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
SWIM: Scalable Weakly Consistent Infection Styl...
Search
Sponsored
·
Your Podcast. Everywhere. Effortlessly.
Share. Educate. Inspire. Entertain. You do you. We'll handle the rest.
→
Paul Hinze
July 22, 2015
Technology
420
2
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
SWIM: Scalable Weakly Consistent Infection Style Process Group Membership Protocol
Papers We Love, Chicago
July 22, 2015
Paul Hinze
July 22, 2015
More Decks by Paul Hinze
See All by Paul Hinze
Getting Good at System Failure Analysis
phinze
0
500
Applying Graph Theory to Infrastructure As Code
phinze
6
2.2k
Infrastructure as Code with Terraform and Friends
phinze
2
350
Smoke & Mirrors: The Primitives of High Availability
phinze
1
790
Git: Everybody's Favorite MMO
phinze
0
240
Shut Up and Pipe! Unix-style Object Collaboration in Rack and Vagrant
phinze
0
210
Freighthop: Vagrant on Rails
phinze
0
350
Puppet Modules Are Our Friends
phinze
0
150
Who Needs Clouds?: HA in Your Datacenter
phinze
1
580
Other Decks in Technology
See All in Technology
ClaudeCodeでセキュリティ監視業務を半自動化_1年の運用でわかったAIに任せる設計の5つのポイント
kintotechdev
3
900
紙 to デジタル
ichien178
0
370
わたしが知り合いゼロの勉強会に 行けるようになるまで
r5ni4
2
830
Sansan Engineering Unit 紹介資料
sansan33
PRO
1
5k
Master Dataグループ紹介資料
sansan33
PRO
1
4.8k
Eight Engineering Unit 紹介資料
sansan33
PRO
3
8.3k
ハッカソンで入賞した話 @ Findy LT
asari194617
0
1.5k
AI駆動開発はどこまで来たのか? ファインディの最新実態調査で読み解く現在地 Devin Con Tokyo
akiratom
4
1.9k
The Django UUID Story - DjangoCon US 2026
pauloxnet
0
310
型落ちシンクライアント端末のPoEモジュールを自作したかった話
logica0419
0
510
When Token Pruning is Worse than Random: Understanding Visual Token Information in VLLMs
sansantech
PRO
0
130
markdown-poster Introduction
kazamori
0
400
Featured
See All Featured
How to Create Impact in a Changing Tech Landscape [PerfNow 2023]
tammyeverts
56
3.4k
The Cult of Friendly URLs
andyhume
79
7k
Agile Leadership in an Agile Organization
kimpetersen
PRO
0
210
Building the Perfect Custom Keyboard
takai
2
850
Why Mistakes Are the Best Teachers: Turning Failure into a Pathway for Growth
auna
0
220
GitHub's CSS Performance
jonrohan
1033
470k
Exploring the relationship between traditional SERPs and Gen AI search
raygrieselhuber
PRO
2
4.3k
Lightning talk: Run Django tests with GitHub Actions
sabderemane
0
240
Bootstrapping a Software Product
garrettdimon
PRO
306
120k
Reflections from 52 weeks, 52 projects
jeffersonlam
356
21k
B2B Lead Gen: Tactics, Traps & Triumph
marketingsoph
0
220
The Pragmatic Product Professional
lauravandoore
37
7.4k
Transcript
SWIM Scalable Weakly Consistent Infection Style Process Group Membership Protocol
Paul Hinze phinze
Paul Hinze phinze death stare
Armon Dadgar armon creator of Serf and Consul
ma
None
None
Process Group Membership Protocol Who is alive
None
Process Group Membership Protocol
SWIM Scalable Weakly Consistent Infection Style Process Group Membership Protocol
Scalable The SWIM effort is motivated by the unscalability of
traditional heartbeating protocols.
Heartbeating A B C A A B B Failure Detection
+ Membership
Heartbeating
Evaluating Protocols Completeness Speed Accuracy Overhead
Evaluating Protocols Completeness Speed Accuracy Overhead heartbeating Yes Limit *
Interval High Nodes2 !
Key Insight Failure Detection State Updates. Solve separately from
Failure Detection ping! ack! A B C D {B,C,D}
Failure Detection ping! ack! A B C D {B,C,D}
Indirect Ping ping(C)! ack! B C D ping(C)! ping ack
fail {B,C,D}
Indirect Ping ping(C)! B C D ping(C)! fail ...C is
dead! fail
Key Insight Failure Detection State Updates. Solve separately from
Key Insight State Updates Failure Detection Piggyback onto messages.
State Updates ping! (B is dead) ack! (D just joined)
A B C D {B,C} {D}
Infection Style A B C D {B,C} {A,B,D} {A,B} weakly
consistent
Evaluating Protocols Completeness Speed Accuracy Overhead SWIM Yes, eventually 1
* Interval High-ish O(N)
Improvements Time Bounded Completeness Increased Accuracy
Completeness A B {B, C, D, ..., N} N
Completeness A B {B, C, D, ..., N} N 1.
Shuffle List 2. Iterate
Completeness Fixed Time
Evaluating Protocols Completeness Speed Accuracy Overhead SWIM Yes, fixed time
1 * Interval High-ish O(N)
Accuracy B C D ...C is MAYBE dead! A
Accuracy B C D I heard C might be dead.
A
Accuracy B C D I'm not dead! A
Accuracy
Evaluating Protocols Completeness Speed Accuracy Overhead SWIM Yes, fixed time
1 * Interval High O(N)
Evaluating Protocols Completeness Speed Accuracy Overhead SWIM Yes, fixed time
1 * Interval High O(N)
None
Limitations Update Latency Problem Solution Separate Gossip Timer
Limitations Cannot Handle Network Partitions Problem Solution Track and Retry
Recently Dead Nodes
Limitations No Concept of Graceful Leave (vs Failure) Problem Solution
Broadcast and Tracking of "Intents"
Limitations New Nodes Take Too Long Materialize Initial State Problem
Solution Anti-entropy TCP State Syncs
Limitations No built-in facility for user data Problem Solution Implement
user payloads (ordering via lamport clocks)
Limitations No peer metadata, only IP Addresses Problem Solution Inject
versioned peer metadata into state messages
Limitations No encryption Problem Solution Implement AES-GCM with key rotation
56 nodes 2K nodes Performance
Implementations Memberlist https://github.com/hashicorp/memberlist Serf lib https://github.com/hashicorp/serf Serf http://www.serfdom.io Consul http://www.consul.io
Implementations events, queries, and scripts serf CLI serf lib memberlist
Implementations service discovery, K/V, health checks consul serf lib memberlist
Implementations service discovery, K/V, health checks
Thanks Cloud icons by Julien Deveaux from the Noun Project