I was interested to read this article as it confirmed something I have felt for some time. That is Opensim Asset growth is a real problem long term.

I set up a small private Opensim grid with a friend sometime ago. We are not developers in the Opensim sense but we do have extensive IT experiance in database systems. So we thought we would investigate the Database behind Opensim and see if we could do anything to help in that way. So far we have built a php based adminstarion tool that allows drill down of the data. We are now working on a utility to identify and archive redundant assets. Its nearly finished. When it is we will be looking for a way to give this to the Opensim community so grid operators can keep their disk storgae requirements down.

We realised that assets get created by Opensim but never destroyed even when those assets have no pointers towards them in any of the tables. Such assets could be removed from the Asset file with no effect on Opensim other than disk space savings and retrieval effeciency improvements. This is especially noticabale in Script assets. A script is written and creates an asset. It is updated as it is developed and a new asset is created each time leaving the older one redundant and unused. The same sort of problems exist on other types of assets too.

In our small and not very active grid we found that 75% of all assets are not pointed to and so can be removed. I would expect on very active grids like OSgrid this percentage would be much higher.

So apply that sort of maths to your current problem and maybe you dont need to upgrade disk so often. Of course it will still grow and your plan is still a good one, but with the event of Hypergrid I think smaller grids with well managed asset archivial would also be a good scalable way forward.

Would love to have your views on what we are doing and will explain our approach in more detail if you are interested.

Microsoft Azure and Google App Engine can be an alternate solution to Amazon hosting.