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
(Advanced) Redis
Search
Sponsored
·
Ship Features Fearlessly
Turn features on and off without deploys. Used by thousands of Ruby developers.
→
Jan De Poorter
February 12, 2013
Programming
8
400
(Advanced) Redis
Everything we learned at a client project using Redis - Talk given at #ArrrrUG
Jan De Poorter
February 12, 2013
Tweet
Share
Other Decks in Programming
See All in Programming
CSC307 Lecture 04
javiergs
PRO
0
660
Grafana:建立系統全知視角的捷徑
blueswen
0
330
例外処理とどう使い分ける?Result型を使ったエラー設計 #burikaigi
kajitack
16
6k
CSC307 Lecture 01
javiergs
PRO
0
690
Vibe Coding - AI 驅動的軟體開發
mickyp100
0
170
Amazon Bedrockを活用したRAGの品質管理パイプライン構築
tosuri13
4
250
Spinner 軸ズレ現象を調べたらレンダリング深淵に飲まれた #レバテックMeetup
bengo4com
1
230
OCaml 5でモダンな並列プログラミングを Enjoyしよう!
haochenx
0
140
【卒業研究】会話ログ分析によるユーザーごとの関心に応じた話題提案手法
momok47
0
190
Fragmented Architectures
denyspoltorak
0
150
カスタマーサクセス業務を変革したヘルススコアの実現と学び
_hummer0724
0
630
なるべく楽してバックエンドに型をつけたい!(楽とは言ってない)
hibiki_cube
0
140
Featured
See All Featured
How to audit for AI Accessibility on your Front & Back End
davetheseo
0
180
Digital Ethics as a Driver of Design Innovation
axbom
PRO
1
170
Designing for Performance
lara
610
70k
Reflections from 52 weeks, 52 projects
jeffersonlam
356
21k
The Web Performance Landscape in 2024 [PerfNow 2024]
tammyeverts
12
1k
The Power of CSS Pseudo Elements
geoffreycrofte
80
6.1k
New Earth Scene 8
popppiees
1
1.5k
How to Grow Your eCommerce with AI & Automation
katarinadahlin
PRO
0
100
Organizational Design Perspectives: An Ontology of Organizational Design Elements
kimpetersen
PRO
1
110
Evolving SEO for Evolving Search Engines
ryanjones
0
120
GitHub's CSS Performance
jonrohan
1032
470k
A better future with KSS
kneath
240
18k
Transcript
(advanced) Redis Advanced is between brackets because we don’t claim
to be advanced users, we just have gained a lot of experience
Jan De Poorter @defv
Redis?
NoSQL
:-D
No, but srsly
Redis is an in-memory key-value data-structure server
in memory with “best effort” persistence
Data structure server Strings SET, SETNX, GET, INCR, DECR, APPEND,
... Hashes HSET, HGET, HGETALL, HDEL, HLEN, ... Lists LPUSH, LPOP, LINDEX, LLEN, LRANGE, ... Sets & Sorted sets SADD, SPOP, SMEMBERS, SDIFF, ZADD, ...
some libraries Ruby redis-rb (gem) hiredis-rb (gem) Node redis (npm)
hiredis (npm)
Q-Music Our use case was the API for the new
Q-Music website and the iPhone and Android applications.
High volume
Pub-Sub
BG jobs with Sidekiq
Some do’s
Serialize values
2 class QApi::Redis 3 class << self 4 def redis
5 @redis ||= Redis.new(config) 6 end ... 181 def set(key, data) 182 redis.set key, to_json(data) 183 end 184 185 def get(key) 186 from_json redis.get(key) 187 end 188 216 private 217 def from_json(result) 218 case result 219 when Array 220 result.map { |r| from_json(r) } 221 when Hash 222 result 223 when nil 224 nil 225 else 226 MultiJson.decode(result) 227 end 228 rescue MultiJson::DecodeError 229 result 230 end 231 232 def to_json(data) 233 MultiJson.encode(data) 234 end 262 end 263 end
Save references
# Write LPUSH plays {“id”: 42, ”title”:”99 Bottles of Rum”}
LPUSH plays {“id”: 49, ”title”:”The Drunken Sailor”} LPUSH plays {“id”: 42, ”title”:”99 Bottles of Rum”} # Read LRANGE plays 0 -1 don’t do this
it’s fine with 100 items
but becomes huge with 90 000 items
or 1 000 000 Justin Bieber mentions ;-)
# Write SET tracks:42 {“id”: 42, ”title”:”99 Bottles of Rum”}
SET tracks:49 {“id”: 42, ”title”:” The Drunken Sailor”} LPUSH plays tracks:42 LPUSH plays tracks:49 LPUSH plays tracks:42 # Read LRANGE plays 0 -1 MGET tracks:42 tracks:49 do this instead
or in Ruby 82 def list_from_references(list, options) 83 references =
redis.lrange list, 84 options[:start], options[:stop] 85 if references.any? 86 from_json redis.mget(*references) 87 else 88 [] 89 end 90 end
Some don’t’s
RDB vs AOF
RDB does fork()
doubling memory usage
Our dataset was 4GB
We had ~8GB of memory
No BGSAVE’s for 4 days
:-(
appendonly yes
Slow(er) start-up
Can be optimized
Monitoring
data:~$ redis-cli redis > INFO redis_version:2.4.16 ... connected_clients:152 ... used_memory:2744963744
used_memory_human:2.56G used_memory_peak:6150240632 used_memory_peak_human:5.73G ... changes_since_last_save:9576 last_save_time:1360661315 redis >
PUBLISH & SUBSCRIBE overload
WebSocket clients
1 client == 1 connection == 1 subscription
Works perfect in development
Works perfect in staging
but in production...
1 new event
sent to 3500 connected clients
Takes about 0.3 seconds
Up to 4 messages per second
:-(
Move logic to websocket server
:-(
1 type of msg == 1 subscription
:-)
MULTI / EXEC
> MULTI OK > RENAME bl:1 bld:1 QUEUED > RENAME
bl:2 bld:2 QUEUED > EXEC 1) OK 2) OK
Works great on 1000’s of commands
Not so much on 1 million commands...
Don’t exceed ~10k commands in 1 transaction
master/slave
no library support
DIY
We didn’t
Limited connections
Redis < 2.6
/* Max number of fd supported */ #define AE_SETSIZE (1024*10)
No more then ~10k connections allowed
Redis >= 2.6
The sky is the limit
(actually, the linux max FD is the limit)
Questions?