO has big improvement in performance. Many more rows fit in memory, so helps there. Doesn’t surprise me in slightest that faster.
Lubor Kollar: Good then. If you tell your boss that you’re doubling database next year, buying another terabytes of SAN, often hundreds of thousands of dollars, if you compress, don’t need that. Easy to show value.
Greg Low: Any way of compressing existing data? Does it only happen when doing inserts? Does it affect data there?
Lubor Kollar: You can. Tricky. Safest way is to reboot cluster index. Bingo, it’s compressed. Tricky to reclaim the space. Working on best practices. If you cluster index, create in different file group then OK. You can do that. You can do online. Enterprise edition. Compression is only enterprise. You can do on only on one partition of table, but not online on one partition. Compressed indexing, we don’t have.
Greg Low: OK. In 2005, one main reason I encourage people to go to enterprise has been ability to do index operations online. Most on bigger end don’t have time window to do anything.
Lubor Kollar: Yes. That’s major high ability requirement. Not only that if fails and comes back fast, but more important, doing daily tasks, when loading data, maintaining data. I have customer in Europe. They haven’t rebooted index since 2006. Huge table, cluster index. They only defragment which can do online. When rebooting online, online index, exclusive lock for brief moment at beginning. Have to wait until everyone out of my table, and then lock. Change it and just short, brief moment. For a moment it’s like I drained workload. Don’t want that. For most customers, online is excellent feature but unfortunately, don’t think we stress enough. Yes, brief moment at beginning and at end, when changing state of index. Need to ensure everyone working on that index is aware of new state. Therefore, a bit disruptive. Much more problem is need to do sometimes, do on partition, online, but then people think so fast, switching out and all kinds of things. Online for partition is requirement. Otherwise, my favorite feature so far in 2008 is compression. People are getting confused. Shipping 2008 with three compressions. Two are data compressions, one is backup compression. Backup compression, you’re creating backup you can compress backup. Makes backups much smaller. Need space and time. Data compression compresses the real database. Two kinds. Page compression and row compression. Unfortunately, didn’t do good job explaining when to use which. Still working on that. Haven’t even shipped and we’re talking this. Rough guidelines. If you’re doing scale, page compression is better. Row for index. Overall, I’m finding page compression giving you more, compresses more. Only if you have problems with CPU, then investigate row compression.
Greg Low: I’ve watched many video’s on this. Could you describe differences between page and row compression?
Lubor Kollar: Page does more than row. Row compresses data by looking at one row. Missing is if I have some common things across my rows, especially if repeating same data in many rows, ending up on one page, dictionary on page, installing value once and point from other rows, using that value. That’s what page compression has. Disadvantage if I want to access rows of page, just one row, have to do much more work than if I did row compression. Row compression more row isolated. Decompress the row. Page compression, decompress entire page to get to rows. That’s where I’m paying more price. We are now holding decompress in buffer pools. Decompress bigger than that. We hold compress space, scale same page, decompress again. If row, decompress again. Accessing same page many times, working with indexes, root pages for every row access, millions of pages, thousand of rows, thousand times hitting. We do compress those. Perfectly fine. Typically non-leaf pages of index are only like one or two percent overall pages.
Greg Low: That’s an interesting point. I often see people with pictures of index structure. Great depth. Reality is when I look at index, don’t see much depth at all,
Lubor Kollar: No. one page can fan out to many pages below. Advantage of that.
Greg Low: Surprised me. I had feeling would be shallow.
Lubor Kollar: Non-leaf pages, only key. Length of that has lots to do with. Looping keys not good practice.
Greg Low: Nice in 2005 was people endlessly adding columns to key to get covering indexes. In 2005 including at leaf level made big difference.
Lubor Kollar: Excellent. That’s great point. Easy to spot. We had customer here in lab, long time ago. Took 19 minutes to do, quite complicated. I looked at plan. Doing like read lookup or key lookup. Doing here, lookups. I said no. this guy new to 2005. We changed and results amazing. From 19 minutes to 45 seconds. Cut more things. Overall, include is good.
