Billions of handles - bottlenecks?

Hi folks,

Could someone please help me understand a little better for high throughput scenarios?

I was under the assumption that the root level infrastructure of dealt with the prefix after which it was the sole responsibility of the registered sites to deal with the suffix resolution. Under this scenario, if one cared to issue billions of suffixes against a prefix it would be their own concern.

However, someone hinted at an additional roundtrip to the root services incurred on each identifier creation which could be a bottleneck. Is this the case please?

Tim, the handle resolution (redirect handle to the registered URL) is done by the handle service provider. In the case of DOIs that is CNRI, which provides a scalable distributed infrastructure able to handle billions of resolutions. CNRI started the handle system, but there are now alternative providers, coordinating via the DONA Foundation.

The organization that registers handles only has to worry about registration of handles and updates of the URLs the handles point to. This is done in the central handle infrastructure. This can be a lot of effort, made a little bit easier when using template handles. Related discussion at Billions of records.

Thanks @mhfenner

I’m sorry to labour the question, but when an organization is registering a new handle does it do a call to the central infrastructure? The architecture section of the technical manual implies that the root services only deal with the prefix resolution and wouldn’t know anything about the suffix (which means the organization can mint handles at it’s own speed without a central bottleneck). On resolution I understood from that document that the root services would only look at the prefix to know where to redirect to.

The handle registration and updates are done via the primary handle server. You could run that yourself or have someone else run that for you, as DataCite is doing for its DOIs (where CNRI runs the primary handle server).