Building a VM-aware Clustering System to aggregate data from Storage Appliances and manage VM’s protection, relocate policies to make sure efficiency and reliability of IO/Space/Flash.
Transparent Failover: Implemented Synchronous Replication Service Group to realize zero RPO for VMs.
VM Scale-out: Built a service to use predictive algorithms (Monte Carlo, Linear regression, KMeans, LogNormal) to find out space/IOLoad/flash imbalanced appliances, then select and relocate VMs to reach overall Load Balance.
200k VM support: Split Aggregator into small aggregators to reduce lock held time and increase throughput of the message queue. Tuned and refactored thread pool to reduce threads coordination time.
Policy Stickiness: Designed a State Machine to carry over VM protection policies after VM migration.
REST API: Responsible for products’ REST API review, maintenance and version compatibility.
Two-time Tintri Hackathon Winner!