bit Obscene: Ordered Column Store
Video Summary
In this video, I delve into the intricacies of ordered columnstore indexes in SQL Server 2022 and 2025, exploring their benefits and limitations. I argue that while these new features offer some advantages, particularly for large ETL loads, partitioning tables by date or another logical key remains a more robust solution for most use cases. To illustrate my points, I walk through a demo using the latest SSMS preview (version 22), showcasing how turning on trace flag 58008 can prevent the creation of unpartitioned columnstore indexes in large tables, ultimately guiding users towards more sensible design practices.
Full Transcript
All right. Welcome, everyone, back to the unofficial maybe relaunch of the Bit Obscene podcast brought to you by Darling Data and BeerGut Magazine. I’ve got my temporary co-host, Joe Obish here. We’ll see how he does this time around. And today, Joe is really excited. You can tell by the look in his face. Look at that face. That’s an excited man. Joe is excited to talk about ordered columnstore indexes, which, correct me if I’m wrong, Joe, that, that, that, that, I’m sorry, I have to get rid of his background. It’s making me look like an insane person. It came up with SQL Server 2022. It’s made maybe some, some progress in SQL Server 2025 that, but no one knows, because SQL Server 2025 is, of course, the unofficial AI and fabric release. SQL Server is billed third on that one.
So maybe there’s nothing good for SQL Server in it. But anyway, take it away. Mr. Robish. First of all, Oh, shoot. Here we go. I learned that my dear friend, Eric is now living under communist rule. And I thought, what’s a small token gesture that I can do to appear like, like I care, and I want to help but not really do much. And I thought, Oh, I can just show up as a guest on his YouTube channel. So that’s why I’m here. I’m, I’m, I’m, I’m, I’m, I’m, I’m in full solidarity with Eric. I’m wearing the plain white t shirt.
He’ll probably be wearing one in a couple of months along with everybody else in New York City. But I, I, I, I’m not here to help Eric. We’re, we’re, we’re all pulling for you. All right. Thanks, Ben. Hopefully, hopefully, my, my means of production are seized in a gentle way.
So for, for order to come store, Eric and I actually both blogged about this a few days of party then back in 2022. And the way I remember it was, that’s right, no clues. The way I remember it was Brent blogs about it. And he was like, Hey, I tried this and it didn’t work. And, you know, Eric and I live to serve. So we, we both took our own little look at it.
But, um, I think Eric can have the links in the description. Oh, yeah. And the viewers can vote on who, uh, did it better. You know, we can get some engagement, maybe. Of course, there’s no, there’s no voting in communism, Joe. So that’s, that’s, that’s, that’s very, that’s a good point. I take it.
Oh, back. Uh, now I’ll admit that I haven’t looked at since. I remember ending the blog post like, Oh, this is CTP 2.0 or whatever. Maybe Microsoft will make it better. Wink, wink. Now they, they like never do. Um, but I haven’t looked at it. I’m assuming it’s not better. If, if they did make it better, feel free to let us know it in the comments.
Now, in terms of what, okay. So why does ordered columnstore exist? So there’s a very old workaround for columnstore where you create a clustered index on the row you want to order by first, and then you create a columnstore index with max. That one second, the idea there being, I was like an, it’s like an unofficial way to get like a, an ordered columnstore after you built it.
That’s where you create the clustered index on the table on some column or key columns. And then you do the clustered columnstore index with drop existing. Yeah. Yeah. Yeah. Okay. And like, you know, like that could be improved. You’re writing the data twice. It’s a bit of a annoying workaround. So I guess like from my point of view, you know, getting rid of that workaround could be a good thing. Um, because now you can just create a, just create a clustered columnstore next with ordering.
Problem is, and this is the thing that Brent blogged about. It’s not always ordered, which, which is not too good start. Um, but like, I’m even gonna set that aside. So I found that when you add ordering, SQL Server is making three changes. First being when you create the index, it’ll add a sort operator. It’s a soft sort. Um, there are all kinds of reasons why the data isn’t going to be fully sorted.
