How to start a performance recording in Tableau Desktop: Quick Tip

0:00Hey, it's Tim here. In this video, I'm

0:02going to show you really, really quickly

0:03how to start a performance recording, and

0:06then get the performance recording workbook

0:07open. And we're just going to have a quick

0:09look at it. It's not a detailed dive deep

0:11dive into the performance recording work

0:14book. This is just showing you how to do it

0:15in desktop. I'll do another video how to do

0:17it in server as well. But for now, let's

0:20get stuck in. Okay, so here I am in Tableau

0:22. All I'm going to do is I'm going to open

0:24the sample workbook here on the bottom left

0:26. If you don't see these, don't worry, just

0:28grab any workbook whatsoever.

0:29You can get hold of, it's going to work

0:31exactly the same. Now, when you click on a

0:33workbook, when you open up a workbook, what

0:35it's actually doing is it's opening up a

0:37couple of bits of information, it's opening

0:38up the workbook file, which is essentially

0:40just XML, then it's communicating with the

0:43data source, it's querying the data source,

0:45then it's loading it. But once it's loaded,

0:47something called the vis qL engine actually

0:49renders all the visualization and puts all

0:51the information you want on screen. So that

0:54's essentially what happens when you open a

0:54workbook. Now, when you're troubleshooting

0:58a workbook, you're going to see that it's

0:58going to be a little bit more complicated.

0:58troubleshooting the workbook, sometimes

0:59maybe you've got slow

1:00performance, a performance recording is

1:02actually a really

1:03good way to sort of diagnose what's going

1:05on. Because what

1:06happens is essentially your tablet could be

1:08building a

1:09really complex query, or you could have

1:11created a calculation

1:12that's causing absolute nightmares to

1:15databases, and all

1:16your tablet workbooks. So how do you find

1:18out how long each of

1:18these behaviors is taking. So for example,

1:21when I click on

1:22this little action over here, how do I find

1:24out the

1:24performance of that interaction, when I go

1:26over here to the

1:27right hand side, and I start filtering on

1:29dates, how do I

1:30find out the load performance of that

1:32particular interaction, I

1:34could just look at it and sort of come to a

1:35conclusion that

1:36it's fast. But how do you get some sort of

1:38quantitative

1:39numbers essentially for this behavior? Well

1:41, it's actually

1:42hidden up here in the Help menu. Now

1:43depending on whether you're

1:45on Windows or on a Mac, it's looks slightly

1:47different, but

1:48it's essentially in the same place, I'm

1:49going to head up here

1:50to the Help menu, he's just behind my

1:51little annotation tool

1:52here. So if you go to help, and you'll see

1:54here, there's an

1:55option to go to Settings and Performance.

1:58Once there, at the

1:59very, very end, there's something called

2:01starting a

2:01performance reporting. And here's the

2:03important thing. Once

2:04you hit Start, it started, you don't get

2:07some sort of visual

2:08callback on the screen to say, hey, it

2:10started, you don't get a

2:11recording icon, nothing happens. You just

2:14click that button. And

2:15now it's kicked off. There's nothing on

2:17screen that will tell

2:18you that it's kicked off, you just have to

2:20trust it's kicked

2:21off. Okay. So now that you've done that,

2:23when you go back to

2:24help, the only way to know that it's

2:25actually kicked off is this

2:27will now say stop performance recording.

