Slide 1

Slide 1 text

Advancing Riak CS Reid Draper Thursday, May 16, 13

Slide 2

Slide 2 text

Riak CS dev github.com/reiddraper @reiddraper Thursday, May 16, 13

Slide 3

Slide 3 text

working on Riak CS since the beginning Thursday, May 16, 13

Slide 4

Slide 4 text

•so far •what’s coming •service integration •performance tips Thursday, May 16, 13

Slide 5

Slide 5 text

1.0.0 1.0.1 1.0.2 1.1.0 1.2.0 1.2.1 1.2.2 1.3.0 1.3.1 March 2012 April 2013 Thursday, May 16, 13

Slide 6

Slide 6 text

Public Release! Garbage Collection DTrace Multi-DC Replication Open Source! Multi-part Upload March 2012 April 2013 Thursday, May 16, 13

Slide 7

Slide 7 text

Since Open Sourcing Thursday, May 16, 13

Slide 8

Slide 8 text

git shortlog --summary | wc -l 29 Thursday, May 16, 13

Slide 9

Slide 9 text

easier to deploy Thursday, May 16, 13

Slide 10

Slide 10 text

How We Work Thursday, May 16, 13

Slide 11

Slide 11 text

when a feature is hard to add, we know it’s time for some refactoring Thursday, May 16, 13

Slide 12

Slide 12 text

Thursday, May 16, 13

Slide 13

Slide 13 text

gitflow (+ some variations) Thursday, May 16, 13

Slide 14

Slide 14 text

Future Work Thursday, May 16, 13

Slide 15

Slide 15 text

Keylisting (1.4) Thursday, May 16, 13

Slide 16

Slide 16 text

3 iterations Thursday, May 16, 13

Slide 17

Slide 17 text

Riak and s3 key listing is different Thursday, May 16, 13

Slide 18

Slide 18 text

Riak CS needs to be like s3 Thursday, May 16, 13

Slide 19

Slide 19 text

aardvark aardwolf Aaron Aaronic Aaronical Aaronite Aaronitic Aaru Ab aba Ababdeh Ababua abac abaca abacate abacay Riak aardvark 102 application/json aardwolf 1024 application/json Aaron 1024 application/json Aaronic 67 application/json Aaronical 462 text/plain Aaronite 105235 image/png Aaronitic 462 text/plain Aaru 1024 application/json Ab 102 application/json aba 67 application/json Ababdeh 67 application/json Ababua 462 text/plain S3 Thursday, May 16, 13

Slide 20

Slide 20 text

Fault-tolerance (1.4) Thursday, May 16, 13

Slide 21

Slide 21 text

Thursday, May 16, 13

Slide 22

Slide 22 text

Admin improvements Thursday, May 16, 13

Slide 23

Slide 23 text

Object Copy Thursday, May 16, 13

Slide 24

Slide 24 text

refcounting vs deep copy Thursday, May 16, 13

Slide 25

Slide 25 text

complexity vs. storage Thursday, May 16, 13

Slide 26

Slide 26 text

Object versioning Thursday, May 16, 13

Slide 27

Slide 27 text

immutable blocks and mvcc makes this much simpler Thursday, May 16, 13

Slide 28

Slide 28 text

BNW integration Thursday, May 16, 13

Slide 29

Slide 29 text

replication policies Thursday, May 16, 13

Slide 30

Slide 30 text

fine-grained control over replication for legal reasons Thursday, May 16, 13

Slide 31

Slide 31 text

a new backend? Thursday, May 16, 13

Slide 32

Slide 32 text

what are the backend needs? Thursday, May 16, 13

Slide 33

Slide 33 text

manifests and blocks Thursday, May 16, 13

Slide 34

Slide 34 text

Manifests sorted <10Kb Thursday, May 16, 13

Slide 35

Slide 35 text

Blocks locality only amongst same blocks in a file > 1MB Thursday, May 16, 13

Slide 36

Slide 36 text

Cloud Integration Thursday, May 16, 13

Slide 37

Slide 37 text

Open Stack Thursday, May 16, 13

Slide 38

Slide 38 text

Thursday, May 16, 13

Slide 39

Slide 39 text

Cloud Stack Thursday, May 16, 13

Slide 40

Slide 40 text

Eucalyptus Thursday, May 16, 13

Slide 41

Slide 41 text

Gotchas and Tips Thursday, May 16, 13

Slide 42

Slide 42 text

leveldb on ssd Thursday, May 16, 13

Slide 43

Slide 43 text

Blocks Manifests Thursday, May 16, 13

Slide 44

Slide 44 text

Bitcask Leveldb Thursday, May 16, 13

Slide 45

Slide 45 text

2GB Manifests / 1TB Blocks Thursday, May 16, 13

Slide 46

Slide 46 text

2TB Manifests / 1PB Blocks Thursday, May 16, 13

Slide 47

Slide 47 text

different access pattern Thursday, May 16, 13

Slide 48

Slide 48 text

archive and access tuning Thursday, May 16, 13

Slide 49

Slide 49 text

access_log_flush_factor access_log_flush_size access_log_flush_size access_archive_period access_archiver_max_backlog storage_schedule storage_archive_period usage_request_limit Thursday, May 16, 13

Slide 50

Slide 50 text

bucket sizing Thursday, May 16, 13

Slide 51

Slide 51 text

Garbage Collection Thursday, May 16, 13

Slide 52

Slide 52 text

leeway_seconds gc_interval gc_retry_interval Thursday, May 16, 13

Slide 53

Slide 53 text

10GbE Thursday, May 16, 13

Slide 54

Slide 54 text

Thursday, May 16, 13

Slide 55

Slide 55 text

zdbbl Thursday, May 16, 13

Slide 56

Slide 56 text

+zdbbl 131072 (128MB) goes in Riak vm.args Thursday, May 16, 13

Slide 57

Slide 57 text

OS config Thursday, May 16, 13

Slide 58

Slide 58 text

swappiness 0 noop or deadline Thursday, May 16, 13

Slide 59

Slide 59 text

dirty_background_ratio to dirty_background_bytes Thursday, May 16, 13

Slide 60

Slide 60 text

Thanks Kelly McLaughlin Scott Lystig Fritchie Andrew J. Stone + a bunch more Thursday, May 16, 13

Slide 61

Slide 61 text

Questions? @reiddraper Thursday, May 16, 13