SLAOrchestrator: reducing the cost of performance SLAs for cloud data analytics

Jennifer Ortíz, Brendan Lee, Magdalena Bałazińska, Johannes E. Gehrke, Joseph L. Hellerstein · USENIX Annual Technical Conference · 2018

SLAOrchestrator is a new system designed to reduce the price increases necessary to support performance SLAs in cloud analytics systems. SLAOrchestrator is designed for SLAs that guarantee per-query execution times. Its core architecture consists of a double learning loop that improves both SLAs and resource management over time. It further utilizes an efficient combination of elastic query scheduling and multi-tenant resource provisioning algorithms to reduce the costs of performance guarantees.

Read the paper · More papers on PaperTik