Apache BookKeeper - A High Performance and Low Latency Storage

1 .Apache BookKeeper A High Performance and Low Latency Storage Service @sijieg (Sijie Guo, Twitter) @jvjujjuri (JV, Salesforce) 

2 .Hello! I am Sijie Guo - PMC Chair of Apache BookKeeper - Co-creator of Apache DistributedLog - Twitter Messaging/Pub-Sub Team - Yahoo! R&D Beijing 

3 .Challenges in Distributed Systems 

4 . Expect Failures up to 10% annual failure rates for disks/servers 

5 .Symptoms “ 

6 .Problem 1: Not Available 

7 .Problem 1: Not Available 

8 .Problem 2: Inconsistencies 

9 .CAP 

10 .More Issues “ 

11 . Problem 3: Split Brain Two Writers Writer A Writer A Writer A Write A’ Write A’ 

12 .Problem 4: Failure Detection B A C 

13 .Problem 5: Recovery B A Consistency C Recovery Protocol 

14 .Solutions “ 

15 .Overview Enter Apache BookKeeper 

16 . BookKeeper - Durable Storage A Durable Storage Optimized for Immutable Data Serve as a building block for reliable systems Client Library Replication Consistency Recovery Durability Commodity Hardware 

17 .Immutable Data Abstraction 

18 .Ledger ◉ Segment ◉ Block / Object ◉ Append-Only File ◉ ... 

19 .Guarantees If an entry has been acknowledged, it must be readable If an entry is read once, it must always be readable 

20 .History ◉ Initial Use Case - Hadoop NameNode HA ◉ 2008: Open Sourced Contrib of ZooKeeper ◉ 2011: Sub-Project of ZooKeeper ◉ 2012: Yahoo! Push Notification ◉ 2012~Now: DistributedLog, Pulsar, Majordodo ◉ 2015~Now: Salesforce Distributed Store 

21 .Details Inside of Apache BookKeeper 

22 .Architecture Metadata Store Bookie Client APP Bookie Ledger Bookie 

23 .Reliable Writes Bookie ◉ Store checksum along with entry ◉ Fsync entries before responding ◉ Ack when Bookie Accepted ○ All Previous Entries by ○ This Entry Quorum Bookie 

24 .Consistency - LastAddPushed Writer Add entries 0 1 2 3 4 7 8 9 10 11 12 LastAdd Pushed 

25 .Consistency - LastAddConfirmed Ownership Changed Writer Writer Add entries Ack Adds 0 1 2 3 4 7 8 9 10 11 12 Fencing LastAdd LastAdd Confirmed Confirmed Reader Reader 

26 .Fencing 

27 . Read Entry & Read LAC Read Entry K Read LAC Client Client Speculative Reads Quorum Read On Timeouts B1 B2 B3 B1 B2 B3 

28 .Long Poll Read Client Speculative Long Poll Read Long Poll B1 B2 B3 

29 .Inside a Bookie