Greg Low: Thing I found, include has been good in toolkit. Helped keep index shallow. I see people worried about number of hops down. What I find is shallow. Those pages above just sit in memory.
Lubor Kollar: Yes. You are right. Function of access and size. How to keep those pages. Non-leaf pages good. But not many of them.
Greg Low: Is there anything else in 2008 that you think will make big difference to customers?
Lubor Kollar: We see amazing stability, robustness of server. We had customers here in November. Demanding applications. It worked like charm, loop increase. We had exercises here because managers were afraid not enough testing on things. Explanation is new process through which it goes, not with one main build like 2005, now split into multiple, smaller pots. You cook there, ensure tastes well before you pour into big pot. Two levels of testing brings quality. Surprisingly good quality.
Greg Low: A couple shows back I talked with David Campbell about this process. The one alternate bit of feedback on that is that problem I see, challenge going forward, is what we’re finding is we only see features when backed. When you try to provide feedback, I often hear that it’s a good idea but too late.
Lubor Kollar: Yes. This is true that internally we’re trying to do testing on private branches to get to the problem you mention, the late feedback. Short leash. Two and one half years. November 2005, we will be two years. All that. Feedback. On other hand, try to take as much as we can, be sensitive. On other hand, shorter leashes, big chance next release as five years anyway.
Greg Low: What concerns me is where there are API related things. For example, minor things like resource governor. There’s a DMV that returns resource governor configuration. If you don’t have classify function, returns zero rather than null. I went to the guy responsible. He came back saying it would have been a lot better, but there was an internal table value of zero. He said it was too late to change because used in other places.
Lubor Kollar: This is good example of when it will impossible to change in future. Build dependencies on. Good info. I’ll try to find some way to fix this.
Greg Low: That’s the creative thing needed. A couple points. One, early point where some of the fundamental design of features, challenge to give many people access to many builds to show what you’re talking. Those that concern me are the API purity things. As you say, hard to change down the track whereas others can be changed.
Lubor Kollar: Yes. Good one.
Greg Low: Mark Souza’s role. Where does he fit? He headed up the team before?
Lubor Kollar: Yes. What happened, Mark got bigger role. I have two peer teams. ISV team. As I work with customers, this team works with ISVs of SQL Server, internal and external as well as vertical. Adoption of new release, bringing feedback, working with their customers. Similarities. Mark’s team is customer and partner team. I head CAT team. Now, I work with customers and find good way to load terabyte data fast. You can add lots of value to this info if you do more testing around the scenario. Tens of thousands of customers. That’s out of my scope. I work with customers and next customer and next customer. We have this best practices team. Takes the learning’s, doing testing, investigations, writing white papers and top ten lists and blogs, and rounding up, helping my team. Taking part of the work, this community facing work off of the shoulder of my team so I can concentrate on working with more and more customers and they take some of that and add specialists with writers. Technically capable as well. Peer team. Besides doing this, we are bringing customers to labs. This team is responsible for running those labs, also best practices. When I bring customer and data, helping me conduct the lab. One week exercise. Preparation takes months or more. Much data, hardware preparation. Challenging clients, conflict of clients. Need to prepare so we can do testing. That’s ark’s team. He’s having this team. Doing lots of this effort across Microsoft.
Greg Low: That brings us to time. Wraps up nicely. Next question, when will we see you in next month’s?
Lubor Kollar: You will see me. I’m coming to Australia TechEd, September 2. I don’t have my schedule yet. Don’t know if I’ll go to New Zealand or not. That’s my nearest plan. October in Europe, Prague. Most probably more travel between now and then. PASS conference in Seattle in November. Nearest public showings.
Greg Low: Outstanding. Thank you for your time today.
Lubor Kollar: I appreciate the opportunity and wish good luck of everyone. Go ahead and download SQL Server 2008. Next week.
Greg Low: Imminent.
Lubor Kollar: Yes. Imminent. Visit SQLCAT.com. When I feel down, not frequently, I look at SQLhike.com with its happy people, see all kinds of smiles.
Greg Low: Indeed. Thank you for that and tell Katarina hi.
Lubor Kollar: Thank you.