Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
A Riak Query Tale
Search
Mathias Meyer
February 01, 2012
Programming
5
1k
A Riak Query Tale
An introduction to the abundance of ways you can get data out of Riak.
Mathias Meyer
February 01, 2012
Tweet
Share
More Decks by Mathias Meyer
See All by Mathias Meyer
Building and Scaling an Distributed and Inclusive Team
roidrage
0
1.3k
cooking infrastructure with chef
roidrage
4
230
The Message Queue is Dead, Long Live the Message Queue
roidrage
4
700
riak-js
roidrage
1
280
designing for concurrency with riak
roidrage
11
1.8k
metrics, monitoring, logging
roidrage
82
15k
design for cloud - jax 2012
roidrage
2
300
Don't Use NoSQL
roidrage
10
1k
Designing Applications for Amazon Web Services (GOTO Aarhus)
roidrage
6
350
Other Decks in Programming
See All in Programming
マルチアカウント環境での、そこまでがんばらない RI/SP 運用設計
wa6sn
0
700
Going Structural with Named Tuples
bishabosha
0
200
Kamal 2 – Get Out of the Cloud
aleksandrov
1
160
AtCoder Heuristic First-step Vol.1 講義スライド(山登り法・焼きなまし法編)
takumi152
4
1k
SEAL - Dive into the sea of search engines - Symfony Live Berlin 2025
alexanderschranz
1
120
新卒から4年間、20年もののWebサービスと 向き合って学んだソフトウェア考古学
oguri
8
7.2k
Compose Hot Reload is here, stop re-launching your apps! (Android Makers 2025)
zsmb
1
450
ベクトル検索システムの気持ち
monochromegane
30
9.8k
AHC045_解説
shun_pi
0
430
リストビュー画面UX改善の振り返り
splcywolf
0
120
PHPでお金を扱う時、終わりのない 謎の1円調査の旅にでなくて済む方法
nakka
4
1.5k
Building Scalable Mobile Projects: Fast Builds, High Reusability and Clear Ownership
cyrilmottier
2
240
Featured
See All Featured
Dealing with People You Can't Stand - Big Design 2015
cassininazir
367
26k
Fashionably flexible responsive web design (full day workshop)
malarkey
407
66k
Embracing the Ebb and Flow
colly
85
4.6k
Responsive Adventures: Dirty Tricks From The Dark Corners of Front-End
smashingmag
251
21k
Let's Do A Bunch of Simple Stuff to Make Websites Faster
chriscoyier
507
140k
How STYLIGHT went responsive
nonsquared
99
5.5k
Improving Core Web Vitals using Speculation Rules API
sergeychernyshev
12
650
Done Done
chrislema
183
16k
Mobile First: as difficult as doing things right
swwweet
223
9.6k
Statistics for Hackers
jakevdp
798
220k
Design and Strategy: How to Deal with People Who Don’t "Get" Design
morganepeng
129
19k
We Have a Design System, Now What?
morganepeng
52
7.5k
Transcript
A Riak Query Tale Mathias Meyer, @roidrage NoSQL Cologne
http://riakhandbook.com
Riak Distributed Database Fault-Tolerant Content-Agnostic Scalable on Demand
Querying Data
Key-Value $ curl localhost:8098/riak/users/roidrage
Links $ curl -‐v localhost:8098/riak/users/roidrage < HTTP/1.1 200 OK <
Link: </riak/users/klimpong>; riaktag="friend"
Links $ curl .../riak/users/roidrage/users,friend,_/
Listing Keys $ curl .../riak/users?keys=true
Don’t do that!
Streaming Keys $ curl .../riak/users?keys=stream
Avoid that!
Loads all the keys.
MapReduce
MapReduce Transform (Map) Aggregate (Reduce)
Warning: JavaScript
MapReduce riak.add("users"). map("Riak.mapValues").
run()
MapReduce var nameLength = function(value) { var doc
= Riak.mapValues(value)[0]; return [doc.length]; }
MapReduce riak.add("users"). map(nameLength).
run()
MapReduce riak.add("users"). map(nameLength).
reduce("Riak.reduceSum"). run()
MapReduce var average = function(values) { var avg
= values.reduce(function(n, sum) { return sum += n; }, 0); return [(avg / values.length)]; }
MapReduce riak.add("users"). map(nameLength).
reduce(average). run()
MapReduce riak.add("users"). map(nameLength).
reduce(average). run() Uh-Oh!
MapReduce riak.add(["users", "roidrage"]). map(nameLength).
reduce(average). run() Better!