When you’re doing an insert in the table, it’ll again, add, add, add a, add a sort. And then once again, the data might not actually be sorted or it’ll be sorted in an intuitive way. And then when you’re doing a rebuild, it once again, adds a sort. And we’re a three for three and that sort may also not fully sort the data. So that’s what the feature does. Um, my big beef with it is how it changes the, the insert queries, but you had a comment here. Oh, I was gonna say we’re now at three sorts and actually no sorting. Yeah, yeah. So.
All right, I’m gonna ask you a non trick question. You ready? I know what that means. Why are clustered indexes generally useful?
Why are clustered indexes generally useful? Uh, well, uh, there, there are lots of reasons why they’re generally useful. Looking for the simplest, the simplest answer.
Well, it, it gives you a free access to all of the columns in the table in an ordered fashion. Yes. Yes. It’s because the data is sorted, right? The data is sorted globally. If you insert some old data, according to your key, it’ll do the, you know, page splitting and it’ll, the data is always sorted.
Now, what if I told you that we created a new type of index and it was kind of sorted. Sort of sorted. And whenever you inserted new data, it would sort the new data you inserted and append at the end of the table.
And it would, it would, it would keep doing that. So whenever you do an insert, it’ll locally sort that data and append it at the end. Does that sound very useful to you?
Let me, let me just make sure I have the right mental model of this. Let’s say you have a table with a million rows in it and it’s the numbers one through 1 million and they’re in order. And then you insert 10,000 rows to it.
Instead of those 10,000 row, let’s say they’re, they’re also the numbers one through 10,000. So they fit in with the numbers one through 1 million. So instead of the numbers being dispersed in the table where they should be, it would just be like an appendage on the table with the rows in order from one to 10,000.
So it would be like one to 1 million and then one to 10,000. Yes. Yes, that’s correct.
And that’s, that’s the fundamental problem. From my point of view with order to columnstore, like I find that kind of worse than though we’re going to not sort data always for you. Other people could disagree because you know, like it’s, if you have clustered indexes, you have your, your, your AK pages.
They can be split all kinds of great things can happen. You have columnstore. You have your compressed row groups.
Like there isn’t any way for, there isn’t any reasonable way. Like if you’re loading a million new rows, like it’s, it’s not going to find the most appropriate compressed row group uncompress it, add the new data and then recompress it. That that would take forever.
Like it’s just, it’s just not going to do that. Like it’s going to append the data at the end. I’m surprised though that on rebuild, it doesn’t like fix that. It could fix it.
It could. If, if, if your sort manages to work fully. Yeah. Okay. Great. Good. Good. So what, what makes us sort work fully or not? Uh, here’s a number of things, how much memory you have, how much data you have, what DOP you’re at.
Um, which brings me to something, which, man, I, I, I have these great notes that I’m like not reading. So, you know, like one of the, uh, like one of the annoying things about the old workaround to get like pseudo ordered column stores before 2022, you know, the cluster index and CCIs. You have to create the CCI at max top one, which, which could be slow for a big table.
Free the, free the documentation for ordered columnstore. It says, if you create the index in parallel, it’s not going to be ordered. And in fact, if you want perfect ordering, it.
Max top one. Yeah. Max top one. So like, so are you better off just doing the old message, the old method rather? The, the, the problem.
I view the sorts during insert as not that useful or, or even harmful in some cases. And you, you get this partial to full store when creating the index, but like the penalty for that is you, you lose online rebuilds. Hmm.
At least on 2022 and 2025, they’re adding online rebuilds or the columnstore. Okay. So if you asked me if it was worth it, I would say you should just partition your tables, but I’m going to try to make my case a bit better. Um, on this subject of insert.
So let’s say you, you have an ordered comp store index. Cause. You’re in 2022. Yeah. Use all the new features as soon as they come out. Read the documentation.
