There is a tremendous amount of awesome work happening in this space. Tsuru is one such great project, thanks for adding it to the thread.
One big difference is that Convox uses managed services over writing and running our own components. For example, we use DynamoDB for saving build and release records. And we use ELBs for load balancing and routing.
This approach ties us closely to AWS, but results in a stack of infrastructure that has minimal operational overhead (e.g. Amazon is keeping DynamoDB Available, you don't have to manage an etcd cluster), and great scaling potential (vs tuning and scaling individual HAProxy nodes).
One big difference is that Convox uses managed services over writing and running our own components. For example, we use DynamoDB for saving build and release records. And we use ELBs for load balancing and routing.
This approach ties us closely to AWS, but results in a stack of infrastructure that has minimal operational overhead (e.g. Amazon is keeping DynamoDB Available, you don't have to manage an etcd cluster), and great scaling potential (vs tuning and scaling individual HAProxy nodes).