Web Service Drawing Response: Classed vs. Unclassed Colors

[ZachBodenner]

Hello,

Is there a best practice for symbolizing large datasets using Classed or Unclassed colors? I have searched and low for some guidance, but I've come up short. I have a dataset that contains ~20k polygons. I've published them as a hosted layer but they still draw rather slow. Currently, they are symbolized using an unclassed color ramp but that doesn't seem to be a requirement for my end user. If I were to resymbolize and bin the values of the polygons, would that improve the response time of my server?

[MarkChilcott]

Hi Zac,

Do you have access to the UC2024 presentations?  There were some good sessions on Enterprise trouble shooting and performance tuning.  

You need to break down the problem and test in each part of the stack.

ie If you had the layer in an Enterprise Geodatabase and connect directly through ArcGIS Pro, what is the drawing speed like?  If it is slow there, you need to work out why.  Is is slow in a file Geodatabase?  Check you have spatial indexes and rebuild if required, check the indexes on fields used for display.  See if it draws OK with simple symbol, check for godzilla polygons etc.

Work out where in the stack the performance issue is occuring.

Download and use the ArcGIS Pro PerfTools to measure performance.   You could also try the ArcGIS Pro Performance Assessment Tool.

If the issue is in the server stack - look over the UC2024 Enterprise Geodatabase Performance Tuning for ares to look into.

Cheers,

Mark

[ZachBodenner]

I was actually in that session. There's a lot to pull apart with this particular layer in terms of why it could be slow, some things not actually within my control. It's a SQL view that joins together five tables, so that right there I know can be a heavy lift. The polygons are what they are; they're parcels that we source and then store locally from the county so we can't really run too much for simplification. On top of that, there's some math and some casing in the statement so it isn't even simply pulling data columns. Next, most of the tables are stored in an off-prem server that we have linked to our EGDB, so more lag time there, plus I then don't really have control over attribute indexing. 

So that's kind of why I'm asking about symbolizing. I know I have a lot going on that will drag the draw speed down so I'm very aware that I'm looking mostly for marginal improvement.