SE401:Group33:Research:Mark:TTL Load Balancing

From Marks Wiki
Jump to navigation Jump to search

Load balancing for us can be done at the DNS level. This should be fairly easy with a static site. Probably use a simple round robin (even better RR2) technique.

This paper

  • They explore using the TTL (Time To Live incase ya didn kno) values by "adaptively setting the TTL value based on the client request rate and/or the server capacity can dramatically reduce system imbalance even in a highly heterogeneous distributed Web system"
  • Round Robin isnt good enough so a two-tiered round robin (RR2) is "found to provide the most robust peformance". The busy local gateways speak up and are moved to a separate round robin with different request rates.
  • Their adaptive TTL; small servers get a few big requests and powerful servers get a lot of small requests. This is combined with RR2
  • RR & RR2 balance load among homogeneous distributed servers. probabilistic routing features may be added to extend this to heterogeteous distributions.
  • They modify the TTL values based on the additional component of server capacity.


Performance

  • They list the assumptions they make for a performance test
Retrieved from ‘https://wiki.kram.nz/index.php?title=SE401:Group33:Research:Mark:TTL_Load_Balancing&oldid=11121’