CANFARFeatures

Yellow Sticky Features:

• Deployment of Cloud Scheduler to provisioning system test bed with dynamic configuration querying. (MOU 4)
• Trigger dynamic configuration update of deployed Cloud Scheduler. (HLD 3.4.2.4)
• Continuous update of Cloud Scheduler resources.
• Deployment of Cloud Scheduler to a single production cluster with dynamic configuration querying (Beta1). (MOU 5)
• Deployment of Cloud Scheduler to multiple production clusters with dynamic configuration querying (Beta2). (MOU 6)

• Deployment of a VM to multiple clusters with no load balancing between clusters. (RM)
• Deployment of a VM to multiple clusters with rudimentary load balancing between clusters. (RM)
• Deployment of a VM to multiple clusters with more sophisticated load balancing between clusters. (RM)
• Deployment of a VM to multiple production clusters with performance enhancements to sophisticated load balancing, as required. (MOU 7)

• Utilize rudimentary VM Deployment scheduling algorithm. (MOU 5)
• Utilize more sophisticated VM Deployment scheduling algorithm. (MOU 6)
• Round Robin VM Deployment scheduling algorithm.
• Utilize worst-case VM scheduling algorithm such that there is an even distribution between users of VM slots. (MOU 6)
• Refine VM Deployment scheduling algorithm on multiple production clusters. (MOU 7)
• Clear finished jobs from the system.

• Support VM parameter specification for memory, CPU architecture. (RM)
• Support VM parameter specification for number of cores, scratch space. (RM)

• More sophisticated VM life-cycle management. (RM)
• Refine VM life-cycle management on multiple production clusters. (RM)

• Rudimentary job execution fairness between users. (RM)
• More sophisticated job execution fairness between users – anti-starvation model. (RM)
• Refine job execution fairness between users on multiple production clusters. (RM)

• Detect VM provisioning failure.
• Performance enhancement of VM provisioning.

• Scalability – submit 100 jobs by one user with two VMs, specify memory, cpu, and network, to multiple provisioning system test beds,all jobs start execution.
• Scalability – submit 100 jobs by multiple users with multiple VMs, specify multiple cores, to multiple provisioning system test beds, all jobs start execution.
• Scalability – submit 1000 jobs by multiple users with multiple VMs, to multiple provisioning system test beds, all jobs start execution.

• Reliability – submit 10 jobs by one user with one VM to multiple provisioning system test beds, all jobs run.
• Reliability – submit 100 jobs by one user with two VMs, specify memory, cpu, and network, to multiple provisioning system test beds, all jobs run.
• Reliability – submit 100 jobs by multiple users with multiple VMs, specify multiple cores, to multiple provisioning system test beds, all jobs run.
• Reliability – submit 1000 jobs by multiple users with multiple VMs, to multiple provisioning system test beds, all jobs run.

• Support job priorities – over-ride queueing for high-priority job execution. (HLD 3.4.1.6)

• Get copy of required VM via interface to VCE Manager Business Service. (HLD 3.4.1.6))

• Logging for administration.
• Logging for job execution.

• VM Lifecycle Management – pause VM. (HLD 3.4.1.6)
• VM Lifecycle Management – move VM. (HLD 3.4.1.6) (not likely ever)
• VM Lifecycle Management – reboot VM. (HLD 3.4.1.6)

• Network Core Node VM Startup. (CND 2.2.3)
• Network Core Node VM Shutdown. (CND 2.2.3)
• Network Core Node – communicate authorization token to user VM. (CND 2.2.4)
• Network Core Node – user VM joins CANFAR network. (CND 2.2.6)
• Network Core Node – monitor currently running VMs (CND 2.2.4)

• Maintain a history of job execution (HLD 3.4.1.8)
• Identify CPU resources consumed by user (HLD 3.4.1.8)
• Check properties of queued jobs (priority + ?) (HLD 3.4.1.8)

References

• CND – CANFAR Network Design
• HLD – CANFAR High Level Design
• MOU – CANFAR Grid Infrastructure Statement of Work
• RM – this Road Map

CANFAR Project Plan
Schedule:: 9.30, Phase 2 – Provisioning System Testbed Development
Schedule:: 9.3, Phase 3 – Initial Production System Deployment and Beta 1 Prototype
Schedule:: 9.4, Phase 4 – Multiple Production Cluster Deployment and Beta 2 Prototype
Schedule:: 9.5, Phase 5 – Bug Fixing and performance Improvements