Cluster schedulers

Quincy, Mesos, & Borg

What’d you think?

What “is” each work?

What is the general focus of each paper?

What did the papers compare against?

Connections to other papers

“a distinguishing feature of the data-intensive clusters we are interested in is that the computers in the cluster have large disks directly attached to them. … high-performance computing clusters traditionally do not have a large quantity of direct-attached storage” [Quincy]

Connections to other papers

Connections to other papers

Connections to other papers

“Quincy [25] is a fair scheduler for Dryad that uses a centralized scheduling algorithm for Dryad’s DAG-based programming model. In contrast, Mesos provides the lower-level abstraction of resource offers to support multiple cluster computing frameworks.” [Mesos]

Which works could you replicate and why or why not?

Which work has the most interesting core idea?

Which work has the most valuable implementation?

Which work has the most valuable secondary ideas?

Is the Borg scheduler queue-based?

Can you model Mesos as a min-cost flow problem?

Can you model Borg as a min-cost flow problem?

What do these works teach you about simulating cluster behavior?