Why Does A SQL Inner Join Bring Performance To A Crawl?

[christappan]

So three things:
1) If you go to the query definition in Alpha and click on Native SQL you can see if there's something weird about the query being run in SQL Server.
2) the demo video in the app is using a where clause and the demo in SSMS isn't. Is the query still lightning fast in SSMS when you filter by the criteria you're showing in the Alpha video?
3) I've found that LEFT OUTER JOIN can perform a lot faster than INNER JOIN at times, so if it will produce the same records, you could try that.

[lgrupido]

Have you tried the view yet? That would take the join out of Alpha completely. (And you need to index the foreign key.)

If you would like, I could hop on a teamviewer session with you to take a look.

If you want to do that, private message me and I'll give you my direct email/phone.

[madtowng]

WOW!!!!!!!!!! that makes quite a difference.
Just to be fair, I use mysql. I'm very inexperienced with lists, but seem to do well with sql.
I created a list that uses the query shown below,
The samsaraugdrivers has 96 rows, the samsaradutylogs table as almost 232000 rows.
The time to load the data, and the responsiveness while scrolling through the data is sluggish at best without using virtualization.
Using virtualization, I get my data relatively quickly and am quite happy with the speed scrolling through the data.
I opted against pagination and turned off the Get count of Records option because I wanted to avoid having the query performed twice.

Now I need to learn how to create a search option and selectively display fields. Being able to show more data by clicking on a cell would be even better.

I know it's late to say in short, but try the virtualization setting. It really makes a difference.


2021-08-26_14-19-51.png

SELECT DL.seq,DL.logdate,DL.samid,SD.idnum,SD.drivername,DL.dutystatus,DL.statusbeg,
DL.statusfin,DL.latitude,DL.longitude,DL.samvehid,DL.codriver
FROM samsaradutylogs DL
LEFT JOIN samsaraugdrivers SD ON DL.samid = SD.samid

[iRadiate]

Rendering 13,000 records into your list is what might be slowing you down. The query is fast in SSMS. Not sure how a view would help. The list still needs to render 13K records.

Have you tried playing with the virtualization options? You can limit the number of rows rendered in the list.

Ashampoo_Snap_Thursday, August 26, 2021_08h17m28s_001_.png

I'm only bringing back 100 rows at a time in the List with pagination.

List Pagination.jpg


Virtualization.jpg

The problem is with what Alpha is doing with the query behind the scenes. Have a look at the comparison videos above.

[iRadiate]

Just to demonstrate how drastic a performance hit Alpha is introducing to the list query, take a look at this:

1. Alpha List - Query Without Join (< 2 seconds)

https://www.screencast.com/t/cGj5l5NEe


2. Alpha List - Query With Join (> 15 seconds)

https://www.screencast.com/t/pSISxYjzy9f


3. SSMS - Query With Join (< 1 second)

https://www.screencast.com/t/mh5S46yGX

[mikeallenbrown]

Rendering 13,000 records into your list is what might be slowing you down. The query is fast in SSMS. Not sure how a view would help. The list still needs to render 13K records.

Have you tried playing with the virtualization options? You can limit the number of rows rendered in the list.

Ashampoo_Snap_Thursday, August 26, 2021_08h17m28s_001_.png

[iRadiate]

Yes, definitely some good options to explore.

Thanks to all for the great suggestions.

[Lance Gurd]

It definitely helped me, views a others have suggested is a great way to go so that SQL does all the work of getting the data you want. Then there is sub-queries to get your data, in my experience these are blindingly fast in comparison to a normal join

[iRadiate]

I don't just mean that the tables are indexed, I mean the the linking field in both tables is indexed in a separate index, in your case this would be an index on MACID in the repairs table and MACHINEID in the machines table

I see what you mean, the 2 tables have clustered indexes on the primary keys. So repairs is indexed on repairid (not macid). And machines is indexed on machineid. These were the auto generated indexes by SQL when the tables were created.
So, I 'could' add a non-clustered index on repairs.macid.

What I really don't understand is why the query runs fast in SSMS but in Alpha it is painfully slow. (only when the query is changed to use the join).
However, I will try your suggestion to see if it helps, though I want to try this on my test system first not the production system.

Thanks

[peteconway]

Lance just use a view.

[Lance Gurd]

Lance, yes the tables are indexed.

I don't just mean that the tables are indexed, I mean the the linking field in both tables is indexed in a separate index, in your case this would be an index on MACID in the repairs table and MACHINEID in the machines table

[madtowng]

I've been avoiding lists because A. my first try created a horribly slow experience, and B. They seem far too complicated to work with compared to a grid.
With that said, for purposes of testing the sql statement, I would try it with a grid or even a simple (non-list/viewbox) ux.

[lgrupido]

From within SSMS, use the following:

create view myviewname as
select statement

That will create the view and then you use it in Alpha like you would any table.

Caveat: Your select statement looks like it should create an editable view, but if not you may need to do some manual updating if you intend to use this list to update records. (other than saving it in a standard detail view.)

[iRadiate]

Ditto on the index, but if it is fast in SSMS, then I am guessing the index is there.

Thought 1)
My next thought would be - do you really need all of those fields in the query?

If not, I would remove any fields you do not need. The amount of data being pulled in will have an effect on the list being populated.

Thought 2)
You should try creating this select statement as a view. So create a view with the exact same statement and then try pulling the view into the list instead of doing the join inside the list query as you are now.

You may find that to be much faster also.

Larry,

Thought 1 - I do need the fields.

Thought 2 - I've never done that, create a view. So not sure how you do that or how you use it, but I will do some reading and look into it.

Thanks for the suggestions.

[iRadiate]

Is repairs.MacID indexed, I recently introduced more indexes to my joined query tables and this helped

Lance, yes the tables are indexed.