50 likes | 69 Views
CalvinFS is a Distributed File System featuring consistent WAN replication and high throughput metadata management. Key benefits include high availability, low latency, linearizable operations, and scalability for billions of files with minimal memory requirements. However, real-world evaluation, handling large files, and background process overhead need further consideration.
E N D
CalvinFS: Consistent WAN Replication and Scalable Metadata Management for Distributed File Systems Scriber- Vibha GoyalDate:- March 03, 2016 Course:- CS 525 University of Illinois at Urbana Champaign
Introduction • Consistent WAN replicated Distributed File System. • Metadata management by high throughput distributed database management system.
Cons Pros • High latency for multiple-file operations. • High latency for updates due to consistent replication over WAN. • System can handle billion of files. • Provides linearizable operations with high availability even in case of datacenter outages. • Low memory requirement for metadata per machine. • High read and write throughput. • No distributed commit protocol required, which reduces latency. • Scheduler is deadlock-free. • Allows concurrent writes on a same file.
Discussion/Comments • Evaluation should have been done on more than 300 machines/real world scenario. • Too strong claim to handle unlimited number of files. • Focus of the work was small files. Is it possible that the benefit of CalvinFS is weakened when the files are larger? • Overhead of background process which is doing the compaction? • Median and percentile numbers are given for latencies (not average).