You’re hot on the heels. Why not? Yeah. Yeah. Why not? Now, if you’re inserting less than a million rows per partition, what’s going to happen? Well, it’s going to sort the data pointlessly. Yeah.
And that’s going to try to compress the new single row group. If you get some kind of, uh, what’s it called? Uh, pressure, dictionary pressure, memory pressure, whatever, you can get multiple row groups. But like in general, like that, that, that sort’s not useful.
Like if it’s, it’s, it’s, it’s, it’s almost like sorting, like a single, like a single 8K page before you insert the data, which maybe it happens, but you know, like, you know, like, like, it’s, it’s not doing anything for you. Now you want me to say, all right, well, what if I insert more than a million rows? What if I insert 10 million rows of time?
Well, It’s an ambitious insert. First of all, if you, if you, if you have like, you know, like, like a normal ETL load where you’re like, like, like a normal ETL load. Like yesterday’s data, it’s going to be like new data probably.
Right. So the thing you want to order by is probably for like yesterday’s day. So it’s, it’s not gonna matter in that case, or like, let’s suppose you’re loading a week of data at a time. Cause you’re, you know, you, you, you, you live under communism.
You’re not going to work every day of the week. Right. Right. So in that case, you, you, you have a week of data, 10 million rows. So it’s, if it’s ordered, then your row groups have like one to three days of data instead of seven.
And if you happen to run a select query, which is like, you know, looking at a single day in that range, then maybe you read a couple of fewer row groups. So it’s, it’s not, it’s not, yeah, maybe like, that’s not really giving you a whole lot in terms of performance. No, but I think, I think that gets, I think that gets back to your point though, about if, if this is something that you care deeply about, regardless of ordered columnstore.
Like, but if your table is truly that large and you’re truly, truly doing that sort of ETL with like what you’re talking about where there’s daily loads. And I think that’s, truly the most sensible thing to do is have that table partitioned by some date that, that, that gets you like, I mean, maybe, maybe not like from a data loading perspective, but from like a, like querying that data later perspective. That at least gets you the partition elimination and then segment elimination based on, you know, the criteria for your query, looking for data for some date range.
I agree fully. And I’m going to jump into my demo ahead of time. Ooh, my, my, uh, four notes.
They’re just, they’re just basically just, you know, the, the, a cast of Erik Darling is just down the memory hole. Yeah. All right.
Can you see the beautiful modern SSMS? And I can’t, is that 21 or 22? It’s the newest one. That’s SSMS 22 preview. It’s glorious.
It’s glorious. Look at it. Oh, it’s not 21. 21.6.1. Wow. I’m already out of date. I guess. Let’s do that. Check for an update too.
All right. All right. Um, I, I, I, I, I, I, I, I hear this stuff. All right. So what’s that trace flag? It looks, it looks, it’s that trace flag, Joe. Microsoft has heard my, my, my desperate please.
Yep. On the subject of partitioning is super important for columnstore. So what was, so why do I say that?
It’s the same reason you said it. If you, if you, if you partition logically, then that’s going to force all the word groups together. It, it, it basically gives you like, like, like, like a minimum sort of segment elimination free.
Like no matter how badly the developers load the data. If it’s a row at a time, it’s a hundred thousand rows at a time. If, if, if they don’t load the data correctly the first time and they delete it and update over and over again, they just still in that partition.
Even if you don’t get a partition elimination, you can still get segmentation. I mean, I’ll go further and say that I’ve never had a columns or forms issue get resolved by, well, you know, we, we had bad segment elimination because the data wasn’t ordered.
So we were reading like, we were reading like 20 row groups and then we made the segment elimination perfect. And now we’re reading three and performance is like amazing.
Cause we, we, we got rid of those 17 compressor groups. Like I I’ve never seen that. The things I’ve seen are you have a big table. It’s not partitioned at all.
