Migrate from ArcGIS Enterprise to ArcGIS Online

[Michael_Kraus]

I have a 4 year old ArcGIS Enterprise Portal - currently at 10.9.1 with about 25 creator and 250 active viewer users.  I am the only ArcGIS Pro user.  Portal is exclusively used internally within the organization.  I don't use any Server extensions, I use ArcGIS Pro for publishing, and I don't use an enterprise geodatabase.  I currently don't connect ArcGIS to our IT department's cloud data stores (Big Query) - I had planned to connect but don't actively use these yet.

With increasing hosting costs and an organizational IT policy of SAAS first, I am contemplating a switch from Enterprise to AGOL.  We do already have one AGOL organizational site which has been earmarked for external users / public sharing.  If we went down an AGOL path I would probably consider an "internal" and "external" version with collaboration used to sync selected datasets.

Our main benefits of Enterprise have been free viewer users (especially when launching and we didn't know how many there would be) and not having to pay for storage / analysis (other than the hosting costs).  Downside of Enterprise has been lagging behind AGOL updates, persistent bugs, hosting costs, and upgrade costs.

Obviously I will have the pay per user costs and storage and analysis credits, but based on our annual hosting costs, this will be a lot less than our current arrangement.

Has anyone made this switch and care to comment on pros and cons?  Functionality? Reliability? Performance?

I have struggled to find much documentation comparing ArcGIS Online with Enterprise which would make it clear what I will miss out on if I made this change - if anyone has found useful resources please post a link.  

Thanks

Michael.

[jcarlson]

It sounds like you already have a good plan forming for the migration. If you're not using an enterprise geodatabase or any server extensions, the list of benefits that Portal can offer gets quite a bit shorter. Our org needs Portal, thanks to a few advanced datasets, but we keep our hosting costs pretty small by using cloud infrastructure.

It's only happened one time that I can remember, but if AGOL has a technical outage, you just get to sit on your hands. If your Portal has an outage, you're able to get straight into the machines and potentially address the problem faster.

If you haven't already, I'd do a full audit of your org's content (or at least of the items you're planning to keep) and get a good estimate on the credit usage it would take to store it.

[berniejconnors]

With AGOL you can publish features services but you cannot publish map services.  Map services give you more control over the cartography which in my experience is most important if you need to have greater control over labeling of lines and polygons.

Bernie.

[RTPL_AU]

Map out requirements vs functionality. AGOL is limited in many ways and there may be one deal breaker hidden from all the marketing material.
Are there any 3rd party tools currently using map services? Authentication or identity may become an issue.

Other option is using a fully hosted 'real' environment "PAAS" such as offered by Esri Australia where the servers and desktops are all in AWS or Azure (S3, EC2, AppStream, Workspaces, etc). Moves all backup, architecture, and below data level responsibility over to them and you can focus on the data upward.  Cheap when you cost the full stack + IT people required to run local hardware stack, etc. You can also use your current investment in licenses as well as retain 'free' viewers. 
For your linked CADD/AEC users someone like ITOC can deploy a similar architecture but catered to a multi-app scenario rather than being Esri-centered. 
Takes a bit of getting used to from a desktop user's point of view but if you have a good & receptive team it will open  up opportunities for remote work, site connections, etc not previously possible without fat pipes going everywhere.
Do you use Enterprise/SQL servers as point of truth for all corporate data, raster/LiDAR, etc, even that which is not published to services? Where is that going?

Performance:
Using AGOL data is much slower than on-prem. Running a query against the updated Qld DCDB is a non-issue on local data but character building in Pro with AGOL. It works fine until it doesn't, which usually is when the solicitor needs a linear infra map series by 13:00.......

Security:
Does your SSO provider play nice with AGOL?

Pricing:
When last did you check AGOL pricing - there was a decent increase recently. 

Lastly - Backup. You will need somewhere to archive AGOL to; backups and having them accessible is the complicated bit. AGOL provide decent uptime but no protection against sneeze deletes. 

[Michael_Kraus]

Following up to my original post - we are currently testing functionality and migration to from Enterprise 10.9.1 to AGOL.  A couple of things we have found so far.

