Instagram: http://instagram-engineering.tumblr.com/post/10853187575/sharding-ids-at-instagram
Flickr: http://code.flickr.net/2010/02/08/ticket-servers-distributed-unique-primary-keys-on-the-cheap/
Ticket Server based on database (auto-increment in DB, ACID guarantee)
Instagram:
id should be chronologically sorted;
Sharding: several thousands "logical" shard, and hosted in far fewer physical shards => makes it much easier to move logical shards out to new physical shards.
Id genration : PL/PGSQL stored procedure to generate next_id() in database (transactionally), which contains : 41bits for current time in mililliseconds, 13 bits for logical shards, 10 bits for auto-incrementing sequence % 1024 => at most 1024 IDs per shard per millisecond;
Given the ID also encodes the shard, it‘s easy to find out shard for a primary key id
CREATE OR REPLACE FUNCTION insta5.next_id(OUT result bigint) AS $$ DECLARE our_epoch bigint := 1314220021721; seq_id bigint; now_millis bigint; shard_id int := 5; BEGIN SELECT nextval(‘insta5.table_id_seq‘) %% 1024 INTO seq_id; SELECT FLOOR(EXTRACT(EPOCH FROM clock_timestamp()) * 1000) INTO now_millis; result := (now_millis - our_epoch) << 23; result := result | (shard_id << 10); result := result | (seq_id); END; $$ LANGUAGE PLPGSQL; CREATE TABLE insta5.our_table ( "id" bigint NOT NULL DEFAULT insta5.next_id(), ...rest of table schema... )
Flickr : http://code.flickr.net/2010/02/08/ticket-servers-distributed-unique-primary-keys-on-the-cheap/
/* The Tickets64 schema looks like: */ CREATE TABLE `Tickets64` ( `id` bigint(20) unsigned NOT NULL auto_increment, `stub` char(1) NOT NULL default ‘‘, PRIMARY KEY (`id`), UNIQUE KEY `stub` (`stub`) ) ENGINE=MyISAM SELECT * from Tickets64 returns a single row that looks something like: +-------------------+------+ | id | stub | +-------------------+------+ | 72157623227190423 | a | +-------------------+------+ /* When I need a new globally unique 64-bit ID I issue the following SQL: */ REPLACE INTO Tickets64 (stub) VALUES (‘a‘); SELECT LAST_INSERT_ID();
In order not to make it a single-point-of-failure:
TicketServer1:
auto-increment-increment = 2
auto-increment-offset = 1
TicketServer2:
auto-increment-increment = 2
auto-increment-offset = 2
presentation on this topic:
http://media.postgresql.org/sfpug/instagram_sfpug.pdf