keynote #tech - google : aperçu de la gestion des services distribués chez google, par bastien...
Post on 22-Jan-2017
448 Views
Preview:
TRANSCRIPT
Bastien LEGRAS, Google
Google : aperçu de la gestion des services distribués chez Google
De Borg à Kubernetes ou l’art d’ordonnancer des tâches
Bastien Legras (@bastienlegras)Google Cloud Platform - Solution Engineering
Image by
Connie Zhou
job hello_world = { runtime = { cell = 'ic' } // Cell (cluster) to run in binary = '.../hello_world_webserver' // Program to run args = { port = '%port%' } // Command line parameters requirements = { // Resource requirements ram = 100M disk = 100M cpu = 0.1 } replicas = 5 // Number of tasks}
10000
User view
(optional) .
User view
What justhappened?
web browsers
BorgMaster
link shard
UI shardBorgMaster
link shard
UI shardBorgMaster
link shard
UI shardBorgMaster
link shard
UI shard
Cell
Scheduler
borgcfg web browsers
scheduler
Borglet Borglet Borglet Borglet
BorgMaster
link shard
read/UI
shard
Config file
persistent store (Paxos)
Binary
User view
Hello world! Hello
world!Hello
world!
Hello world!
Hello world!
Hello world! Hello
world!
Hello world!
Hello world!
Hello world!
Hello world!
Hello world!
Hello world!
Hello world!
Hello world!
Hello world!
Hello world!Hello world!
Hello world!
Hello world!
Hello world!
Hello world!
Hello world! Hello
world!
Hello world!
Hello world!
Hello world!
Image by Connie Zhou
User view
Hello world!
Hello world!
Hello world! Hello
world!
Hello world!
Hello world!
Hello world!
Hello world!
Hello world!
Hello world!
Hello world! Hello
world!
Hello world!
Hello world!
Hello world!
Hello world!
Hello world!
Hello world!
Hello world! Hello
world!
Hello world!
Hello world!
Hello world!
Hello world!
User view
task-eviction ratesand causes
9
Failures
Images by
Connie Zhou
A 2000-machine service will have >10 task exits per dayThis is not a problem: it's normal
Failures
Multiple apps per machine
Sharing clusters between prod/batch
Advanced bin-packing algorithms
Efficiency
stranded resources
available resources
onemachine
Multiple apps per machine
Sharing clusters between prod/batch
Advanced bin-packing algorithms
Efficiency
A few other moving parts
web browsers
BorgMaster
link shard
UI shardBorgMaster
link shard
UI shardBorgMaster
link shard
UI shardBorgMaster
link shard
UI shard
Cell
Scheduler
borgcfg web browsers
scheduler
Borglet Borglet Borglet Borglet
BorgMaster
link shard
read/UI
shard
Config file
persistent store (Paxos)
Binary
app
agent
masterjob config
A few other moving parts
app
agent
master
system config
monitoring
security
accounting/planning
binaries + data distribution
job config
storage
Diagram from an original by Cody Smith.
A few other moving parts
A few other moving parts
app
agent
master
system config
monitoring
security
accounting/planning
binaries + data distribution
job config
storage
Diagram from an original by Cody Smith.
Building what’s next 17
Containers at Google
Developed as the only practical way to manage Google-scale compute
Everything at Google runs in a containerWe launch over2 Billioncontainers per week.
κυβερνήτης: pilot or helmsman of a ship
http://kubernetes.io
Kubernetes
Kubernetes
Direct Borg analogues:
● Borg containers => Docker containers
● alloc (task group) => pod (container group)
● Borglet => Kubelet● Paxos => persistent, declarative
specs● reconciliation loops
New / improved:
● labels + label queries● service abstraction● composable microservices● IP per pod
Kubernetes
Kubernetes RoadmapMilestones• 1.0 stable in June 2015• 1.1 Nov 9th 2015 (perf, pod autoscaling, http LB,
no APi change)
Cloud Native Computing Foundation• Linux Foundation• VMWare, RedHat, Joyent, Mesosphere, Docker,
IBM, CloudFoundry, CoreOS, Cisco, Google, …• cncf.io
Kubernetes: zero downtime update at 1 million requests per secondScales
Google Container Engine (GKE)Google Container Registry
bastien@google.com@bastienlegras
http://kubernetes.io http://goo.gl/1C4nuo (Borg paper)
Images by Connie Zhou
RETROUVEZ TOUTE L’ACTUALITE DE L’EVENEMENT SUR LE
COMPTE TWITTER OFFICIEL @OSS_PARIS
top related