Data is basically in a random order. You know, you can get hurt there. Um, or you have like a billion soft deleted rows. I never got cleaned up.
I’ve seen them production, you know, that’ll hurt you. Or even just like, like all of the normal performance issues that you can find with really any query.
Like for me, it’s, it has never been like I’m partitioning my data by quarter or whatever, but well, my data is too unordered within the three month window. And like, I really need that perfect segment elimination.
Like I’ve never seen that. Maybe there’s somebody with high frequency trading or online gambling, or who was like super hardcore. And that they really need those perfect pristine row groups.
In which case I would think that, you know, a sort that might not start isn’t good enough. Definitely not. And you, you gotta fix your UTL anyway.
Anyway, I’m a big fan of partitioning and Microsoft has heard my fleet, my please. They gave me permission to introduce this new trace leg 58008. We go ahead and turn it on.
All right, it’s turned on. And now I’m going to try to create a columnstore index, a big table, which I’m creating an, an, an unpartitioned index.
I’m trying to, that’s, that’s, that, that, oh, look at that. This is not safe for children. Look at that. The operation failed because the table is too big. Like, create a partitioned commister index instead.
Finally, SQL Server is finally stopping us from, from making dumb mistakes and bad decisions. Isn’t this great? This is fantastic.
Now, you know, I’m sure, you know, like, now, like, you know, maybe your DPA thinks they always, they always know better or, you know, like, it’s, it’s important to be able to, to, you know, have full control of the software.
So we had this new create bad index mode, uh, option. And if we, if we set it to on, then, uh, we can, uh, avoid Microsoft’s guardrails. So I, I’m really excited about this new feature.
We can finally, you know, encourage people to partition their damn tables, which I think is really, it’s probably like the, I’d say it’s the most important thing if you’re doing columnstore at any like reasonable size, you got those deals partitioned.
Great performance, ETL, maintenance. There’s just so many reasons to do it. And, uh, I’ve seen what happens when you, when you don’t do it and it is not, it is not pretty.
Yeah. I mean, uh, you know, I don’t, I don’t do a ton of production work with, uh, with column store. I do a bit, but, um, you know, most of the, most of the tables that I run into that are, are, that are column stored are well beyond like, you know, my ability to, um, have my consultant way with them and like, maybe go back and partition them or, you know, do anything useful in that way.
Um, you know, I wish, I wish more people would involve me in like the implementation of something rather than like the, Hey, all of a sudden this sucks. I think I’m almost like, well, guess what? At 3 billion rows, we’re a little past the point where we can easily do this. So, um, you know, not, not undoable, but certainly, uh, what makes, makes the experience a little bit more painful.
But, um, you know, a lot of what I run into is, um, people who have, uh, you know, like row store partition tables, cause they bought into the, the meme that, uh, you know, partitioning tables makes queries faster with Rome, with row stores indexes.
And boy, are they surprised when they start doing things like men queries and max queries, and even some, like, you know, some other, some other types of queries, all of a sudden they, they, they can’t just like, like start at one end of an index and get something like they, they, all of a sudden they have to scan every single like index and like the, like all of the partitions.
And you’re like, again, like, like, why didn’t you call me like, you know, five years ago when you first got this bad idea in your head. So for the, for the folks out there who have perhaps large columnstore tables implemented, but did not have, uh, the foresight and certainly, uh, at this point may or may not have the high hindsight to, uh, to partition their, their large clustered columnstore tables.
Uh, is there any hope? Is there anything they can do to make their lives better or easier? Or do you just recommend creating the partition columnstore table and loading data over and doing the old SP rename switcheroo?
I don’t remember how I did cause it was so long ago, but we had that same problem where we had like, you know, we had 300 customers at the data warehouse, nothing was partitioned. Some tables had billions or tens of billions of rows.
And we did ship the code to make that happen. Um, how did we do it? I don’t remember how we did it.
