Gungho::Component::Throttle - Base Class To Throttle Requests
package Gungho::Component::Throttle::Domain; use base qw(Gungho::Component::Throttle);
If you create a serious enough crawler, throttling will become a major issue. After all, you want to *crawl* the sites, not overwhelm them with requests.
While the concept is simple, implementing this on your own is relatively costly, so Gungho provides a few simple ways to work with this problem.
Gungho::Component::Throttle::Simple will throttle simply by the number of requests being sent at a time, regardless of what they are. This simple approach will work well if your client-side resources are limited -- for example, you don't want your requests to hog up too much bandwidth, so you limit the actual number of requests being sent.
# throttle down to 100 requests / hour components: - Throttle::Simple throttle: simple: max_iterms: 100 interval: 3600
In most cases, however, you will probably want Gungho::Component::Throttle::Domain, which throttles requests on a per-domain basis. This way you can, for example, limit the number of requests being sent to one host, while letting the remaining time slices to be used against some other host.
# throttle down to 100 requests / host / hour components: - Throttle::Domain throttle: domain: max_iterms: 100 interval: 3600
This component utilises Data::Throttler or Data::Throttler::Memcached for the main engine to keep track of the throttling. Data::Throttler will suffice if you are working from a single host. You will need Data::Throttler::Memcached if you have a farm of crawlers that may potentially be residing on different hosts.
By default Data::Throttler will be used. If you want to override this, specify the throttler argument in the configuration:
components: - Throttle::Domain throttle: domain: throttler: Data::Throttler::Memcached cache: data: 127.0.0.1:11211 max_items: 100 interval: 3600
Starting from 0.09003, you can stack throttlers. For example, you can throttle by Throttle::Simple first, and if Throttle::Simple allowed the request to go, then you can throttle with Throttle::Domain as well to make sure that the same host doesn't get beaten up.