I help manage a cluster of sites related to course curriculum. Each site is a separate Drupal 7 install with its own theming, modules, user list, content types, entity structure, and codebase - all of them identical (not shared, but replicated across all the sites). They all run on the same server as different Drupal instances. I'm not a huge fan of managing each individual site and its modules, libraries and theming, but this is the setup as I inherited it.
Now we are looking at adding a few sites to the cluster. This is an opportunity to organize differently, and I haven't found any adequate data on which method is better. These are my three choices as I see them.
- Multiple Drupal instances (the way we have it now).
- If code for one instance gets broken, the others will continue to run.
- Opportunity for separate userlists, taxonomies, etc. (Right now we are not leveraging this, but I suppose we might in the future.)
- Every upgrade has to be tested and implemented on all sites individually.
- Performance hit from running all those instances (cron runs, session variables, caches, etc.)
- One Drupal instance running Multisite.
- One codebase means one set of modules to maintain.
- One Drupal instance would smooth out the performance issues (maybe?)
- Each site still has its own database, requiring separate userlists, taxonomies, etc.
- One Drupal instance with all the content in one place. (This is the way I'd really like to go.)
- Much, much easier to maintain.
- Performance savings? (This is what I really want to quantify somehow)
- Only one site to upgrade when we finally go to D8...
- Lose flexibility among sites' user lists and access levels. (I can get some of this back using taxonomy terms, but for now it's immaterial.)
- Lose flexibility with content types and structure. (The sites are very nearly identical as it is, and the differences can be smoothed out in development.)
- If one site gets broken, they're all broken. (Already true to an extent because all the sites are on one server.)
- Performance hit from all that content being in one database?
What I'm wondering is, what are the performance savings for each option over #1 (the status quo)? Quantitative data is great, but even an explanation of where the savings come from would go a long way.