administering many DSpace/manakin instances

Does anyone have any experience and/or advice on administering multiple installations of DSpace/Manakin applications?  Our plan is to have about a dozen installations of the DSpace and Manakin web applications by early Spring, and then build up to approximately 100 installations as they are needed.  Each installation will belong to a separate institution or organization, typically an academic library.

We will be performing customizations to the software.  Some customizations will apply only to an individual instance; others will be global.  We will also need a way to search/browse for items globally across all instances.  It’s likely that we’ll collect OAI/PMH records into an external tool to implement the global search.

I’m interested in collecting information about memory/disk cache requirements, any tools that have been developed for applying updates and customizations to multiple instances of the web applications, and any changes to tools, handle servers, etc. that are necessary/helpful for this type of setup.

I recently saw an example of a multiple institution repository that is running under a single instance of the application here:

http://dspace.nitle.org/ (NITLE DSpace Service ).

Managing multiple institutions under a single instance of the DSpace application is an interesting approach.  It looks like it would simplify administration for the developers and engineers, make global search work automatically, and possibly reduce system requirements.  I’m not sure that it would work in our situation here, though, for several reasons:

1. Customizations.  We’re unsure specifically which customizations each institution will need, and we may not be able to implement a customization for one institution without affecting another if they’re running in the same application space.

2. Control.  Our goal is to give each institution as much control over their space as possible, and we’re not sure that we’d be able to do this if they all shared the same application space.  One example is that we’ve had problems giving institution heads control to add subcommunities/collections to a community in Manakin, without giving them full administrative access to the system.  If each institution “owned” their own instance, we’d avoid this problem, and possibly other unknown issues.

3. Scalability.  How much load can a single set of DSpace/Manakin web applications handle?  If necessary, does the DSpace application cluster well?  If clustered, would the clustered applications still talk to a single database, or should the database also be replicated?

Any advice, statistics, anecdotes from anyone about experiences managing DSpace for multiple parties with differing needs?

Thanks,
Keith Gilbertson

Systems Developer

OhioLINK