Toward Eidetic Distributed File Systems. Xianzheng Dou, Jason Flinn, Peter M. Chen

Transcription

1 Toward Eidetic Distributed File Systems Xianzheng Dou, Jason Flinn, Peter M. Chen

2 Rich file system features Modern file systems store more than just data Versioning: retention of past state Provenance-aware: connections between file data Problem: High costs for providing these rich features Xianzheng Dou 1

3 Versioning FS tradeoffs Frequency of versioning Less frequent Lower storage cost More frequent Higher storage cost 2

4 Versioning FS tradeoffs Frequency of versioning Less frequent Lower storage cost More frequent Higher storage cost Ext4 2

5 Versioning FS tradeoffs Frequency of versioning Less frequent Lower storage cost More frequent Higher storage cost Versionfs WAFL 2

6 Versioning FS tradeoffs Frequency of versioning Less frequent Lower storage cost More frequent Higher storage cost Elephant FS 2

7 Versioning FS tradeoffs Frequency of versioning Less frequent Lower storage cost More frequent Higher storage cost CVFS Wayback 2

8 Versioning FS tradeoffs Frequency of versioning Less frequent Lower storage cost More frequent Higher storage cost Any past user-level state? 2

9 Versioning FS tradeoffs Frequency of versioning Less frequent Lower storage cost More frequent Higher storage cost Any past user-level state? Any past file system state and any transient state 2

10 Provenance FS tradeoffs Details of connection information Lower granulartiy Lower storage cost Higher granularity Higher storage cost 3

11 Provenance FS tradeoffs Details of connection information Lower granulartiy Lower storage cost Higher granularity Higher storage cost Ext4 3

12 Provenance FS tradeoffs Details of connection information Lower granulartiy Lower storage cost Higher granularity Higher storage cost Connections 3

13 Provenance FS tradeoffs Details of connection information Lower granulartiy Lower storage cost Higher granularity Higher storage cost PASS 3

14 Provenance FS tradeoffs Details of connection information Lower granulartiy Lower storage cost Higher granularity Higher storage cost Complete byte-level provenance? 3

15 Background: eidetic systems[osdi 14] Recall any past user-level state By How pervasive to use deterministic X-ray How to record use and X-ray replay Record RECORD Log Latency Metric RECORD Logs of PLAY Log non-deterministic events Latency Metric Replay PLAY Mail Scope Request Mail Scope Request debug_peer_list List of root causes: mydomain.com 1) queue_directory 2) 3) Michael Chow 12 debug_peer_list List of root causes: mydomain.com 1) queue_directory 2) 3) Michael Chow 12 Xianzheng Dou 4

16 Background: eidetic systems[osdi 14] Recall any past user-level state By How pervasive to use deterministic X-ray How to record use and X-ray replay Record RECORD Log Latency Metric RECORD Logs of PLAY Log non-deterministic events Latency Metric Replay PLAY Mail Scope Request Mail Scope Request Provenance at the byte granularity debug_peer_list List of root causes: mydomain.com 1) debug_peer_list List of root causes: mydomain.com 1) queue_directory 2) queue_directory 2) Intra-process lineage: 3) dynamic information 3) tracking Michael Chow 12 Michael Chow 12 Inter-process lineage: data flow dependency graph Xianzheng Dou 4

17 A clean-sheet design of FS Eidetic systems prototype Graft eidetic support onto an existing FS Consider only local storage An eidetic distributed file system A small number of personal devices + cloud servers New design choices Fundamental unit of persistent storage File transfer Xianzheng Dou 5

18 Traditional distributed FS Xianzheng Dou 6

19 Traditional distributed FS Xianzheng Dou 6

20 Traditional distributed FS Xianzheng Dou 6

21 Traditional distributed FS Xianzheng Dou 6

22 Eidetic distributed file systems Xianzheng Dou 7

23 Eidetic distributed file systems Xianzheng Dou 7

24 Fundamental unit What is the fundamental unit of persistent storage? Xianzheng Dou 8

25 Fundamental unit What is the fundamental unit of persistent storage? Xianzheng Dou 8

26 Fundamental unit What is the fundamental unit of persistent storage? Replay Xianzheng Dou 8

27 Fundamental unit What is the fundamental unit of persistent storage? Xianzheng Dou 9

28 Fundamental unit What is the fundamental unit of persistent storage? Xianzheng Dou 9

29 Fundamental unit What is the fundamental unit of persistent storage? Fundamental unit: Logs of non-determinism Files are only considered as caches Xianzheng Dou 9

30 File persistency When is a file considered persistent on the server? Xianzheng Dou 10

31 File persistency When is a file considered persistent on the server? As long as logs generating the data are persistent Xianzheng Dou 10

32 File persistency When is a file considered persistent on the server? Xianzheng Dou 10

33 Updating server cache Should the server cache the file version?? Xianzheng Dou 11

34 Updating server cache Should the server cache the file version?? Probability of future access Costs for regeneration Xianzheng Dou 11

35 File transfer methods How are files transferred to the server? Xianzheng Dou 12

36 File transfer methods How are files transferred to the server? Xianzheng Dou 12

37 File transfer methods How are files transferred to the server? Xianzheng Dou 13

38 File transfer methods How are files transferred to the server? Xianzheng Dou 13

39 File transfer methods How are files transferred to the server? Compare computation costs with communication costs - by value (file data) - or by replay Xianzheng Dou 13

40 Read path How should a client read a particular version? Xianzheng Dou 14

41 Read path How should a client read a particular version? Xianzheng Dou 14

42 Available transfer methods How should a client read a particular version? Xianzheng Dou 15

43 Available transfer methods How should a client read a particular version? Xianzheng Dou 15

44 Available transfer methods How should a client read a particular version? Xianzheng Dou 15

45 Available transfer methods How should a client read a particular version? Xianzheng Dou 16

46 Available transfer methods How should a client read a particular version? Xianzheng Dou 16

47 Available transfer methods How should a client read a particular version? Xianzheng Dou 16

48 Available transfer methods How should a client read a particular version? Xianzheng Dou 17

49 Available transfer methods How should a client read a particular version? By value By replay on the client By replay on the server From peers Xianzheng Dou 17

50 Choosing the right method How should a client read a particular version? Server has the most complete knowledge Metrics User waiting time Monetary cost Client energy consumption Xianzheng Dou 18

51 Feasibility Eidetic system overheads Record 4 years of workstation data on a 4TB hard disk Under 8% performance overhead on most benchmarks Applications (log size vs. diff size) Logs are smaller image/audio editing, latex, make, slides editing Diffs are smaller: text editing File sharing Most files are not shared Xianzheng Dou 19

52 Conclusions A new point in the design space of Versioning file systems Provenance-aware file systems Hypothesis More effective in versioning and provenance Achieving reasonable overheads Under implementation Xianzheng Dou 20

53 Thank you! Xianzheng Dou 21