Tag Archives: msst13

A Deduplication Study for Host-side Caches in Virtualized Data Center Environments

msst.gifJingxin Feng and Jiri Schindler.

This paper explores the effectiveness of content deduplication in large (typically hundreds of GB) flash memory-based caches inside VM hypervisors.

Flash memory-based caches inside VM hypervisors can reduce I/O latencies and offload much of the I/O traffic from network-attached storage systems deployed in virtualized data centers. This paper explores the effectiveness of content deduplication in these large (typically 100s of GB) host-side caches. Previous deduplication studies focused on data mostly at rest in backup and archive applications. This study focuses on cached data and dynamic workloads within the shared VM infrastructure. We analyze I/O traces from six virtual desktop infrastructure (VDI) I/O storms and two long-term CIFS studies and show that deduplication can reduce the data footprint inside host-side caches by as much as 67%. This in turn allows for caching a larger portion of the data set and improves the effective cache hit rate. More importantly, such increased caching efficiency can alleviate load from networked storage systems during I/O storms when most VM instances perform the same operation such as virus scans, OS patch installs, and reboots.

In Proceedings of the IEEE Symposium on Massive Storage Systems and Technologies 2013 (MSST ’13).

Resources

  • The author’s version of this paper is attached to this posting. Please observe the following copyright:

© 2013 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

Host-Side-Cache-Deduplication.pdf

Paragone: What’s next in block I/O trace modeling

msst.gifRukma Talwadker and Kaladhar Voruganti.

In this paper, we present Paragone, an algorithm that helps designers fundamentally rethink how I/O traces should be modeled and replayed.

Designers of storage and file systems use I/O traces to emulate application workloads while designing new algorithms and for testing bug fixes. However, since traces are large, they are hard to store and moreover inflexible to manipulate. Thus, researchers have proposed techniques to create trace models in order to alleviate these concerns. However, the prior trace modeling approaches are limited with respect to 1) number of trace parameters they can model, and hence, the accuracy of the model and 2) with respect to manipulating the trace model in both temporal and spatial domains (that is, changing the burstiness of a workload, or scaling the size of the data supporting the workload). In this paper we present a new algorithm/tool called Paragone that addresses the above mentioned problems by fundamentally re-thinking how traces should be modeled and replayed.

In Proceedings of the IEEE Symposium on Massive Storage Systems and Technologies 2013 (MSST ’13).

Resources

  • The author’s version of this paper is attached to this posting. Please observe the following copyright:

© 2013 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

Paragone.pdf