I need to build a social network that is expecting very high traffic, 10k - 50k + registrations in a day, not even including visitors, pageviews, image uploads, etc..., a lot of database activity. We have a solid quad core dedicated server right now with 8GB ddr3 and 2tb hdd sata. Will doctrine be to slow as far as queries and load time to handle such a high traffic website?
Short Answer: You will certainly worsen performance, and hide database specific features that could provide more performance.
Suggestion: Use MongoDB - and a Database only where highly relational - transactional ACID compliant data is required. No ORM needed. Orm's hide the implementations of the underlying data-store. This is a great thing and an awful thing at the same time. We use MongoDB and Postgres at my corporation; This will not change for a long time. Our code is easy to read, to the point, and doesn't rely on "middleware". When new Postgres 9 features became available we could take advantage immediately. When the Python MongoDB driver was updated we could refactor very easily - no need to hack the ORM layer to accomplish this.
If you are building something like this you will need a combination of databases. A text search database, a relational database, and a fast schema-less database perhaps? How does an ORM help solve this problem? - it doesn't.
Answer is always Use normal queries instead of ORM to get better performance. If your are planning to change the database after some longer period ORM helps to done easily.You need to sacrifice performance for convenience.
I suggest you to create the separate class where you can perform DML operations by getting the inputs as parameter and create the query and use it to execute it. Instead of writing insert statement each time or for many occurrences . but Make sure too create the query which is more optimized .