Post on 27-Jun-2015
description
I work on a PaaS…
“For us, PaaS is a set of services aimed at developers that helps them develop and test apps without having to worry about the underlying infrastructure. Developers don't want to have to worry about provisioning the servers, storage and backup associated with developing and launching an app.”
[1] http://www.javaworld.com/article/2078801/java-app-dev/forrester--paas-makes-developers-happy.html[2] http://www.infoworld.com/d/cloud-computing/paas-primer-what-platform-service-and-why-does-it-matter-212696
And I believed ‘em…
I work on a PaaS…
“For us, PaaS is a set of services aimed at developers that helps them develop and test apps without having to worry about the underlying infrastructure. Developers don't want to have to worry about provisioning the servers, storage and backup associated with developing and launching an app.”
[1] http://www.javaworld.com/article/2078801/java-app-dev/forrester--paas-makes-developers-happy.html[2] http://www.infoworld.com/d/cloud-computing/paas-primer-what-platform-service-and-why-does-it-matter-212696
And I believed ‘em…
… but then I learned otherwise
Application Instances and Availability Zones
Rout
erDEA
DEA
DEA
Cloud Foundry Elastic Runtime
DEA
DEA
DEA
Zone 1 Zone 2
App Ops
Application Instances and Availability Zones
Rout
erDEA
DEA
DEA
Cloud Foundry Elastic Runtime
DEA
DEA
DEA
Zone 1 Zone 2 Application instances are evenly distributed over two availability zones.
App Ops
Application Instances and Availability Zones
Rout
erDEA
DEA
DEA
Cloud Foundry Elastic Runtime
DEA
DEA
DEA
Zone 1 Zone 2 Application instances are evenly distributed over two availability zones.
Loosing an AZ keeps instances running and available.
App Ops
Rout
er
Failed Application Instances Replaced
BlobstoreCloud Controller
Health Manager
Messaging(NATS)
DEA DEA DEA
App Ops
Cloud Foundry Elastic Runtime
Rout
er
Failed Application Instances Replaced
BlobstoreCloud Controller
Health Manager
Actual State
Messaging(NATS)
DEA DEA DEA
App Ops
Cloud Foundry Elastic Runtime
Rout
er
Failed Application Instances Replaced
BlobstoreCloud Controller
Health Manager
Actual StateDesired State
Messaging(NATS)
DEA DEA DEA
App Ops
Cloud Foundry Elastic Runtime
Rout
er
Failed Application Instances Replaced
BlobstoreCloud Controller
Health Manager
Actual StateDesired State
Messaging(NATS)
DEA DEA DEA
App Ops
Cloud Foundry Elastic Runtime
Rout
er
Failed Application Instances Replaced
BlobstoreCloud Controller
Health Manager
Actual StateDesired State
Messaging(NATS)
DEA DEA DEA
App Ops
Cloud Foundry Elastic Runtime
Message Bus
ERS Processes are Monitored
IaaSCloud Foundry BOSH
Health Monitor Health Manager
DEA
Cloud Controller
A
A
A
Responders:
pager
monitoring
…
PaaS Ops
Message Bus
ERS Processes are Monitored
IaaSCloud Foundry BOSH
Health Monitor Health Manager
DEA
Cloud Controller
A
A
A
Responders:
pager
monitoring
…
PaaS Ops
Message Bus
ERS Processes are Monitored
IaaSCloud Foundry BOSH
Health Monitor Health Manager
DEA
Cloud Controller
A
A
A
Responders:
pager
monitoring
…
PaaS Ops
Message Bus
VMs are Monitored
IaaS
Health Monitor Health Manager
Cloud Controller
A
DEA
A
A
Responders:
pager
monitoring
resurrector
…
BOSH Director
Actual State
DEA
A
PaaS Ops
Cloud Foundry BOSH
Message Bus
VMs are Monitored
IaaS
Health Monitor Health Manager
Cloud Controller
A
DEA
A
A
Responders:
pager
monitoring
resurrector
…
BOSH Director
Actual State
DEA
A
PaaS Ops
Cloud Foundry BOSH
Desired State
Message Bus
VMs are Monitored
IaaS
Health Monitor Health Manager
Cloud Controller
A
DEA
A
A
Responders:
pager
monitoring
resurrector
…
BOSH Director
Actual StateDesired State
PaaS Ops
Cloud Foundry BOSH
Message Bus
VMs are Monitored
IaaS
Health Monitor Health Manager
Cloud Controller
A
A
Responders:
pager
monitoring
resurrector
…
BOSH Director (CP
I)Actual StateDesired State
DEA
A
PaaS Ops
Cloud Foundry BOSH
Four levels of HA in PCF
Elastic Runtime (ERS):
Distribution across availability zones
Application health management and recovery
BOSH (cluster management):
Process monitoring, recovery and alerting
Virtual machine health monitoring, recovery and alerting
Too slow for you?
How about <28 seconds?
http://blog.gopivotal.com/cloud-foundry-pivotal/products/the-four-levels-of-ha-in-pivotal-cf
Thank You!
@cdavisafc