更好地查询数据库或将表数据加载到CLR程序集中的对象中？

Question

I'm importing a flat-file into a DB via CLR assembly.
For each row in the flat-file the assembly does a few quality checks. I've noticed that storing DB tables in DataTables and querying these DataTables is much slower than querying the DB directly. HashSet on the other hand seem just as fast as querying the DB.

At the moment, my code sometimes loads data into a HashSet and queries the HashSet for each row, at other times it checks the DB separately for each row. For example when I check if a key in the source-row exists in the database for ~10 000 source-rows and ~1000 possible correct keys.
HashSet :
+ I query the DB only once, and the assembly can perform its checks on the HashSet.
- Why replicate something that already exists in the DB?
Query the DB :
+ DB holds the structure of the table, and is optimized for these kinds of queries.
- I have to manage the DB-connection, which may include opening/closing the DB-connection multiple times.

I want to standardize my code, and need help deciding which option to use? I don't see a difference in performance for me. If opening a DB-Connection from a CLR-assembly is a non-issue, then I would prefer to query the DB, since I can then just write SQL code into my CLR-assembly and execute it, rather than having to code multiple objects.

Is there a technical reason to use one over the other?
A coding-style recommendation?

Note : I am working with static data, so I don't need to worry about data-changes while the import-assembly is running.

Perhaps a related question.

Answer 1

I will vote for HashSet . As there no significant difference in DB vs HashSet. Once you have everything with you in system you don't need to worry about DB calls connections and all.

- Why replicate something that already exists in the DB?

This replication will free you from hitting DB time and again, you can play around data with in the system.