Statewide Advokit?
Submitted by springmp on Sat, 07/14/2007 - 01:06.
AdvoKit Development
I've been reading many of the posts and the geographical areas seem quite small. Can we run one huge statewide Advokit site, or should we split our state up and run smaller sites?
California version
Submitted by brucekdan on Mon, 09/03/2007 - 10:37.
Yes, we now have about 95% of California voters or about 15 million on-line at all times to support our permanent precinct captains program at http://PrecinctCaptains.org.
We are using sort of Advokit version 0.9.9 for this. That is the version that we started with. However, some of the later code changes and bug fixes have been incorporated. We have also made a number of enhancements. For example, we were able to get our hands on all the precinct maps for the state. So there is a UI for user to get their map. We have also made some schema enhancements to allow us to utilize voting history for each voter.
Another Post Mentioned Scale, Please Enlighten
Submitted by springmp on Sat, 07/14/2007 - 01:43.
Pat,
Does the following post made by you mean that we should split the state (Colorado)into instances? If so, how do we determine how much each instance can handle?
****************************
Scale may be an issue, but otherwise sounds doable...
Submitted by patdunlavey on Wed, 07/04/2007 - 12:59.
Advokit's already pretty good about allowing you to manage access to your voting population and access to data that you collect about voters in a fairly flexible and fine-grained way. And of course it's open source, so if you need to change the way Advokit works, the sky is the limit. The real question in my mind is whether the number of voters and users can practically be handled under a single Advokit instance.
I'll contact you offline and hopefully post summaries of what we figure out together here so everyone else can follow along.
-Pat
I'd keep instances at a
Submitted by patdunlavey on Wed, 07/18/2007 - 16:41.
I'd keep instances at a million voter records or less. Advokit performance is primarily a function of two things: voterfile size (number of voter records); and the number of users on the system.
If you are doing a phonebank style operation, there is absolutely no need to have more voters in your database than you will actually try to call. I have seen a number of campaigns load up Advokit with a huge number of voters, and only end up actually contacting something like 1% of them or less. It is much smarter to perform whatever targeting analysis you can on the voter data prior to moving it into Advokit, and then import only those voter records that are your highest-value. You will get your best bang for the buck that way.
On the other hand, if you are styling your operation as a friend-to-friend campaign, then you will want to have as close to complete voterfile coverage in the areas that you will be working. While Advokit was designed with large scale friend-to-friend campaigns in mind, I've never actually seen a f2f campaign tried at a large scale (hundreds of thousands or millions of voters and hundreds or thousands of activists). Unfortunately, I do not think Advokit would perform suitably in that kind of scenario. Splitting into multiple instances, divided by geography, would be the best solution for that.
In summary, if you're anticipating using voterfiles of a million voters or more, you definitely should be thinking about dividing into multiple instances; but only do so after carefully examining your assumptions about the need to have that many voters in your system.
Yeah.
Submitted by fredsambo on Fri, 08/03/2007 - 17:13.
I agree with Pat. You can do all of your list work out-side of Advokit (like in *shudder* MS Access) and only use records that are *complete* for the current campaign, as opposed to dumping all records into Advokit and then querying through all of the BS.
California example
Our California system (PrecinctCaptains.org) might be a relevant example. We have the state broken up into counties. One reason for doing it this way was because we receive our data separately from each County Registrar of Elections. So this is the natural unit of data load and reload. Each county is a single separate Advokit database.
We have indeed noticed some slowdown with our largest county Los Angeles at 3.5M voters whereas Orange at 1.5M and San Diego at 1.4M work OK. So as suggested, it would seem that 1M is a good limit. This organization also allows us to do work on a county without bringing down the whole system.