RAID+ - Deterministic and Balanced Data Distribution for Large Disk Enclosures

Sep 12, 2026

Upload by Melati

Existing RAID partitions large disk enclosures so each RAID group uses dedicated disks, delivering isolation but causing disk under-utilization and slow recovery after failures. RAID+ is proposed as a new construction mechanism that balances both normal I/O and reconstruction workloads across a larger disk pool. Using deterministic addressing based on mutually orthogonal Latin squares, RAID+ builds 3-D templates mapping logical data to uniformly distributed disk blocks, engaging more drives without increasing total read/write volume. Evaluation on a 60-drive enclosure with synthetic and real workloads shows faster recovery and improved normal I/O and multi-tenant throughput.

LanguageEnglish

CategoryPresentations

FormatPDF

Pages16

Reading Time6 min

AccessFree to download

Abstract

1 Introduction

RAID+: Deterministic and Balanced Data Distribution for Large Disk Enclosures Guangyan Zhang and Zican Huang, Tsinghua University; Xiaosong Ma, Qatar Computing Research Institute, HBKU; Songlin Yang, Zhufan Wang, and Weimin Zheng, Tsinghua University [https://www.usenix.org/conference/fast18/presentation/zhang](https://www.usenix.org/conference/fast18/presentation/zhang) This paper is included in the Proceedings of the 16th USENIX Conference on File and Storage Technologies. February 12–15, 2018 • Oakland, CA, USA ISBN 978-1-931971-42-3 Open access to the Proceedings of the 16th USENIX Conference on File and Storage Technologies is sponsored by USENIX . RAID+: Deterministic and Balanced Data Distribution for Large Disk Enclosures Guangyan Zhang􀀃 , Zican Huang􀀃 , Xiaosong Ma†, Songlin Yang􀀃 , Zhufan Wang􀀃 , Weimin Zheng􀀃 􀀃 Tsinghua University, † Qatar Computing Research Institute, HBKU Abstract Existing RAID solutions partition large disk enclosures so that each RAID group uses its own disks exclusively. This achieves good performance isolation across underlying disk groups, at the cost of disk under-utilization and slow RAID reconstruction from disk failures. We propose RAID+, a new RAID construction mechanism that spreads both normal I/O and reconstruction workloads to a larger disk pool in a balanced manner. Unlike systems conducting randomized placement, RAID+ employs deterministic addressing enabled by the mathematical properties of mutually orthogonal Latin squares, based on which it constructs 3-D data templates mapping a logical data volume to uniformly distributed disk blocks across all disks. While the total read/write volume remains unchanged, with or without disk failures, many more disk drives participate in data service and disk reconstruction. Our evaluation with a 60-drive disk enclosure using both synthetic and real-world workloads shows that RAID+ signiﬁcantly speeds up data recovery while delivering better normal I/O performance and higher multi-tenant system throughput. 1 Introduction For the past 30 years, Redundant Array of Inexpensive Disks (RAID) [40] has been used pervasively in servers and shared computing platforms. With paritybased RAID levels (e.g., RAID-5 and RAID-6), users obtain high performance via parallel accesses and reliability via data redundancy. With continued advance in disk capacity and slow improvement in speed, however, RAID rebuild time keeps increasing [13, 54] . For example, a recent NetApp document speciﬁes that a 2TB SATA 7200-RPM disk takes 12.8 hours to rebuild on an idle system [12] . When performed online on a heavily loaded system, rebuild can take dramatically longer. Such slow rebuild brings two consequences. First, it raises the risk of a second failure and consequently data loss. Second, prolonged recovery subjects foreground applications to long periods of I/O performance degradation. Note that highperformance solid-state drives (SSDs) actually exacerbate this problem, as their growing deployment promotes storage system construction using more high-density, low-performance hard disks [43] . One inherent reason for such slow recovery is that, with conventional RAID, each disk drive involved in RAID reconstruction is read or written entirely. Despite the growing width of RAID arrays (with each array typically containing several to around a dozen disks), the recovery time is determined by reading/writing an entire disk. No matter how many disk arrays coexist in a shared/virtual storage system, resources are isolated between underlying RAID arrays. Idle or lightly-loaded disks cannot offer help to peers in other RAID arrays, who might be overwhelmed by high access trafﬁc, RAID recovery, or, in the worst case, both. Many approaches to enhancing the reconstruction performance have been proposed [3, 18, 21, 28, 42, 52, 55, 56], which fall into three categories: 1) designing better data layout in a disk group [18, 48, 52, 56], 2) optimizing

FAQs

What problem does RAID+ aim to solve in large disk enclosures?

Conventional RAID isolates disks per RAID group, which leads to disk under-utilization and slow reconstruction because each involved disk must be fully read or written, delaying recovery under load.

How does RAID+ spread workloads during normal I/O and failure reconstruction?

RAID+ deterministically addresses data using properties of mutually orthogonal Latin squares to build 3-D data templates that map logical data to uniformly distributed disk blocks across a larger pool.

What benefits are reported for RAID+ compared with conventional RAID?

With the same total read/write volume, RAID+ enlists more drives for data service and reconstruction, delivering faster recovery, better normal I/O performance, and higher multi-tenant throughput.

Sep 12, 2026

Upload by Melati