10.4. Data Reorganization

An important goal for any computer cluster is to achieve a balanced load distribution across all the nodes. Both round-robin and random data placement techniques, described in Section 10.2, distribute data retrievals evenly across all disks drives over time. A new challenge arises when additional nodes are added to a server in order to a) increase the overall storage capacity of the system, or b) increase the aggregate disk bandwidth to support more concurrent streams. If the existing data are not redistributed then the system evolves into a partitioned server where each retrieval request will impose a load on only part of the cluster. One might try to even out the load by storing new media files on only the added nodes ...

Get Streaming Media Server Design now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.