Jul 24 2006
The reliability of web services
A few months back I posted a link to my ten favorite Web Mapping Services. The post included live links directly to the WMS servers. At first I questioned this move as locally hosted images would be far more reliable. But I thought it would be a neat experiment to see the downtime of each site. So I checked it daily just out of curiosity…
Well with today’s apparent disappearance of the NASA JPL site, all but one of my WMS layers mentioned have been down for at least a significant portion of a day. (The only one that’s been consitently up has been http://mesonet.agron.iastate.edu) .
This echos back to what I was complaining about with the whole USGS National Map debacle. The bottom line is that whenever we rely heavily on a web service to deliver essential data, we are risking the integrity of the end product. The chain is only as strong as it’s weakest link and, unfortunately as the USGS and NASA have shown, those links can and will fail completely from time to time.
Not to be an ESRI “fan boy” (as Sean Gilles would call me), but this is one reason that I really like the new geodatabase replication that’s coming with ArcGIS 9.2 - you can cascade data between organizations, bi-directionally, at the database level. Since you have all the data locally, and it’s kept in synch with the original source, your application is not dependant on the uptime of the remote service. Of course this functionality is not “free”, but it’s better than relying on inconsistent services for mission criticla datasets
Here’s a link to the web help about GDB Geodatabase Replication
Cheers,
Dave
Good post Matt. We have a project where we integrate around 30 map services in a targeted portal. We use everything from large National services (seamless, Geography Network, etc) to smaller scale services specific to the project (project partners). The major organizations are quite stable actually. The services that are run out of University’s with mostly graduate student help are the ones that go down the most. Check it out here if you are interested. The fun thing about the project is that we have been doing this for about two years and we have statistics/uptime for that entire time period.
Jeremy
Interesting experiment, Matt. I am working on a Colorado WMS/WFS project at the moment and
reliability is one of the issues that I am curious about. Do you think it has to do with the
particular server hardware or an issue specific to WMS services?
Very interesting.
Chris
Great discussion topic. I’ve been thinking about this a lot lately, and so have many others here: http://wiki.osgeo.org/index.php/WMS_Tile_Caching. The problem with WMS is that there’s a cost/benefit mismatch between what you want–a pretty map–and what it takes to deliver that to you–lots of computation to rasterize vector data. That’s why OGC and others are working on a tiling scheme for WMS so that you can pre-generate your maps and just serve out files, like all the big street mapping sites do now anyway.
In publishing a service that will be connected to, lights-out, by the public and stakeholders one might not even be aware of, the hope is that the owner of the service considers these aspects in advance. This goes toward the overall strategy and service level agreement (SLA). What is the backup and replication strategy? How reliable is the service expected to be in terms of throughput and uptime? How about data latency - is it going to be updated on a reliable basis? Data accuracy? Is the service something that has solid and ongoing funding as part of a bigger program, is it just a proof-of-concept or pilot, or are there other contingencies to consider in terms of longterm viability? These things may have some impact on whether others will even consider the published service worthwhile… Some plan, or better yet, a published metadata describing these aspects should be on the radar of any service provider.
Very good article. I don’t believe web services give right information on their frontpages especially about uptime percentages.