• Batch Jobs
• Take from a few seconds to a few days to complete
• E.g. MapReduce, Spark
• Long-running Services
• Should “never” go down, handle short-lived latency-sensitive
requests
• E.g.Presto/Nesto, Spark Streaming, Web Services
• YARN: Data Operating System
• ResourceManager HA & Recovery
• NodeManager Recovery
• Resource Scheduling(e.g. label-based scheduling)
• Apache Slider & Apache Twill
DISTRIBUTED APPLICATION TYPES
评论2