• Most things work fine!
• Publishing from Pro with the "support all clients" seems to be required to render services correctly in AGOL Map Viewer Classic vs not being required in 10.9.1 Map Viewer Classic
• Seems to be zero support for Annotation in AGOL.  We work with a lot of CAD and have been converting annotation to points before publishing to Portal anyway (it seems faster to display labels from points than anno), but Enterprise does have some support for annotation
• Labelling supported by Enterprise from a published web layer / map image layer is not supported by AGOL.  Particularly rotated labels which is quite annoying.  Several in the AGOL community have been asking this from ESRI for some years, but nothing on the horizon to support point labels rotated by an angle (attribute driven) in AGOL.  Workaround would be publishing tiled layers which do provide the label visual you see in Pro, but at the downside of losing dynamic labelling and not being able to move the labels to the top of the TOC.  Again another workaround is to add the labels to the basemap and then make the labels a reference layer in the basemap but these multiple steps are beyond the skill level and tolerance of my viewer users.
• No support for configured / customised map templates.  I was hoping to add a logo and standard disclaimer to a print template, but all of the AGOL documentation points to having an ArcGIS server configured to support this.  It seems standalone AGOL cannot do this?  We will have non-maintained perpetual Enterprise licences available but the biggest driver to migrate from Enterprise is the hosting cost, so don't want to rely on keeping an Enterprise licence alive somewhere to support this.
• Some bugs we have had to deal with from 10.8.1 to 10.9.1 are fixed in AGOL

Our migration process will include scripts / processes that will be reused for AGOL archives / backups, so seem to have a solution there.

Still looking for some definitive documentation that provides a comparison.

Also considering some 3rd party add-ons such as those from GeoJobe to support some administration things.

[Michael_Kraus]

Adding a few more observations and issues to the list.  We are now in the middle of migration.

Size of data - there is a bug somewhere.

Running an item report in ArcGIS Enterprise (10.9.1) we used the feature size to estimate how much space we would need in AGOL and how many credits we would use.  Unfortunately there is a bug somewhere that means this information is quite meaningless.

Enterprise "Item Report" was telling me that our feature storage was about 3Gb total.  For our 2500 or so items I thought this was a bit low.  During migration we quickly realised this was wrong.  A typical example - a polygon layer with 6000 suburb boundaries - ArcGIS Pro catalog view tells me this is 20Mb.  Published to Portal as a web feature layer, it is reported ad 4Mb, and doing the same to AGOL gives me 56Mb of feature storage.  So this error of 1400% makes our estimation of future AGOL credits fairly pointless (ESRI Case Number 03425687)

We thought with 25 creator users and the 12500 credits their user type includes would satisfy our annual storage, but with the current numbers it looks like we will run out in 3 months.

Map Services - features

This one fell through the cracks for us in our initial assessment.  Publishing a web layer from Pro as a web map with feature access was a quick way of getting multiple layers to our clients to consume in the map viewer.  They were presented a single layer with sub-layers they could toggle on and off in one motion, and this was grouped for logical reading in the map TOC.

I have now gone through all of these map services and made a choice to republish as feature services in most cases, but occasionally map tile services where there was annotation involved or the grouping / hierarching was significant.  The users won't like the flat structure / non-grouping when they add these to their maps.

As a workaround, I will need to do further work to understand the way these are used and it could end up that tile services are sufficient for the users, and give that simpler user experience and more user friendly handling / grouping of layers.  Alternatively we may accellerate our move to the "new map viewer" which does have layer grouping, but was hoping to delay the user training / documentation update that the change of map viewer will imply.

Backup solution

Moving from a hosted solution which included nightly backups to SAAS without any backups has us looking for a solution.  We are leaning towards GeoJobe Backup My Org.  No worries with the fairly low cost annual licence, but this needs a server for hosting the backup software and cloud storage for the backups.  So while we are trying to move away from reliance on IT department and the issues associated with needing to spin up a server and manage it, it looks like we will be needing a server for this.  

[RTPL_AU]

Hi @Michael_Kraus Interesting journey (for spectators, not you!) Having a web server on hand is not a bad thing - especially when looking at serving tile packages for Field Maps, etc. when the credit consumption process is a bit of a black box. You may also need it for widgets and other things if you stray outside the walled garden with web apps.

Additional credits are not too costly but last I requested it there was no detailed audit/reporting functionality for credit balance & usage. Also keep in mind that purchased credits have a shelf life. If you are lucky to not use them all they will disappear on you; which can be inconvenient unless you have a process in place to keep track and top up before they do.

Please post follow ups on your backup journey - it is a complicated topic often glossed over in the marketing materials. I have yet to find an easy to use 'anti-sneeze-delete' process to quickly restore something from an arbitrary point in time.
My go-to is to keep things simple and make sure data is backed up. A map/app can always be recreated quickly if it is not complex.
With Experience Builder replacing Web AppBuilder the process of managing App processes are a bit more complex too - it is more powerful, but a painful & slow from idea to app for very basic yet functional apps where you need just a bit more than what the templates give.

Good luck.