Windsor W. Hsu

WAN-optimized replication of backup datasets using stream-informed delta compression

Replicating data off site is critical for disaster recovery reasons, but the current approach of transferring tapes is cumbersome and error prone. Replicating across a wide area network (WAN) is a promising alternative, but fast network connections are expensive or impractical in many remote locations, so improved compression is needed to make WAN replication truly practical. We present a new technique for replicating backup datasets across a WAN that not only eliminates duplicate regions of files (deduplication) but also compresses similar regions of files with delta compression, which is available as a feature of EMC Data Domain systems. Our main contribution is an architecture that adds stream-informed delta compression to already existing deduplication systems and eliminates the need for new, persistent indexes. Unlike techniques based on knowing a files version or that use a memory cache, our approach achieves delta compression across all data replicated to a server at any time in the past. From a detailed analysis of datasets and statistics from hundreds of customers using our product, we achieve an additional 2X compression from delta compression beyond deduplication and local compression, which enables customers to replicate data that would otherwise fail to complete within their backup window.

RAIDShield: Characterizing, Monitoring, and Proactively Protecting Against Disk Failures

Modern storage systems orchestrate a group of disks to achieve their performance and reliability goals. Even though such systems are designed to withstand the failure of individual disks, failure of multiple disks poses a unique set of challenges. We empirically investigate disk failure data from a large number of production systems, specifically focusing on the impact of disk failures on RAID storage systems. Our data covers about one million SATA disks from six disk models for periods up to 5 years. We show how observed disk failures weaken the protection provided by RAID. The count of reallocated sectors correlates strongly with impending failures. With these findings we designed RAIDS hield , which consists of two components. First, we have built and evaluated an active defense mechanism that monitors the health of each disk and replaces those that are predicted to fail imminently. This proactive protection has been incorporated into our product and is observed to eliminate 88p of triple disk errors, which are 80p of all RAID failures. Second, we have designed and simulated a method of using the joint failure probability to quantify and predict how likely a RAID group is to face multiple simultaneous disk failures, which can identify disks that collectively represent a risk of failure even when no individual disk is flagged in isolation. We find in simulation that RAID-level analysis can effectively identify most vulnerable RAID-6 systems, improving the coverage to 98p of triple errors. We conclude with discussions of operational considerations in deploying RAIDS hield more broadly and new directions in the analysis of disk errors. One interesting approach is to combine multiple metrics, allowing the values of different indicators to be used for predictions. Using newer field data that reports an additional metric, medium errors, we find that the relative efficacy of reallocated sectors and medium errors varies across disk models, offering an additional way to predict failures.