ArcGIS in the Cloud
Posted on November 21, 2008, 10:05 am, by James Fee, under
GIS.
A couple people have commented recently about this blog or via twitter about the possibility of ArcGIS and the Cloud. The current licensing of ArcGIS Server pretty much precludes the possibility of running it in the cloud so until ESRI changes the licensing or works with a company to provide ArcGIS Server in the cloud don’t get your hopes up.
I’m guessing it is virtually impossible for ESRI to license ArcGIS Server in the cloud as the product stands today.
Good and timely topic. This needs to happen, but it needs to be affordable.
On a side note, I have been able to configure Geoserver/PostGIS in the cloud (I used gogrid.com but I am also looking at Amazon EC2/S3) and it works extremely well. Even tested out ZigGIS to make direct edits to my PostGIS tables from a stand alone ArcView license. Very nice!
Can we throw ‘the cloud’ on the heap pile of technojargon please?
Maybe once ESRI gets their product working well on my desktop they can tackle this issue.
@Rain Maker: Desktop is irrelevant to server no matter what ESRI wants us to believe. This whole Server and Desktop relationships hurts everyone.
What about the ArcGIS server license does not allow this? How is running a windows server instance on EC2 and installing a license of ArcGIS Server on it any different from installing ArcGIS Server on a windows server in my own data center. They are both windows servers that I administer via remote desktop. It seems like I am bound by the same license no matter if the server I am using is in my office, your data center, or in EC2. Whats the diff?
How do you handle the scalability of EC2? I don’t want to license 16 cores if 99% of the time I only need 2 cores.
Running ArcGIS Server in a virtual environment is easy, but the scalability of EC2 throws a wrench into the model.
Thinking Out Loud – How will seat pricing work in the cloud?
Exactly the problem we are talking about with ESRI’s current license plans.
Sure I see what you mean, so you want an elastic license scheme.
Generally, you have exactly the same problems of ArcGIS Server licensing list prices scaling high even in your own data center on all of your own hardware, as soon as you start building out a multi-tier server platform (GIS servers on one tier, GIS web apps on another tier, geospatial data on another tier), and then you size the whole thing for say, 2000 concurrent users – you might need 20 ArcGIS Server licenses because you have 20 servers. Then again, if you really want to do that, call ESRI and see what they can do.
But even without an elastic license scheme, ArcGIS Server on EC2 is still really great. Now you have the ability to setup a multi-tier server environment w/o purchasing all the hardware, and the ability move capacity up and down while in product development to get the production system sized right, in a way that most people could not afford to do in the past is pretty cool.
But its nothing really to do with the “cloud’s elasticity”, as you have exactly the same problem in your own office as you fire up more servers or VMs. Maybe we just notice it more now that you can start up servers “in the cloud” quicker and cheaper than you can setup new servers in your own office.
Even with licensing as it is, I think now you will see people building all kinds of ArcGIS Server web apps “in the cloud” because they can make big multi-server implementations without having to build their own data center, colo, or lease, etc – all of which is expensive and takes tons of time and money to setup.
But yeah, it would be cool to pay by the instance hour for an ArcGIS Server license, like you do with SQL Server on an EC2 instance. I would pay an extra $.20\hour on EC2. It would also be cool if Amazon would let you suspend a server (and thus the meter), without killing the server forever – although I can see why they don’t – it would significantly reduce their cloud revenue because we would all turn our EC2 instances off at night.
Amazon is now hosting public data sets in the cloud as well. How long before we see geospatial data?
http://aws.amazon.com/publicdatasets/
Interesting this begins to put them into Google’s territory. Wonder if Google will do something similar with AppEngine?
James:
One of the memes popping up re the incoming Obama administration is that “a crisis should never be wasted” i.e. the ever-worsening crisis we’re going through opens up the opportunity for deep systemic change that wouldn’t be possible in “normal times”.
While the “Cloud” has attracted a fair bit of hype, I think it is quite relevant in the GIS sector because, frankly, there’s a great deal of unhappiness in too many shops with how the burdens of server and database administration leave way too few resources leftover for things like cartography and spatial analysis (the reasons we all got into this field in the first place).
While there may be few signs of “cloud” activity from Redlands, look at who is already working with the technology: WeoGeo/SAFE, Arc2Earth, and, oh, a little company from Redmond, WA. Not to mention that open source options such as PostGIS, MapServer, and GeoServer are by definition natural options (a worthy undertaking by the OSGeo would be to make it VERY simple to fire up a stack-on-demand).
Finally, NoLove’s comment about using zigGIS with PostGIS in the Cloud should give pause. What if Microsoft had a similar direct-connect extension for SQL Server 2008 that they gave away for ArcMap users?
In short, unhappiness + the value proposition = change coming sooner rather than later.
Brian
@James – How is Desktop irrelevant to server when most people are using mxds to create their services? Clearly w/ Google/VE APIs, SQL 2008, etc., you don’t need Desktop to create a Server application, but right now, most of us are still using mxds as our primary publishing base.
The idea of Server in the cloud is great, however, from the company that built their flagship product on COM when COM was nearly dead, I’m not expecting much in this arena very soon.
Desktop’s problems shouldn’t keep Server from reaching its potential and Server shouldn’t keep Desktop from being improved.
That is all I’m saying. Getting caught in the weeds with blaming server for desktops problems or blaming desktop for servers problems is how none of us will move forward. ESRI’s fault for designing their tools this way for sure, but I’m thinking bigger picture here.
What I’m interested in is whether there are licensing limitations in ArGIS Server that prevent you from creating a cache of tiles and then using VE or Google Maps to share them with 1,000’s of users. In a scenario like that I may only need a single CPU box running ArcGIS Server, yet I can serve geospatial data from it to 1,000’s of users concurrently.
(I’m working on the assumption here that you can convert the naming of the ArcGIS tiles in their R-Tree directory structure into something that VE or Google Maps can use directly without the ArcGIS Server JavaScript extensions.)
Here’s an interesting story about the use of Open Source in the Cloud:
Obama Campaign – Mapping voters with Mapserver, PostGIS and Openlayers
This has got to be on the short list of most important GIS stories of the year.
Forget about licensing issues for AGS in the cloud… as Rain Maker said, deal with the performance and architecture issues first; and make it work on the desktop and server before even thinking of creating vapourware in the cloud!
I have been waiting to see AGS application demos and case studies but no luck so far. When I scroll through the ArcNews newsletters, all articles mentioning “ArcGIS Server” in the title, are actually ArcIMS/ArcSDE implementations, but for some strange reason (!) still use the ArcGIS Server label.
The whole ArcObjects/COM thing has been so flaky for so long now that frankly speaking, I don’t give a damn… whether ESRI is working towards the cloud or not.
I know that the current licensing is an issue for running ArcGIS Server on the cloud, and I agree that that would be huge if ESRI could come up with an elastic license that could handle this that wasn’t prohibitiveley expensive. I am currently working on an ArcGIS server FLEX API application that uses Amazon s3 to store the tiles, app, and json service descriptions, and only uses ArcGIS Server to respond to queries (loading the map, turning layers on and off, and zooming and panning requires no server response). I will be posting a link to my blog about this soon. I am very interested in seeing how scalable the performance is with this setup (anyone interested in doing a large scale user test feel free to comment here about it).
On a side note, the negativity about server from this list of responses is disheartening. If everyone in the community is waithing for working examples before starting development, we will be in the same place 5 years from now.
Hi Adam did you ever get your Flex Ap up and riunning in teh cloud? wiould be interested in testing it if it is still going?