ArcGIS Portal Deployment

[lenaburag]

I currently work for a small local government GIS that has been an ESRI shop (enterprise geodatabase, server) for 20+ years. We are currently upgrading to 10.8x, and thinking about including Portal for ArcGIS (Enterprise). We have 46 services (map, geoprocessing, feature, geocode) that feed both internal and external GIS applications.

We are looking at a multi-machine deployment, and would like hear from anyone that has deployed Portal for ArcGIS in their organization about what your current deployment scenario is and how many and what types of servers (Web, App, SQL) comprise your setup both on the outside (DMZ) and internal to your network.

We have spoken to ESRI, and were quoted $25,000 for professional services to help.

[berniejconnors]

Lena,

        We have used Esri professional services to give us some guidance on a past upgrade.  We should chat off-line and I can share our experience.  You can reach out to me on LinkedIn - https://www.linkedin.com/in/berniejconnors/

Bernie.

[jcarlson]

We have a multi-machine deployment, and it works wonderfully! We're very happy with it. We initially migrated to our current setup with the help of GIS Inc (now part of Axim, I think?), but in truth, it really isn't that difficult to do. We've moved to performing all Portal administrative tasks (upgrading, installing / removing things, etc) ourselves, and it has been just fine. Nice to have an IT person on hand, though, in case you run into some issue.

Specifically, we're using separate Windows machines for each of:

• ArcGIS Server
• ArcGIS Data
• ArcGIS Portal
• ArcGIS Web Adaptor
• An internal-only PostgreSQL server for our enterprise GeoDatabase

We're on AWS, so our instances can scale up or down (or even turn off) as load / scheduling dictates.

All our public stuff runs through Portal as Hosted Feature Layers, as those rarely need advanced capabilities from the EGDB, which significantly simplifies things for us.

I would take a close look at your services and think about how much each are used, and what your current server's workloads look like. At later versions, you can take full advantage of the shared instance pool, and it's possible to provide many services to your users without taxing your server or seeing serious impacts to performance. We have dozens of highly active services, apps, integrations with other tools, etc., and even on a 4-core machine, rarely see our server's CPU cross 20%, so we've got a lot of room to grow before additional Server licenses are even worth considering.

The documentation on installing the various pieces of a multi-machine portal is very good, and as long as your networking is set up correctly (easy to do on AWS), shouldn't take too long, either.

[BrianWilson]

We started with everything on one machine (virtual). About a year ago we migrated from that machine to separate virtual machines for the Server and Data Store, and we shut off the MS SQL server and moved to using the countywide SQL Server.

I had ESRI customer support help when there were glitches. I ended up with Data Store and Server on one machine and the standalone Data Store is in replicated backup mode. Customer Support was not able to help with flipping them around and I lost patience because we ran out of (weekend) down time. I will deal with that someday, like probably when I upgrade to 11. 

We went to a multimachine set up mostly to get SQL Server off the container with the Esri software. Everytime there was a problem, customer support would ask me to roll back the entire machine to its previous state. That's not an option when it means telling my users all their work from the previous day(s) is gone.

Having an enterprise-wide SQL Server means someone else now cares for it, so it's up to date and optimized and on a very fast server + network connection.

We have a license restriction for 4 cores only but now we have 4 cores for Server, 4 cores for Portal, and "many" for SQL Server. I also moved the ArcPro License Manager into a standalone machine. If I screw up any one component then restarting or rolling back only that one machine is much less stressful. It's also nice to be able tell our IT dept to reboot the entire server (for example the license manager server) instead of having to tell them how to manage each individual service on a single machine deployment.

We are a small rural county so the entire concept of a $25,000 consulting fee for what I do just fine on my own would never fly. I suppose they might go that route when I retire, they might have a hard time finding a GIS person with a deep dark IT past so maybe it's actually economical?  Or maybe I will be consulting again soon?

[BrianWilson]

I meant to say we left the original deployment on the first machine and then as I moved each of the other services off, after each was running I turned it off on the first machine. So there is a standalone machine running Portal and the Web Adaptors but at any moment I had the option to move things back there. I guess I still do technically.

I highly recommend using separate servers. I don't really see any advantage of a single machine when each "machine" these days is just a partition on a disk (on our server or in JCarlson's case in AWS)

Web Adaptors + IIS + Portal all run still on one machine but that's fine for us. It was easiest to leave it that way.

[Brian_Wilson]

When I hear some crazy number tossed out like that "were quoted $25000" usually I just go about my day. Then at lunch it struck me, is that one of those one line quotes or do they break it out? I mean, they spend so much energy at product launches and conferences saying how easy everything is but then to do a standard operation like this, they want $25,000? Is that 25 employees each working on it for a day? It should take one experienced knowledgeable person one day or maybe two?  Or a week? Let's say it's a week. so $25,000 / 40 = $625/hour. Maybe it's two people and I can't imagine what the second person would be doing. Getting everyone coffees. Maybe Jack Dangerman flies in himself on a jetpack to deliver a block of clear plastic with your name engraved on it. (Do they still do that award thing?)

I have to assume the quote includes something like "we will migrate every service for you and make sure every one of them works" though of course there's nothing to migrate since they are already running....

Please tell me what you got a quote for!

I'd say going for a bunch of instructor-led classes would be a far better investment of your time and money and you'd be more comfortable in your job in the future. Also you'd qualify to get a raise and can charge your small local government $625/hour!!! Well, maybe just $300. Make your mom and pop so proud. Unless of course they are taxpayers.

[TonyContreras_Frisco_TX]

Similar to what the previous posts mention, I would recommend splitting out the different components to different servers. Some components can have multiple machines (arcgis server - active/active, portal - active/passive, data store-active/passive) which helps for High Availability (HA). How or if you utilize this feature, what hardware specs you decide on, other choices you will have to make are things you will have to consider in relation to what your current needs are and what you expect those needs to be in the next few years.

That is where consulting services come in if that is beyond your experience. I'm not trying to justify the $25,000 price tag, and someone out there is probably willing to do it for a more reasonable fee, especially if there are less requirements to gather and consider. However, this planning portion is important and can save time and money in the future as well as result in a highly performant environment without over-building.

If you are part of a GIS User or other Professional GIS Networking group, see who has some experience in this type of planning and see if they will give you some time to talk it out and offer suggestions.

To answer your question-

We are a mid-size City in Texas, but the GIS is utilized in some way by almost every department in the city. We host data and apps that are used internally and by the public. We went through an Architecture review and have a new environment planned for later this year. Our current system looks like this:

• 4 GIS Servers (2 GIS sites of 2. The hosting servers are running relational Datastore in Primary/Secondary setup)
• 1 Portal Server 
• DMZ Server with Web adaptors for Portal and each GIS Site
• 1 Geoevent Server (not federated, just used for writing live data to Enterprise Geodatabase)
• SQL Server Always On Cluster hosting Enterprise Geodatabases

You should be able to do this type of move with some rigorous planning. You should expect some hiccups, but will learn the quirks and limitations of the software. Do your best to create helpful documentation so you don't have to re-learn those things the next time. Going through this type of process will help you better administer your environment and have a better understanding of how the components work so you can use it to its full potential. Good luck!