Acunu & OCaml: Experience Report, CUFP
2 likes1,431 views
This document summarizes Tom Wilkie's presentation on Acunu & OCaml. It discusses how Acunu has evolved from small databases in 1990 to distributed, shared-nothing databases today. It presents the architecture and components of Acunu's storage core, which is built using OCaml. Performance results are shown for Acunu's prototypes using different data structures like doubling arrays, demonstrating high insertion rates.
![References
[LSM] The Log-Structured Merge-Tree (LSM-Tree)
Patrick O'Neil, Edward Cheng, Dieter Gawlick,
Elizabeth O'Neil Stratified B-trees and versioned dictionaries, - Andy
http://staff.ustc.edu.cn/~jpq/paper/flash/1996-The Twigg, Andrew Byde, Grzegorz Miłoś, Tim Moreton,
%20Log-Structured%20Merge-Tree%20%28LSM- John Wilkes, Tom Wilkie, HotStorage’11
Tree%29.pdf http://www.usenix.org/event/hotstorage11/tech/
final_files/Twigg.pdf
[COLA] Cache-Oblivious Streaming B-trees,
Michael A. Bender et al [RDA] Random duplicate storage strategies for
http://www.cs.sunysb.edu/~bender/newpub/ load balancing in multimedia servers, 2000, Joep
BenderFaFi07.pdf Aerts and Jan Korst and Sebastian Egner
http://www.win.tue.nl/~joep/IPL.ps
[DSST] Making Data Structures Persistent - J. R.
Driscoll, N. Sarnak, D. D. Sleator, R. E. Tarjan, Making Apache, Apache Cassandra, Cassandra, Hadoop, and
Data Structures Persistent, Journal of Computer the eye and elephant logos are trademarks of the
and System Sciences,Vol. 38, No. 1, 1989 Apache Software Foundation.
http://www.cs.cmu.edu/~sleator/papers/making-
data-structures-persistent.pdf](https://image.slidesharecdn.com/cufp-110924103541-phpapp02/85/Acunu-OCaml-Experience-Report-CUFP-32-320.jpg)
Acunu & OCaml: Experience Report, CUFP
- 1. Acunu & OCaml: Experience Report Tom Wilkie Founder & VP Engineering tom@acunu.com @tom_wilkie
- 2. What do we do? 1990 Small databases BTree indexes BTree File systems RAID Old hardware
- 3. What do we do? 2010 Distributed, shared-nothing databases Write-optimised indexes Write-optimised indexes BTree file systems BTree file systems RAID ... RAID New hardware New hardware
- 4. What do we do? 2011 Distributed, shared-nothing databases Castle Castle ... New hardware New hardware
- 6. What does this have to do with Functional Programming?
- 7. Big Appl Java, Amazon S3 compatible Open API Erlang, OCaml Management C ... Deployment Monitoring Python, ... ... ... ... Bash, Acunu Storage Core C Perl ... ... Cros Manag
- 9. Autogeneranted HTML5/JavaScript External Monitoring OCaml CLI User Interface Tools (Munin etc) Routerd Another Routerd enumeration, routing, clustering on a different machine FSd Cassandrad S3d Miscd Clusterd Statsd AlertsD Default_ Alert_ Disk Keyspace Bucket Base Host Report Rule Column NamedObje Version Group Report Alert Family cts Collection Stat Cassandra_ Filesystem S3_Node Service Source Node Castle Cassandra BigS3
- 10. FSd Cassandrad S3d Miscd Disk Keyspace Bucket Base Column NamedObje Version Family cts Bridges to Collection other systems Cassandra_ Filesystem S3_Node Node Castle Cassandra BigS3
- 11. Miscd Clusterd Statsd AlertsD Default_ Alert_ Base Host Report Rule Clustering NamedObje Group Report Alert cts Failure Detection Stat Monitoring Service Source Alerting
- 12. neranted HTML5/JavaScript External M ml CLI User Interface Tools (Mu Routerd enumeration, routing, clustering drad S3d Miscd Clusterd St Routing & Aggregation De ace Bucket Base Host Re
- 15. Aim: Investigate algorithms for KV storage • CoW BTrees • Fractional Cascading • Mod List BTrees • Stratified DAs • LSM Trees • Multidimensional keys • Doubling Arrays • Z curve packing
- 16. Doubling Array 2 2 9 9
- 17. Doubling Array Inserts 11 2 9 2 8 9 11 8 8 11 etc... Similar to log-structured merge trees (LSM), cache- oblivious lookahead array (COLA), ...
- 19. 8KB @ 100MB/s, w/ 8ms seek 100 / 5 = 100 IOs/s = 20 updates/s ~ log (2^30)/log 100 = 5 IOs/update Range Query Update (Size Z) Log Structured O(logB N) O(Z/B) B-Tree random IOs random IOs O((log N)/B) O(Z/B) Doubling Array sequential IOs sequential IOs ~ log (2^30)/100 8KB @ 100MB/s 13k / 0.2 = 0.2 IOs/update = 13k IOs/s = 65k updates/s B = “block size”, say 8KB at 100 bytes/entry ~= 100 entries
- 20. Block Index BTree Disk Trace Time (s)
- 21. Block Index Doubling Array Disk Trace Time (secs)
- 22. Insertion Rate (kvps/s) OCaml Prototype Performance # inserted kvps
- 23. The Dark Side...
- 24. Insert Rate (keys/s) Java Prototype Performance Time (s)
- 29. SH OT S* SN AP * And clones!
- 30. I’ll explain how.... “Castle: Re-inventing Storage For Big Data” London, 27th September http://bit.ly/rduBia
- 31. Questions? tom@acunu.com @tom_wilkie http://www.acunu.com http://bitbucket.org/acunu http://github.com/acunu
- 32. References [LSM] The Log-Structured Merge-Tree (LSM-Tree) Patrick O'Neil, Edward Cheng, Dieter Gawlick, Elizabeth O'Neil Stratified B-trees and versioned dictionaries, - Andy http://staff.ustc.edu.cn/~jpq/paper/flash/1996-The Twigg, Andrew Byde, Grzegorz Miłoś, Tim Moreton, %20Log-Structured%20Merge-Tree%20%28LSM- John Wilkes, Tom Wilkie, HotStorage’11 Tree%29.pdf http://www.usenix.org/event/hotstorage11/tech/ final_files/Twigg.pdf [COLA] Cache-Oblivious Streaming B-trees, Michael A. Bender et al [RDA] Random duplicate storage strategies for http://www.cs.sunysb.edu/~bender/newpub/ load balancing in multimedia servers, 2000, Joep BenderFaFi07.pdf Aerts and Jan Korst and Sebastian Egner http://www.win.tue.nl/~joep/IPL.ps [DSST] Making Data Structures Persistent - J. R. Driscoll, N. Sarnak, D. D. Sleator, R. E. Tarjan, Making Apache, Apache Cassandra, Cassandra, Hadoop, and Data Structures Persistent, Journal of Computer the eye and elephant logos are trademarks of the and System Sciences,Vol. 38, No. 1, 1989 Apache Software Foundation. http://www.cs.cmu.edu/~sleator/papers/making- data-structures-persistent.pdf