BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:2.0
BEGIN:VEVENT
DTSTART:20141118T203000Z
DTEND:20141118T210000Z
LOCATION:393-94-95
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: The growing size of modern storage systems is expected to soon achieve and exceed billions of objects, making metadata operation critical to the overall performance. Many existing parallel and cluster file systems only focus on providing highly parallel access to file data, but lack a scalable metadata service. In this paper, we introduce a middleware design called IndexFS that adds support to existing file systems such as HDFS and PVFS for high-performance operations on metadata and small files. IndexFS uses a tabular-based architecture that incrementally partitions the namespace at per-directory basis, preserving disk locality for small directories. We also propose two client caching techniques: bulk insertion for creation intensive workloads and stateless metadata caching for hot spot mitigation. By combining these techniques, we scaled IndexFS to 128 servers for various metadata workloads. Experiments demonstrate that its out-of-core metadata throughput out-performs PVFS by 50% to an order of magnitude.
SUMMARY:IndexFS: Scaling File System Metadata Performance with Stateless Caching and Bulk Insertion
PRIORITY:3
END:VEVENT
END:VCALENDAR