In the words of Jonas Boner "There is no such thing as a stateless architecture It's just someone else's problem". While a stateless broker can be easily made HA by using a load balancer it is the stateful once which are challenging. Service Brokers are critical for PaaS providers and a downtime has a potential to create outages. A line of defense can be BOSH resurrection however it takes minutes to bring up all the various processes for example a Stemcell update may take around 5-10 minutes. And this is unacceptable by any standards where the expectation is to be 99.99% available.
In this talk Ashish and Shashank will share how Service Fabrik broker from Cloud Foundry community has been implemented as a highly available, disaster resistant with failover time of few seconds. In this talk implementation specifics on AWS, Azure, GCP and OpenStack platforms will be covered followed a by a demo on one of the IaaS.
Shashank has 20 years of work experience with 8 years in cloud and distributed systems domain. Shashank holds more then 30 patents and has been a speaker in various cloud foundry summits and other conferences.