Cloud-based Trellis

[Aaron Coburn]

Hi All,

I wanted to let you know that I've started work on a "cloud-based" Trellis here: https://github.com/trellis-ldp/trellis-ext-aws

The code is designed to run in the AWS Cloud. It is still in a very early state (there are no tests and no documentation), but all of the basic structure is in place. The key thing is that this will allow one to run Trellis in such a way that it is entirely stateless with respect to its immediate runtime environment. That is, nothing is stored on disk, so one will be able to run this entirely in Lambda, or as part of an elastic fleet of EC2 instances.

The project is actually rather simple -- the EventService makes use of AWS' simple notification service, and both the BinaryService and MementoServices use S3 for storage. Otherwise, it uses AWS' Neptune as a triplestore backend. I would like it to also support the RDBMS-based ResourceService implementation, too.

Let me know if there are questions or ideas for this.

Regards,
Aaron

[Diego Pino]

Wow. Now we have 2 services that do 95% the same (esmero is ours), have not checked your code but ours is node.js based, linda reminds me when php disk based thing happened. We were waiting a few more months before releasing because of other ongoing but thighly related dev and also because of the moving specs, but you are a coding machine! congrats and thanks for sharing, really interesting . Will compare implementations further down the road :)

--
You received this message because you are subscribed to the Google Groups "Trellis LDP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to trellis-ldp...@googlegroups.com.
To post to this group, send email to trell...@googlegroups.com.
Visit this group at https://groups.google.com/group/trellis-ldp.
To view this discussion on the web visit https://groups.google.com/d/msgid/trellis-ldp/42B06673-130B-4DAF-9391-84E19EE220A1%40apache.org.
For more options, visit https://groups.google.com/d/optout.

--

Diego Pino Navarro

Digital Repositories Developer

Metropolitan New York Library Council (METRO)

[Soroka, Adam]

It seems to me that the biggest attraction of this kind of setup is elasticity. (Keep resource allocation just high enough to serve normal traffic, then when a big ingest comes along or your site is mentioned in the local paper, spin up a fleet just for the duration, then spin back down-- much cheaper than sizing to the maximal load, much less complicated to build or run than ingest queuing systems or the like.)

With that in mind, it would be great to have a demo setup or some other tooling to deploy that elastic fleet of EC2 instances you mention, even if only a very simple example.

---
A. Soroka : Senior Solutions Architect
Research Computing : Office of the CIO : the Smithsonian Institution