How to modify System Index?

Question

DSEE7 on Redhat 5 64bit.
I'm trying to modify the objectclass index to index substring searches. However, In DSCC it shows System Indexes as Read-Only and everything is greyed out.
How do I modify the objectclass index?

Answer 1

I don't think that doc explicitly goes into the difference in how the entry cache itself is managed when an unindexed search happens as opposed to when an indexed search happens - i.e., when are entries removed from the cache. That doc is more about the details of satisfying the search itself. Your inference that the entry would always be retrieved from the entry cache is incomplete, since there is always the case when the entry is not already in the entry cache.
I am going off of a previous escalation I worked on when ENG told us the entry cache gets blown away as part of processing an unindexed search. If we really want to prove it, we could look at the entry cache monitor and do some unindexed searches. But for the purpose of the current discussion, all we really need to know about unindexed searches and the Directory is that they are very resource-intensive and can interfere with the timely processing of your other workload. Presumably, that's why you are working on "making" this query be indexed, but the problem with the approach you are taking is that the nature of the query itself makes it unsuitable for indexing.
I disagree that this leads you back to your original question. Technically, the way to force such a query to be indexed is to raise the allidsthreshold so high that it doesn't do its job. I do not recommend that, as it will likely have negative performance implications for your entire Directory.
My opinion about where this should be leading you is either:
1) Towards an offline parsing job (if this query is part of a periodic report), or
2) To a dedicated reporting server (if this query and others like it need to be done in real time).
Of course, right now, you may not need to take either of these steps, if your current Directory infrastructure has enough capacity to handle the unindexed queries as well as its other workload.