2:29Okay, so that's how

2:30you know, everything's kicked off. But for

2:33now, we've just

2:34started it. Now we want to do some sort of

2:36interactions. Now

2:37what I've done is when I loaded this work

2:39book, I purposely

2:40didn't go to other tabs. So you can see the

2:42load performance of

2:43those tabs, I'm going to go here to product

2:45, and it's going to

2:46have to load it. And then I'm going to go

2:48to customers. And

2:49it's going to have to load that too. And it

2:50's going to go to

2:51shipping here. And it's loading that to

2:54performance, it has to

2:55load that to forecast, it loads that very

2:57quickly. And the what

2:59if forecast is as such, now that I've cycl

3:01ed through all of them,

3:02I should have some information in the

3:04recording about each of

3:05those. Now what I'm going to do is go back

3:07to customers, and I'm

3:08going to just sort of interact with the

3:09filters just by

3:10deselecting all. And this time, I'll just

3:12select furniture, then

3:14I'll select technology. And you can see

3:15that the animation and

3:16everything sort of kicks in and select

3:18office supplies again,

3:20and that kicks in again, last thing I'll do

3:22is a deselect

3:23corporate, then deselect home office, then

3:26deselect consumer,

3:27and then select all again, just to bring

3:29everything back in. The

3:30reason I'm doing this is because some of

3:32these queries are going

3:33to be loading from the data source. And

3:36some of them are

3:36going to be cached. Essentially, a cached

3:39load is like Tableau

3:41has already got the details. It's already

3:42remembered. So for

3:44example, when I go to the shipping tab,

3:46this is a cache

3:47load because it already had information

3:49about that saved in

3:50its memory. So whenever you go back to a

3:52view that's already

3:53been previously loaded, it's not loading it

3:55from the data source,

3:56it's just loading it from its own memory.

3:58And now that we're

3:59done that we go to help, we go to settings

4:01and performance, and

4:02then we hit start performance recording.

4:04And when you do that,

4:05nothing will happen immediately. But just

4:07trust that Tableau will

4:09eventually open a workbook. If you've got a

4:10slow computer, this

4:11will take a while depending on how big your

4:13performance query

4:14was. But here we are, we've got the

4:16performance recording ready

4:17to go. And what you can see is you get this

4:20sort of it's not a

4:21greatly designed workbook, but it's

4:23actually pretty good. What

4:24you can tend to do sort of take this

4:26information out and do your

4:27own thing with it. So just stop this

4:29looking confusing, what I'm

4:30going to do, I'm just going to close the

4:31actual one that I was

4:32using for my performance recording and make

4:35this large.

4:35Now, something people often don't realize

4:38is that this

4:38filter here at the top is set sometimes all

4:41the way over here.

4:42It's not always set exactly to the far left

4:44. And the reason

4:45that is, is because some interactions are

4:47so fast, they

4:48don't even sort of clock anytime. So you'll

4:50see here that

4:51they set it to just 0.01 like that. And you

4:54can actually see

4:55the number of things that come in disappear

4:57. So those things

4:58that sort of really sort of don't matter.

5:00But if you want to

5:00see everything, just slide that to the left

5:02. And you can see

5:03this list of activities gets a lot longer.

5:06Now, as we scroll

5:07down, you can actually see the things that

5:09took the longest. So

5:10connecting to the data source took 0.62

5:13seconds, that was

5:15actually the longest activity, rendering

5:17the visualization was

5:18the next sort of most sort of intensive

5:21tasks. And then as I

5:23scroll down, what you're typically actually

5:25analyzing is

5:26the query performance. So if I was to just

5:28go in here, and

5:29select query performance, and let's just

5:32say keep on there and

5:33see if this applies to the whole hub

5:35unfortunately doesn't fill

5:37to the does that's really annoying. It

5:39doesn't fill to the

5:41charts on the left. So let's see if we can

5:45set this to do that.

5:48So it will actually highlight it, but we

5:49won't actually see

5:50it. So I've managed to find one here, which

5:53is good. So now if I

5:54click on that, you'll actually see the

5:56query. And so what

5:57essentially is happening is here, if you're

5:59connected to a

6:00live database, let's say SQL Server, Oracle

6:03or Snowflake, the

6:04query that Tableau is sending to the

6:06database is going to load in

6:08this box, okay. And this box here that we

6:10're coming from was

6:12this box here that we're drawing from is

6:14essentially just

6:15ranking how long each activity took from

6:18slowest to the

6:19fastest. Now, you will register times that

6:22are absolutely

6:23ridiculous in here, because you for example

6:25, this took 0.00.

