Investigation of Transactions in Cassandra

Investigation of Transactions in Cassandra

1. 1. Investigation of Transactions in Cassandra Cassandra Meetup in Tokyo, Fall 2015 December 2, 2015 Ian Chen Rakuten Inc.
2. 2. Cassandra usage at Rakuten •  Cassandra  nodes:  400+  (many  clusters)   •  Data  size:  5B+  records   •  Rows  accessed/sec:    450k+  (busiest  cluster)   •  Largest  cluster:  90+  nodes   •  Versions:  0.8,  1.2+,  2.0+   •  Community  feedback  since  v0.8   2
3. 3. New Uses for Cassandra •  Not  just  for  “big  data”   •  Global  scale  applicaQon  data   •  TransacQons  and  more   3
4. 4. 4 Agenda    Consistency  (CAP)  1    Atomicity  (ACID)  2    Isola8on  (ACID)  3    Example  4  
5. 5. 5 The  “C”  is  for  Consistency   “..  any  transacQon  will  bring  the  database  from   one  valid  state  to  another.”   “…all  clients  have  the  same  view  of  the  data.”   CAP  Theorem  “C”   ACID  “C”   Eventually..  
6. 6. 6 Tunable  Consistency   •  Manage  consistency  in  your  applicaQon   •  Decide  consistency  per  transacQon   CONSISTENCY ONE; UPDATE users SET name='IAN' WHERE name='ian'; CONSISTENCY ALL; SELECT * FROM users WHERE name='ian';
7. 7. 7 Strong  Consistency   *  QUORUM  =  (replicaQon  factor  /  2)  +  1     nodes     wri_en   nodes     read   replicaQon   factor   R  +  W  >  N   Write  ONE   Read  ONE   =   WEAK   Write  ONE   Read  ALL   =   STRONG   Write  ALL   Read  ONE   =   STRONG   Write  QUORUM   Read  QUORUM   =   STRONG  
8. 8. 8 “A”  is  for  Atomicity   START TRANSACTION; INSERT INTO users (name, country) VALUES(‘ian’, ‘Australia’); INSERT INTO users (name, country) VALUES(‘meg’, ‘Japan’); COMMIT; // or ROLLBACK aka  transacQon  consistency,  linearizable  consistency,  serial  isolaQon  level   All  succeed  or  none  succeed   SQL:  
9. 9. 9 Cassandra  BATCH   BEGIN BATCH INSERT INTO users (name, country) VALUES(‘ian’, ‘Australia’); INSERT INTO users (name, country) VALUES(‘meg’, ‘Japan’); APPLY BATCH; Guarantees  ALL  will  succeed  or  ALL  will  fail   CQL:  
10. 10. 10 “I”  Is  For  Isola8on   START TRANSACTION; SELECT current_caps FROM bottlecaps WHERE name='ian' FOR UPDATE; UPDATE bottlecaps SET current_caps = current_caps + 1; COMMIT; Is  anyone  else  changing  this  data?   SQL:  
11. 11. 11 Cassandra  Lightweight  Transac8ons   INSERT INTO users (name, country) VALUES(‘ian’, ‘Australia’) IF NOT EXISTS; UPDATE users SET country=‘Japan’ WHERE name=‘ian’ IF country=‘Australia’; Allows  you  to  set  condiQons  of  an  update   CQL:  
12. 12. 12 Op8mis8c  vs  Pessimis8c  Locking   PessimisQc  (SQL)   OpQmisQc  (CQL)   SELECT     FOR  UPDATE   SELECT     FOR  UPDATE   COMMIT   COMMIT   lock   SELECT   SELECT   UPDATE  ..  IF  …   UPDATE  ..  IF  …   Thread  1   Thread  2   lock   Thread  1   Thread  2  
13. 13. 13 PuTng  It  Together   1.  Record a transaction 2.  Keep track of the total 3.  Ensure they are updated atomically Example  
14. 14. 14 name current_caps tx_id amount balance ian 0 tx001 null null Example: Bottlecaps CREATE TABLE bottlecaps ( name text, current_caps int STATIC, tx_id int, amount int, balance int, PRIMARY KEY ((name), tx_id) ); // initialize user INSERT INTO bottlecaps (name, tx_id, current_caps) VALUES ('ian', 'tx001', 0);
15. 15. 15 name current_caps tx_id amount balance ian 10 tx001 null null tx002 10 10 Get Some Caps BEGIN BATCH UPDATE bottlecaps SET current_caps = 10 WHERE name='ian' IF current_caps = 0; INSERT INTO bottlecaps (name, tx_id, amount, balance) VALUES ('ian', 'tx003', 7, 17); APPLY BATCH;
16. 16. 16 name current_caps tx_id amount balance ian 17 tx001 null null tx002 10 10 tx003 7 17 Get Some More Caps BEGIN BATCH UPDATE bottlecaps SET current_caps = 17 WHERE name='ian' IF current_caps = 10; INSERT INTO bottlecaps (name, tx_id, amount, balance) VALUES ('ian', 'tx003', 7, 17); APPLY BATCH;
17. 17. 17 name current_caps tx_id amount balance ian 2 tx001 null null tx002 10 10 tx003 7 17 tx004 -15 2 Spend 15 Caps BEGIN BATCH UPDATE bottlecaps SET current_caps = 2 WHERE name='ian' IF current_caps = 17; INSERT INTO bottlecaps (name, tx_id, amount, balance) VALUES ('ian', 'tx004', -15, 2); APPLY BATCH;
18. 18. 18 Summary   Do  the  work  in  your  applicaQon  code.     1.  Consistency  across  all  nodes   •  Tune  Cassandra  access  to  behave  with  “strong  consistency”     2.  Atomicity  of  TransacQons   •  Use  Cassandra  BATCH   3.  IsolaQon  of  TransacQons   •  Use  a  Lightweight  TransacQon  
19. 19. 19 Thanks  ☺     Ian  Chen   ian.chen@rakuten.com   @icchanosaur