Uh, I, I, I think we just did the usual scam of, oh, you know, you’re now in a maintenance window and we’re going to rewrite all of your data. Um, I think we had like a lot of parallel jobs doing the inserts.
Sure. There is, there is some trickery there. It, it, it definitely wasn’t easy, but it’s, you know, if you want to make the best long-term decision, you gotta have partitioning.
It’s only going to get worse the longer it goes on without partitioning. Um, in terms of this new feature or columnstore, I mean, maybe it’s better than nothing. You know, there probably are some use cases where it’s okay or bar than status quo.
I do think it’s kind of fundamentally flawed and that, you know, as we talked about your word groups already compressed, you’re not going to uncompress them to get perfect segment elimination.
And that’s just like an unsolvable problem pretty much like, you know, the, the, uh, new data you load at best. It’s going to sort that, but it, it, uh, might not even do it. You know, you might get a partial sort, or if, if, if, if you’re doing parallel insert, it’s just, it’s just not going to work in the way you’re expecting.
That’s covered in the, in the blog post I did. Yep. So I’m looking, I’m looking at the 2025 notes here. And, uh, apparently Microsoft has really expanded its repertoire of, of 75% complete features by adding ordered non-clustered columnstore indexes.
So we’re very excited for that. Aren’t we? Yeah. Yeah. That’s a big, that’s, that’s, that’s going to change, change the data landscape forever.
I was actually trying to fathom like why order to columnstore even exists. Uh, I do remember hearing a rumor about the, the, uh, SQL variant data type where like some big, important customer gave Microsoft a really big bag of money and then Microsoft implemented the SQL variant data type just for them.
And all of us could depend on it. So like maybe it was that, uh, maybe there was some customer who was like, Hey, you know, we have, we got this big problem.
We really need to order to come store. And they managed to convince them with, without the big bag of money. Yeah. Maybe I have a third theory though, without, without any factual basis, what’s the thing whatsoever, but I feel like it’s my favorite kind of theories.
So I’m just going to throw it out there. Like they’re like, really there, there, there are like two things that, uh, DBAs love. Right.
First one is a good clean demo. Second one is attempting to remove fragmentation. And the problem with calm store is it offers you both. Cause like, you know, like, like what’s the easiest possible demo to do with calm store?
Oh, I load my data. Like, uh, like a total dumb ass. And now I’m not getting any segment of elimination. I can, I can, I can view the little print screen and it says segment with eliminate zero segments red 99 or whatever.
Yeah. Then I heroically load my data the right way. And now it’s ordered.
And now, you know, the, the, uh, little numbers and the segments, you know, I’m, I’m skipping segments, doing less work. Like, you know, like I’ve done that demo. You’ve done that demo.
Like everyone’s on that demo. It’s really easy to do that demo. Yeah. You see all these blog posts talking about how great and easy and important. The elimination is nobody talks about partitioning. So, so like, you know, like, like, like the, the important thing, do the ease of demoing it, become a single elimination.
And then there’s just anything, right? Who doesn’t like anger or fragmentation? How many people are still rebuilding all their indexes every night?
Too many, too many, too many Joe. You gotta get rid of fragmentation. Right. So I feel like there’s this way too much importance on that subject. Um, I mean, I, I already said, practically speaking, what I think people should do.
I think partitioning is, is way more important and, you know, like it’s, it’s, it’s even worse. Cause like, I don’t know, like I I’ve seen playing questions like what comes or indexes aren’t ordered. This is totally useless.
Microsoft needs that ordering. Like I’ve seen that come up a lot. And, you know, like, like maybe finally Microsoft listened to the community when they shouldn’t have. And now we had this super half baked feature that can like never truly work, which source the data sometimes and removes online index rebuilds and ads point in the sorts.
And, and so on. Um, you know, it, it, pointless sorts are like, I don’t know. I think a good chunk of the SQL Server source code.
Um, so in, in, in terms of what I want, I already showed what I want the, uh, magic new trace lag. So that’s, that’s great. Look forward to that.
