Data Visualizer embedded content on the BI Answers dashboard taking too long to render initially

Dhaval Parikh Stantec

We are currently on OAS 7.0 and have a workbook built in data visualizer. We have embedded this DV workbook in BI answers dashboard and it is working fine. However, the problem is that there is an20-25 seconds lag before the contents starts to appear. Can someone throw a light on how can the problem be fixed? Are there any dashboard settings that needs to enabled to counter this?

Ashish-Oracle

@Dhaval Parikh Stantec Does the workbook open without any delay when opened directly from catalog and only has a delay when you try to open it from the Embedded Dashboard? If this is the case, did you perform a comparison between the HAR files for both the attempts and notice anything?

Dhaval Parikh Stantec

@Ashish-Oracle The dv workbook is loading super-fast. There are no problems with the DV workbook. But when the same workbook is embedded into the dashboard, there is an initial lag before the content shows up on the Dashboard.

Ashish-Oracle

@Dhaval Parikh Stantec Can you refer HAR Analyzer

, generate a HAR file for your attempt from the dashboard and check where the initial delay/lag is?

Also check if the delay is consistent across different browsers?

Dhaval Parikh Stantec

@Ashish-Oracle Thanks for your prompt response! We will certainly look into it. Can you think of any dashboard properties settings that might be contributing to this delay? Can agent caches be used to make it load faster?

Gianni Ceresa

Can agent caches be used to make it load faster?

Until you find out what is taking time, don't try to find if X or Y can make it faster ;-)

If the slowness is because of SSO between "classic" and DV, and the network calls between the two (because embedding involves all that), you can maybe try to find if there some kind of caching exists to make the SSO piece faster.

But before to look for how to make it faster, you need to find what makes it slow.

The dashboard doesn't have properties to make embedding faster, because embedding is just loading "something else".

Dhaval Parikh Stantec

@Ashish-Oracle This request seems to be taking the most amount of time. There is also a time warning for this. Can you help me interpret this? Let me know if you need any further details from the analyzer

Gianni Ceresa

That call is part of the usual embedding process, it generally comes before to executing the query to get the data and return it to the embedding javascript.

Last time I had a look at embedding it was sending that POST query by sending the XML defining all the pieces of data present in the workbook and was returning a bunch of metadata about it, describing the source datasets and the various columns etc.

And then you should have another request getting the real data back based on the workbook definition.

It's the normal process when embedding a DV content (and also when doing the export to formatted Excel of a DV table or pivot, but that's a different topic).

A random path to explore: what is the source of your workbook? A subject area? A dataset based on something else? Does your workbook contains lot of different data objects or just a handful?

Dhaval Parikh Stantec

@Gianni Ceresa Thanks for your prompt response! No, the data source is not a subject area. The data source is a bunch of database tables (about 17 tables) joined together to create unified dataset. You think the data model here might be the issue?

Other thing I found is that when I look at the HAR file while loading the DV workbook directly in DV, it is only taking 0.42 seconds to execute this above metadata request. However, when the same DV workbook is embedded in the BI Answers dashboard, it is taking about 8.512 seconds (screenshot above) to execute the metadata request.

Gianni Ceresa

I still didn't start analyzing the details of the metadata request (it's something I'm working on in the coming days to analyze a different thing in DV).

At this point in time I don't know (still didn't look) if the metadata loaded is only the piece required by the visualization or is the full dataset metadata. In this case there could be a difference making a dataset as small as possible, with as little customizations and transformation inside as possible.

I also believe it's some of those internal things that aren't covered much in how things works and why (it's an aspect I find quite weak and not optimal in DV). Maybe some Oracle employees knows exactly what happen during an embedding and also if the difference in time is because of some extra authentication involved or maybe some "caching" effect because when you are in DV some of that has already been loaded or whatever (DV is a tons of calls to various APIs for everything, it's quite challenging to follow what does what without having a full documentation or the source code of the product).

[Deleted User]

@Dhaval Parikh Stantec Please open an SR for this, describe the issue and upload the HAR file. Your observations so far point to an issue that's best handled in an SR as product development can be informed and kept appraised.