JavaScript M/R Breaks with Millions of Objects Uses External Libraries
Serializes Data for JavaScript
Warning: Erlang
MapReduce riak.add('tweets'). map({language: 'erlang',
module: 'riak_kv_mapreduce', function: 'map_object_value'}).run()
MapReduce $ riak attach > {ok, C} = riak:local_client().
MapReduce C:mapred([{<<"users">>, <<"roidrage">>}], [{map, {modfun, riak_kv_mapreduce, map_object_value}, none, false}, {reduce,
{modfun, riak_kv_mapreduce, reduce_count_inputs}, none, true}]).
MapReduce ExtractFirstName1 = fun(RObject, _, _) -‐>
Value = riak_object:get_value(RObject), [FirstName, _] = re:split(Value, " "), [FirstName] end.
MapReduce C:mapred([{<<"users">>, <<"roidrage">>}],
[{map, {qfun, ExtractFirstName}, none, true}]).
Erlang M/R Much more efficient than JavaScript No serialization No
ad-hoc functions through HTTP
Key-Filters Reduce MapReduce input Based on key matches
Key-Filters riak.add({bucket: 'users', key_filters: [["matches", "^roid"]]})
Key-Filters riak.add({bucket: 'users', key_filters: [["to_upper"],
["matches", "^ROID"]]})
Key-Filters riak.add({bucket: 'users', key_filters: [["to_upper"],
["to_lower"], ["matches", "^roid"]]})
Key-Filters riak.add({bucket: 'users', key_filters: [["to_upper"],
["ends_with", "RAGE"]]})
Key-Filters riak.add({bucket: 'users', key_filters:
[["and", [["string_to_int"], ["less_than", 10]], [["string_to_int"], ["greater_than", 5]]]]})
Don't use key filters.
Riak 2i Sorted Secondary Indexes Simple Reverse Lookups Maintained Manually
Requires LevelDB
Riak 2i curl -‐X PUT .../riak/users/roidrage -‐d @-‐ \
-‐H "Content-‐Type: text/plain" \ -‐H "X-‐Riak-‐Index-‐firstname_bin: mathias" \ -‐H "X-‐Riak-‐Index-‐lastname_bin: meyer"
Riak 2i X-‐Riak-‐Index-‐firstname_bin: Mathias X-‐Riak-‐Index-‐lastname_bin: Meyer
Riak 2i X-‐Riak-‐Index-‐firstname_bin: Mathias X-‐Riak-‐Index-‐lastname_bin: Meyer X-‐Riak-‐Index-‐age_int: 34
Riak 2i X-‐Riak-‐Index-‐firstname_bin: Mathias X-‐Riak-‐Index-‐lastname_bin: Meyer X-‐Riak-‐Index-‐age_int: 34 X-‐Riak-‐Index-‐topics_bin: nosql,cloud,operations
Riak 2i # Match $ curl .../buckets/users/index/firstname_bin/Mathias
Riak 2i # Range $ curl .../buckets/users/index/firstname_bin/Mathias/Till
Riak 2i # Key $ curl .../buckets/users/index/$key/roidrage
Ordered Keys! (sort of)
MapReduce riak.add({bucket: 'users',
index: 'lastname_bin', key: 'mathias'}). map('Riak.mapValuesJson').run()
Riak 2i No Multi-Index Queries Requires Extra Work in the
App Returns only keys Document-partitioned
Riak Search Full-Text Search Solr-ish Interface Integrates with Riak
Riak Search curl -‐X PUT localhost:8098/riak/users -‐d @-‐ \
-‐H "Content-‐Type: application/json" {"props":{"precommit": [{"mod":"riak_search_kv_hook","fun":"precommit"} ]}}
Indexing Riak Objects curl -‐X PUT .../riak/users/roidrage \
-‐d "Mathias Meyer" -‐H "Content-‐Type: text/plain"
Solr-ish Interface curl .../solr/users/select?q=value:Mathias
Riak Search value:Mathias OR value:Till value:Mathias AND value:Meyer value:Mat* value:[Mathias
TO Till]
MapReduce riak.addSearch("users", "value:Mathias"). map("Riak.mapValues").run()
Riak Search Full text search of structured data Term-partitioned Efficient
for one term queries Multiple Interfaces No Anti-Entropy
When?
Key Listings Never! Almost
MapReduce Analytical Queries Fixed Dataset
Key Filters Never!
Riak 2i Simple Lookups and Range Queries Unbounded Queries Full
Fault-Tolerance
Riak Search Larger documents Full indexing Flexible queries Low frequency
terms
Questions?