Um, I kind of feel like maybe like, uh, you know how there, there, there’s the sword and, uh, temp TV option, which only applies during your next creates or rebuilds. Yep. I think that would have been better.
Like, Hey, you know, you, you’re creating columns or index, like say which court, which column you want to order by, it’ll make a. Attempt to do it. Mm-hmm . And it’s not going to fundamentally change every insert and it’s not going to remove online rebuilds because it’s just something that happens like once during creation. Mm-hmm .
And, you know, if you’re the kind of guy who’s like rebuilding your indexes every night, well, it’s going to be offline now. Like maybe that works for you. Maybe it doesn’t. I don’t know, but.
At maxed up one. Right. Yeah. Like, I don’t know. I, I don’t think. I don’t think this is a very good feature. And if, if you read the documentation, like there isn’t anything even about inserts, you know, it’s just all like, Oh, we’ll, we’ll sort the data and, or maybe we won’t, but even partially sorted data will improve query start performance.
Like they make it sound like a no brainer and. I, I, I, I just think it’s, it’s, it’s, it’s fairly flawed, you know, do the whole. Once again, we’re not going to uncompress row groups and, you know, slice and dice and surgically insert things.
No, I, I honestly like this should be, this should be one of those things where like when you use it. Uh, like. So there, there are other database systems where like you can.
Uh, like create a table and partition it all in one go. Right. Like, like SQL Server, you have to like create the partition function and partition scheme other databases.
You can be like create table with like, and like has like this, like with partition as syntax type thing, which is like, like when you write a window function and you’re like partition by blah, blah. Like you can write the table is like partitioning by that. And you can specify like the column and like the partitioning like ranges for it.
Uh, I think that. With ordered columnstore. Uh, like if it’s not on a partition table, like it should force you to choose a, like, like whatever columns you’re ordering. By it should force partitioning on those columns.
Like, I don’t care. I don’t, I don’t care if it like, like, I don’t care if it makes a stupid choice doing that, not doing that as a stupider choice than any choice you could make in that. Like just.
Bucket things somehow. Like just like why. That’s interesting. I mean, I don’t, yeah, you know, it’s. I think they’d be scared to try to pick partitions for you because you know, of course they would, but yeah.
But yeah, no, it’s, it’s definitely a good point. Possibly even superior to my, you know, you can be able to trace that to tell you when you’re being a idiot or not. So how did, how does that trace flag decide what a too big table is?
That’s, that’s, that’s a, under NDA. So, okay. It’s proprietary. All right. Gotcha. Gotcha.
Well, maybe, hopefully someday, uh, you’ll, you’ll be able to write anonymously on, on some blog about the thresholds for, uh, when a column is too big. Well, I mean, so, so, so the great thing is if I violate NDA on your blog, everyone should think that you wrote it. So I want to get in trouble.
That’s true. I won’t get in trouble. I won’t get in trouble. I don’t have an NDA. What NDA do I have? Oh, okay. All right. Can’t sue me if there’s no paperwork. Okay. All right. That’s, that’s, that’s, that’s the defense you’re going with, huh? Yeah.
Okay. I was drunk when I wrote it. I was just guessing. Uh, I didn’t sign any paperwork unless I was drunk when I signed the paperwork. So make your own decisions there. Well, I mean, I understand why the NDA is certainly annoying.
I remember I was once told by, uh, uh, Dr. Now that, you know, here’s some information that’s under NDA, but I already knew it. Yeah.
But you’re gonna tell me something I already know and now it’s under NDA and I can’t tell the people. No, like I already, I already had that information. I already knew it. That’s why you have to, that’s why you have to blog about everything.
Yeah. Believe it or not, we, oh, let’s see what you did there. Yeah. Believe it or not, we can figure out things on our own sometimes. So what, so the real trick of what to do is to write the blog post, but put this, the posting date is like a month before anyone said anything to you and then publish it and then be like, well, when did you tell me that? Oh, such and such a date.
