You need a very systematic approach to identify and resolve OOMs.
- Do not assume that if you have lot of RAM free it is all for your application.
- OOMs are very notorious to disguise themselves, you might see the Stack Trace pointing somewhere else and the problem actually would be elsewhere, in the sense, some other code is leaking/consuming large memory and has put the memory usage on the verge and a legit code somewhere else trying to secure some limited memory would push it to OOM.
- Use some profiler and identify the area where the memory is increasing/leaking. (Looks like you have identified this as the code to serialize your large data.)
- Then work on those areas and try to minimize the memory usage.