azure, arcgis enterprise 11.5 and poor browser performance

LachlanWainwright

Hi Everyone

We are experiencing poor performance with display data in the browser since we moved from On-Prem to AZURE with our ArcGIS Enterprise Environment.

The Problem:

The day we migrated everything we received incidents from users where the web apps they were using for years and years are now slow in the browser. Pro and Field Maps are rolling without incident/issue.

Setup:

Up until last week we were on 10.9.1, we just upgraded using DSC, like the 3 Jims, such beauties..apart from their tattling.

AE11.5, fully patched.  Separate machines for portal, ags, ds and all well spec'ed, portal has all the WA's. Shared config on Azure SMB share, EGDB SQL Server.  Pipe to our office is Express route, AZURE data centre is 800km away/500mile.  None of the machines are red lining/thrashing or like.  Nothing in the logs saying obtuse things.

The AZURE team have looked a server performance, iOPS, logs and have now raised a call with Microsoft.  A few years ago we have our ESRI reseller take a look and nothing could be found....

A friend of mine is experiencing a similar issue with just AGS on AZURE and they have not found a solution either. Good old AI is agreeing with with me....distance.....but who know if that a hallucination due to my shaping of the questions 😀

The Question:

Has anyone else experienced this? Does anyone have a silver bullet for me?

My thoughts:

I believe it's the tyranny of distance pulling uphill from Sydney to Brisbane, the AZURE team make comment on map/app optimisation....but it was speedie as on-prem!

The hope:

Someone says, press this big red button that says "speedie in browser again"

Late news:
I have found that we should be not be using SMB for our directories etc, recommended to use Azure Blob. Could this be the handbrake we are experiencing?

Enjoy our weekend and thanks for reading this and hopefully helping

Lachlan

JoshuaBixby

Troubleshooting ArcGIS Enterprise performance problems is challenging enough when you have access to the systems and lots of detailed information, doing it with limited information via a forum is nearly impossible.  There are so many factors that can impact ArcGIS Enterprise site and service performance that one has to be very detailed and specific when sharing information.

Saying the machines are " all well spec'ed" doesn't really tell anyone anything.  There are literally hundreds of Azure VM SKUs, and what one person might consider "well spec'd" another might few as barely up for the job.  The comment "shared config on Azure SMB share" is also ambiguous since there are many ways to consume files over SMB in Azure ranging from Azure fully-managed cloud offerings to self-managed infrastructure, and a similar comment goes for saying "EGDB SQL Server."

I am not saying the above to nitpick or detract from the situation and question, but to offer context for why it will be difficult to get meaningful feedback from community members.  Even Esri Support has its limits in situations like this because it is focused more around addressing breakage rather than performance matters, and you stated nothing looks overtly broken from the logs.  Maybe Esri Professional Services or some other consulting firm could help if the issue is impactful and important enough the business is willing to spend money on resolving it.

I will say that I doubt the 500 mile WAN distance is the primary cause.  That distance add 15 ms give or take to a network round trip, and modern web-based apps running in browsers are meant to tolerate 100s of ms of latency.  

I would start by taking the apps out of the equation and performance benchmarking the services they consume outside of the apps.  The more layers you can peel away and isolate, the easier troubleshooting becomes.