Well, this post was published a month before that. It’s not covered by NDA. I bet everyone who’s watching this is learning a lot, getting a lot of good tips. Look, there’s one thing I’m good for.
It’s sound legal advice. So that was, that was all my complaints about, or account store. Um, I don’t think it’s very good. I think you should partition your tables and that’s going to do way more for you.
So, but I mean, you know, maybe there’s some workload out there where this is the best thing since that communism and they’ll get a lot of value out of it. Well, I think the only thing that we can say about this feature that it truly is for the people, it’s free kind of for $7,000 a core. It’s free.
Maybe not. All right. Yeah, I’m gonna, I’m gonna have to workshop that one. But, uh, I don’t, I don’t have any specific complaints about ordered columnstore. Uh, you know, my, my complaints are far more generalized than that, like, you know, like, okay, you’re gonna spend development time and cycles on another feature that’s like, you know, 60, 70% done.
And like, try to get us excited about it. And then, you know, they trot poor Bob Ward out to another conference and everyone’s like, Oh, like, what are the internals like get the debugger Bob. And then poor Bob has to sit there and be like, there’s no, there’s no debugger for it.
All right. Sorry. Like, you know, it’s just like one of those, one of those things. It’s like, we’re supposed to get excited about like this, this, this thing that, that, that is like not solving a problem for 99% of the SQL SQL Server. So I think the analogy you wanted before was, uh, you know, think about a, a, a parallel rebuild of an order to come for index.
You have all these individual threads. They’re all working together, but by the end of it, they haven’t accomplished anything. Just like communism.
Man, you’re good. You’re going to get me like flagged in the EU with this one. I’m, I’m going to, I’m going to get, I’m not going to be a big fan of the EU. You’re going to be allowed to travel internationally when, when this, when this goes live.
No, it’s, it’s the, you should have it go live after you’ve, uh, fled the earth. Uh, well, maybe, maybe Elon will take me to Mars. I can open up a saloon somewhere in a gully or that’d be nice.
Probably not have a lot of demand for SQL Server consultants on Mars. But there will be for saloons. Cause it’s me, the new wild west, nice place.
You can, you know, tie up your robot horse, buggy thing, go in for a nice cold beer, Mars, Mars beer, and take a load off. You know, we would be like just the ultimate. We’ll have one of those pianos that plays itself.
Be great. Got a whole vision here. Okay. No, my original, my original plan to open a bar on the beach in Bora Bora. I think that market is saturated at this point.
So I’m going for a saloon on Mars. I wish you well, you know, I don’t drink as you know, it’s true. So I’m not going to support you.
That’s why you’re going to be the perfect bartender. I can trust you not to drink my whole supply of Mars beer. Unless you know, Mars beer is just so delicious.
You change your mind or you get bored on Mars. Then again, you’re not, you’re not bored in Wisconsin. So you can, I don’t know. I don’t know where you could get bored. Sorry.
Did I give away too much personal? Oh yeah. I actually thought you stopped recording a long time ago, but apparently you’re still going. Oh no. Do you want me to stop recording? Yeah. Oh, all right. Cool.
All right. Well, thank you everyone for tuning in to the mildly resuscitated bit obscene podcast with my temporary cohost, Joe Obish. As always, this is brought to you by Darling Data and the kind folks at Beer Gut Magazine who supply me with stickers and magazine covers.
So thank you. And I’m going to hit the button now. All right. But I don’t know the beauty of the light in the internet. Now, I mean, we can do continues. Coming.
Going Further
If this is the kind of SQL Server stuff you love learning about, you’ll love my training. Blog readers get 25% off the Everything Bundle — over 100 hours of performance tuning content. Need hands-on help? I offer consulting engagements from targeted investigations to ongoing retainers. Want a quick sanity check before committing to a full engagement? Schedule a call — no commitment required.