Jedna myśl nt. „Dużo RAMu na Ubuntu i sysctl.conf

  1. I decided to use nodes bescuae each connection needs to efficiently find out who it’s neighbors are. If each node knew it’s neighbors only by ID, as opposed to holding a direct reference, each traversal step would require a look-up into an associative array. For finding the neighbor 2 positions to your left and 2 positions up, for example, a graph structure is just a little easier to work with IMO (e.g. pNode->pLeft->pLeft->pTop->pTop). Also, although chromium’s associative arrays may be O(1) complexity, they involve the overhead of a hash function and potential for collisions, which when dealing with hundreds of thousands of connections, add up enough to be a prohibiting performance bottleneck.

Dodaj komentarz

Twój adres e-mail nie zostanie opublikowany. Wymagane pola są oznaczone *