Sorry for the previous mail with wrong format, I had fixed it.
Cloudera, Tencent, Taiwan University, Xiaomi, ByteDance, JD, Didi, Alibaba
Improvements: like CS global scheduling
Scale: 40,000 hosts last year!!
Flink + Spark + MR + other computations
Not server side federation: client side to route apps to clusters
Largest cluster 10,000 nodes
Version: 3.1.0 (upgraded 1.5 year before)
CS + Global scheduling
Overcommitment with opportunistic containers -> increase utilization
Scheduler activities (share more details in upcoming meetup)
2.6.0 -> 3.1.0 (upgrading in progress)
Migrate FS to CS
TODO: Open umbrella for migration
Hadoop Meetup 8/10 discussion:
Location TBD, most likely will be a one day event
Setting up a review-board, we are glad to have 1 or 2 contract-point
from each company in this board, please nominate to [hidden email] Call for more talks!!!