Background:
I have a 2 class A & B. Class A contains 1 int64, 1 int, 1 datetime, 6 strings, 1 byte[] and a list of Class B objects.
Class B contains 2 Int64, 1 enum (4values), 5 strings.
Class A and B have a master detail relationship between them. Class A can have 0 or more objects of Class B in its list.
These objects are used to store search results from our db.
We have a SOA architecture. The client can search withing a particular date range and with a fixed set of criteria. We use these on the service to gather the search results and push them to the client in batches. After sending each batch, we delete those objects from the service.
Problem:
If a user does a search that returns 200k results (Class A + Class B counts combined), the service memory consumption jumps 200mb. But if subsequent searches return only few results, the memory doesn’t come back to its previous state. When I look at the performance counters for the service, I see that lot of memory is taken up by gen2 objects.
Without trying to do a GC.Collect(), how can I make sure that that memory is returned more quickly? I though of using the IDisposable interface on Class A and Class B, but since most of the fields in both of them are strings, I don’t know how I can forcefully dispose them.
You might want to consider attacking the problem from a different angle. Rather than trying to optimize low-level things like Garbage Collection details, why not take a step back and ask do you really need to return 200k results at once?
If this is being displayed in a UI then showing 200k results is completely useless to a human being as that’s just way too much data. I’d suggest doing some kind of paging and only showing 100 at a time (and only retrieving 100 at a time from the service). Or maybe showing the user the first 100, then have them click a button if they really want to retrieve all 200k.
Improvements like that are going to have a much bigger impact than micro-optimizing memory management.