impala-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Antoni Ivanov"<>
Subject Slow Impalad nodes
Date Fri, 07 Apr 2017 16:54:11 GMT

Is impala able to handle if some nodes are slower? For example in Hadoop (Spark/ Mapreduce)
you have the concept of speculation which enables a job to finish in reasonable time even
if node is particular slow. 

For example how will Impala handle those two cases: 
1. Because for some reason (especially if there are other tenants) a particular node has very
high CPU utilization. 
2. If some part of nodes are on slower or separate network. E.g: 10 nodes are in the same
LAN but 5 nodes are a few switches away :)


View raw message