No icon

MetaFlow: a Scalable Metadata Lookup Service for Distributed File Systems in Data Centers

MetaFlow: a Scalable Metadata Lookup Service for Distributed File Systems in Data Centers

Abstract:

In large-scale distributed file systems, efficient metadata operations are critical since most file operations have to interact with metadata servers first. In existing distributed hash table (DHT) based metadata management systems, the lookup service could be a performance bottleneck due to its significant CPU overhead. Our investigations showed that the lookup service could reduce system throughput by up to 70%, and increase system latency by a factor of up to 8 compared to ideal scenarios. In this paper, we present MetaFlow, a scalable metadata lookup service utilizing software-defined networking (SDN) techniques to distribute lookup workload over network components. MetaFlow tackles the lookup bottleneck problem by leveraging B-tree, which is constructed over the physical topology, to manage flow tables for SDN-enabled switches. Therefore, metadata requests can be forwarded to appropriate servers using only switches. Extensive performance evaluations in both simulations and testbed showed that MetaFlow increases system throughput by a factor of up to 3.2, and reduce system latency by a factor of up to 5 compared to DHT-based systems. We also deployed MetaFlow in a distributed file system, and demonstrated significant performance improvement.

Existing System:

Existing approaches, e.g., focused on building overlay-based metadata management systems, which could be centralized or decentralized. These systems usually provides two main operations: lookup and I/O. In particular, the lookup operation aims to locate the desired metadata; and the I/O operation retrieves the metadata itself from the storage server using the address returned by the lookup operation. However, the throughput and latency of these overlay-based systems could be significantly degraded due to the bottleneck created by a large number of lookup operations. Such operations actually compete for CPU cycles with I/O operations, which might lead to reduced system throughput. Meanwhile, it could take a long time to locate a metadata object in overlay-based systems, which might increase the system latency.

Proposed System:

We propose MetaFlow, a new, efficient and fast distributed lookup service for metadata management. Rather than setting up a separate lookup operation, MetaFlow utilizes network components to locate the desired metadata with two techniques: software-definednetworking (SDN) and a B-tree based overlay network. SDN provides the ability for network switches to forward packets based on metadata identifiers. A B-tree based overlay is constructed over all physical switches and servers in the data center to generate and maintain forwarding tables for SDN-enabled switches. In this way, a metadata request can reach its target server directly, without the need for a separate lookup operation to query the destination. As a result, the latency in metadata operations could be reduced; and more available CPU cycles would be dedicated to I/O operations to improve the overall system throughput.

Comment As:

Comment (0)