6:26That's just because the query was

6:28exceptionally quick. So

6:30that's actually probably why they load this

6:32workbook like

6:32this, because they're trying to filter out

6:35sort of in main tasks

6:36that don't really take any any sort of

6:38amount of time. Now,

6:39this detailed view is actually quite a nice

6:41new addition. What

6:43it essentially does is it basically shows

6:45you a

6:46chronological sort of order of what's going

6:48on from the moment

6:50you started the step the test until

6:52basically the finish. So

6:53here you've got different depths, and I

6:56believe the depths

6:57are just sort of different tracks. Think of

7:00them actually

7:01says here the fact the depth view shows the

7:02high level

7:03activities at the top activities lower down

7:06are spawned by high

7:07level activities. So essentially, whatever

7:09you're

7:09doing in row one is activating some of the

7:12subsequent rows and

7:13so on and so forth. And so stuff at the

7:15bottom here is sort of

7:16activated by stuff at the top. So the

7:18performance recording

7:19here, and you've got on hover action, this

7:22this performance

7:23recording actually started back there and

7:25stopped there. That's

7:26why it's sort of a really, really long line

7:28. It's not

7:28because it's taking up a lot of resources.

7:30It's just a Gantt

7:31chart, showing you what's going on. The

7:33render activities, these

7:35are the ones at the top executing the

7:36queries sort of

7:37303 milliseconds or something like that or

7:41activities,

7:42although the processor time sorry, here is

7:450.9. So all of

7:48this information is available to you. Now,

7:49this is not an

7:50interesting workbook. It's just a normal

7:52one. If we were to

7:53connect to a live database, this this

7:55particular sheet would get

7:57more interesting because you'd get more v

7:59itals about what's

8:00going on. Anyway, I've ranted enough.

8:02Anyway, I've rambled

8:03enough about this, hopefully you find this

8:05useful. I just wanted

8:06to show you how to get at the performance

8:08recording. And then

8:09maybe in another video, we'll do a deeper

8:11dive into performance

8:12recording itself, and how to optimize

8:14performance for

8:15workbooks. So nevertheless, that's how to

8:17open up the

8:18performance recording window. Now you're

8:20done, you can

8:21actually save this workbook if you want to.

8:24And you know, take

8:25it wherever you want. Another thing people

8:27do is actually

8:28don't bother saving the workbook, but they

8:29take an extract

8:30of this particular data source. And what

8:33they do is they

8:34memorize the particular thing that they

8:37were doing. So they'll

8:38open the workbook, they'll click on a

8:40specific set of different

8:42things. And then they'll capture all the

8:44details. And then what

8:45they'll do is they'll come back to that

8:47workbook once they've

8:47performance optimized it, do exactly the

8:50same steps. And

8:51essentially just see how long all the

8:53activities take and see,

8:55okay, what have they optimized? What have

8:56they reduced? And sort

8:57of compare the two tools like Tableau prep,

9:00you can absolutely

9:01go down and maybe interrogate that further.

9:03But nonetheless,

9:04you sort of don't need to really worry

9:06about that. This is enough

9:08really, this workbook is enough to do

9:09everything you need to do.

9:11And yeah, you can just go ahead, save it,

9:13we'll save it here as a

9:14demo. Performance recording, I'll save it

9:17to my desktop demo

9:18perf recording. And now that's done, we're

9:21going to hit save.

9:23And that's it. That's a performance

9:25recording done. And

9:26we can call that a day. Now I've obviously

9:28still got the actual

9:29workbook open. If I go over here to the

9:31right hand side, you'll

9:32see that it's there. But that's it. You can

9:34do this on Tableau

9:35server. I'll show you that in another video

9:37. But for now,

9:37thanks for watching, and I'll catch you in

9:39the next one.

9:39Transcribed by https://otter.ai

9:43(upbeat music)

9:46start a new wave.