)]}'
{"ironic/common/hash_ring.py":[{"author":{"_account_id":6773,"name":"Lucas Alvares Gomes","email":"lucasagomes@gmail.com","username":"lucasagomes"},"change_message_id":"1dc716ca7ebf1ffdf5e76b1c2c8dffe9a320bedb","unresolved":false,"context_lines":[{"line_number":27,"context_line":"               help\u003d\u0027Exponent to determine number of hash partitions to use \u0027"},{"line_number":28,"context_line":"                    \u0027when distributing load across conductors.\u0027),"},{"line_number":29,"context_line":"    cfg.IntOpt(\u0027hash_distribution_replicas\u0027,"},{"line_number":30,"context_line":"               default\u003d2,"},{"line_number":31,"context_line":"               help\u003d\u0027Number of replicas in the hash ring.\u0027),"},{"line_number":32,"context_line":"]"},{"line_number":33,"context_line":""}],"source_content_type":"text/x-python","patch_set":1,"id":"AAAATn%2F9%2FUc%3D","line":30,"updated":"2013-12-03 18:18:33.000000000","message":"I\u0027m not expert in consistent hashing, but how did you reach this number? For what I\u0027ve been reading I think it would be better to have more replicas by default instead of only 2 [1][2][3], the more replicas you have the more balanced the distribution of the keys will be.\n\nMaybe lifeless would have a suggestion? \n\nBy looking at the comments of [2], I would say let\u0027s put at least 10~20 replicas by default.\n\nOr maybe even have the number of replicas to be dynamic? Based on the number of conductors?\n\nThoughts?\n\n[1] http://amix.dk/blog/post/19369#Consistent-hashing-Improving-distribution\n\n[2] http://michaelnielsen.org/blog/consistent-hashing/#comment-29489\n\n[3] http://www.tom-e-white.com/2007/11/consistent-hashing.html","commit_id":"d2aceb5b9172494f9d164ab9bc11738e69e01135"},{"author":{"_account_id":2889,"name":"Aeva Black","email":"aeva.online@gmail.com","username":"tenbrae"},"change_message_id":"20ce319592622fd141554c20df3c022e5043133d","unresolved":false,"context_lines":[{"line_number":27,"context_line":"               help\u003d\u0027Exponent to determine number of hash partitions to use \u0027"},{"line_number":28,"context_line":"                    \u0027when distributing load across conductors.\u0027),"},{"line_number":29,"context_line":"    cfg.IntOpt(\u0027hash_distribution_replicas\u0027,"},{"line_number":30,"context_line":"               default\u003d2,"},{"line_number":31,"context_line":"               help\u003d\u0027Number of replicas in the hash ring.\u0027),"},{"line_number":32,"context_line":"]"},{"line_number":33,"context_line":""}],"source_content_type":"text/x-python","patch_set":1,"id":"AAAATn%2F9%2FDI%3D","line":30,"in_reply_to":"AAAATn%2F9%2FUc%3D","updated":"2013-12-03 19:10:30.000000000","message":"18:44:37 \u003c devananda\u003e lucasagomes: i think i did not explain what replicas means in this context\n18:45:23 \u003c lucasagomes\u003e devananda, oh\n18:45:35 \u003c lucasagomes\u003e is not the same as the replicas on the other consistent hashing algorithms?\n18:47:51 \u003c devananda\u003e lucasagomes: you may be thinking of buckets\n18:48:14 \u003c devananda\u003e where # of buckets affects the smoothness of distribution of keys across the ring\n18:48:32 \u003c devananda\u003e here taht is the hash_partition_exponent\n18:48:42 \u003c devananda\u003e # of buckets \u003d\u003d 2 ** hash_part_exponent\n18:49:09 \u003c devananda\u003e replicas \u003d\u003d # of conductors which prepare  the deployment environment for each node\n18:49:23 \u003c devananda\u003e replicas\u003d1  --\u003e no backup.\n18:49:25 \u003c lucasagomes\u003e ahhh\n18:49:27 \u003c lucasagomes\u003e righttt\n18:49:49 \u003c devananda\u003e replicas\u003d2  --\u003e one backup. any conductor goes down, all the nodes which were mapped to it can be serviced \n                      by other conductors (without downloading new images from glance)","commit_id":"d2aceb5b9172494f9d164ab9bc11738e69e01135"}]}
