AMP Spatial Match Vs e1 Spatial Match

[rasutine]

Hi,

I have noticed I get different results every time I run a Spatial Match process using the AMP engine (Poly intersects a Poly).

It runs a lot quicker which is good however as none of the results match the e1 Spatial Match answer I cannot use it.

Has anyone else used this tool with the AMP engine? If so, any pointers?

I have tried the universe both within the tool and outside. I have tried AMP only and standard Alteryx database files.

Examples below of same process run on the same computer a few minutes apart.

Example1 - Run 1 using AMP Engine and AMP files

Example2 - Run 2 using AMP Engine and AMP files

I have also run the process using AMP engine with standard Alteryx database files as well as using e1.

The only consistent answer was the e1 run. Every time it was the same.

The correct result is only 1 unmatched polygon.

The only conclusion I can make is the AMP engine is not doing all to all when distributing the work.

Any help would be much appreciated

[TonyaS]

Hi Ross,

Thank you for the early feedback on the new AMP Engine. We are working hard to address issues that our customers are finding now that AMP is available. 

One thing of note with the Spatial Match and AMP (from our documentation): In mode "Use records from File or Database" only CSV, YXDB and SQLite formats have been converted. Other file formats will fall back to original Engine. I don't think that is what is causing this though.

We have started looking at this to see how we can help. I will reply again once we know more.

For now, here are some helpful links to available documentation about AMP Engine: 

https://help.alteryx.com/current/designer/alteryx-amp-engine

https://help.alteryx.com/current/designer/Alteryx-Engine-and-AMP-Main-Differences

https://help.alteryx.com/current/designer/AMP-Memory-Use

https://help.alteryx.com/current/designer/tool-use-amp

https://community.alteryx.com/t5/Analytics/Accelerate-Your-Analytic-Processes-with-the-New-AMP-Engin...

https://community.alteryx.com/t5/Engine-Works/AMPlify-your-Workflows/ba-p/617590

https://community.alteryx.com/t5/Alteryx-Designer-Discussions/Best-practice-with-the-new-Multi-threa...

AlterEverything Podcast: https://community.alteryx.com/t5/Alter-Everything-Podcast/66-The-Alteryx-AMP-Engine-Explained/ba-p/5...

https://community.alteryx.com/t5/Engine-Works/AMP-Engine-Technical-Deep-Dive-Part-1-Why-AMP/ba-p/570...  

https://community.alteryx.com/t5/Engine-Works/AMP-Engine-Technical-Deep-Dive-Part-2-Key-concepts-of-... 

[MarqueeCrew]

Please try this macro without AMP:

https://gallery.alteryx.com/#!app/CReW-Spatial-Match/5c335c5f0462d70ba8a5b2cb

cheers,

 mark

[rasutine]

Thank you. 

It would be good to find an answer as it is worrying that I get different answers every time I run the process using AMP.

[rasutine]

Thank you. I will experiment with this macro.

[TonyaS]

@rasutine 

We are not seeing this on our latest build versions. 

Can you please let me know what version of Alteryx you are on? 

We have actually fixed multiple issues with polygons in AMP in the last couple releases 2020.3 and 2020.4. 

[rasutine]

Hi Tonya,

Thank you for getting back to me.

I saw the issue in 2020.3 so updated to 2020.4. The issue was still there in 2020.4.

Any ideas?

Ross

[TonyaS]

Hi Ross, 

We have been looking into this and attempted to reproduce it in 20.4 and 20.3 in addition to our latest builds and we have not been able to reproduce the issue. Is it possible that it is specific to something with the particular data you are using? Would you be able to send me a workflow that demonstrates the issue for you with a small sub-set of data? 

[rasutine]

Hi Tonya,

How can I transfer the data to you securely (and privately)? The 2 datasets are 250MB and 56Mb in size.

(This are smaller versions of what we have been using but showcase the issue)

After some more analysis we have found the following on the above files:

• it is sensitive to file size, results for smaller files seem not vary (Test 2)
  • increasing the files sizes the test varies (Test 1)
• it is also sensitive to which way round the inputs go in - Test 3 at least has no unmatched, but the volumes fluctuate

You would expect Test 1 and Test 3 to get the same results (as you do in the non AMP version). However in the 6 runs using amp none of them are the same as each other or are the same as the non amp version.

We saw these results in both 2020.3 and 2020.4

Any help you can provide will be extremely useful.

Kind regards

Ross

[TonyaS]

Hi Ross,

I'm looking into how you can get me the data. 

I'll private message you once I have the details.