Adaptive Overlapped Declustering: A Highly Available Data-Placement Method Balancing Access Load and Space Utilization

Watanabe, Akitsugu; Yokota, Haruo

doi:10.1109/icde.2005.16

Cited by 17 publications

(18 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The adaptive overlapped declustering method proposed in [46] is a good means of reducing the restoration time. As future work, we plan to combine adoptive overlapped declustering with CompIndexCDR.…”

Section: Discussionmentioning

confidence: 99%

A Compound Parallel Btree for High Scalability and Availability on Chained Declustering Parallel Systems

Luo

Watanabe

Yokota

2011

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

SUMMARYScalability and availability are the key features of parallel database systems. To realize scalability, many dynamic load-balancing methods with data placement and parallel index structures on sharednothing parallel infrastructure have been proposed. Data migration with range-partitioned placement using a parallel Btree is one solution. The combination of range partitioning and chained declustered replicas provides high availability (HA) while preserving scalability. However, independent treatment of the primary and backup data in each node requires long failover times. We propose a novel method for the compound treatment of chained declustered replicas using a parallel Btree, termed the FatBtree. In the proposed method, a single Fat-Btree provides access paths to both the primary and backup data of all processor elements (PEs), which greatly reduces failover time. Moreover, these access paths overlap between two neighboring PEs, which enables dynamic load balancing without physical data migration by dynamically redirecting the access paths. In addition, this compound treatment improves memory space utilization to enable index processing with good scalability. Experiments using PostgreSQL on a 160-node PC cluster demonstrate the effectiveness of the high scalability and availability of our proposed method.

show abstract

Section: Discussionmentioning

confidence: 99%

A Compound Parallel Btree for High Scalability and Availability on Chained Declustering Parallel Systems

Luo

Watanabe

Yokota

2011

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

show abstract

“…Request permissions from Permissions@acm.org. kinds of redundant declustering strategies have been proposed: 1) parity-based schemes [1]; and 2) replica copybased schemes [2,3,4,5,6]. As a consequence of the ever-increasing capacity and decreasing price of disk drives, exploiting replica copy-based declustering to increase availability for cluster systems has been widely studied.…”

Section: Introductionmentioning

confidence: 99%

“…This so called data accessing skew may severely decrease overall system performance when some servers receive relatively larger proportion of accessing workload and be overwhelmed. And the proposed replicated declustering schemes [3,4,5,6] are able to introduce skew-tolerant ability to cluster systems, in which skewed accessing on a primary server can be shared with its replica data on other servers. However, we conclude that these pre-existing schemes introduce a trade-off problem between skew tolerance ability and data reliability.…”

Section: Introductionmentioning

confidence: 99%

A fragmented data-declustering strategy for high skew tolerance and efficient failure recovery

Luo

Dang

Yokota

2014

Proceedings of the 18th International Database Engineering &Amp; Applications Symposium on - IDEAS '14

View full text Add to dashboard Cite

Data declustering is a common technique to improve data I/O performance by retrieving data in parallel from multiple storage nodes. Data-declustering methods with replicated data also increase system availability, reliability and skew tolerance. Current replicated declustering schemes fall into two types. In the first, each storage node has its data fully replicated on one other storage node. In the second, each storage node has its replica data fragmented and distributed over several storage nodes. The different schemes have different trade-offs between skew tolerance and data reliability. In this paper, we introduce Fragmented Chained Declustering (FCD) to provide better load balancing and efficient failed-data restoration, with the same or lower data loss rates than previous solutions. Experimental results show the load-balancing and failure recovery efficiency of FCD.

show abstract

“…Besides balancing load, replicas can improve reliability [7], [8]. In an ideal data placement system, the crash of a physical disk can degrade the service of the whole storage system dramatically.…”

Section: Reliable Replicas and Overheadmentioning

confidence: 99%

“…The latter is complicated and hard to scale. In [8], data object and its replica are placed to neighboring disks to improve reliability. This method can not utilize the parallelism of multiple disks adequately.…”

Section: Related Workmentioning

confidence: 99%

Design and Implementation of a Random Data-Placement System with High Scalability, Reliability and Performance

Liu

Xue

Wang

et al. 2006

Computational Science – ICCS 2006

View full text Add to dashboard Cite

Abstract. As storage system scales to thousands of disks, data distribution, load balance and the support for heterogeneous disks become increasingly important. In this paper, we present a new data-placement method named Weighted Interval Algorithm (WIA) for heterogeneous disks. Through it is not optimal in some circumstances, the difference between WIA and the optimal algorithm is trivial. Combined with replication, WIA can nearly balance access load and space utilization and improve reliability simultaneously. For the first time, we implement a data-placement system with high scalability, reliability and performance. The experimental results show that WIA reduces the average response time by 14.8% and decreases coefficient of relative load from 78.09% to 47.46% while the difference of the ratio of space utilization between disks is not more than 0.79%.

show abstract

Adaptive Overlapped Declustering: A Highly Available Data-Placement Method Balancing Access Load and Space Utilization

Cited by 17 publications

References 11 publications

A Compound Parallel Btree for High Scalability and Availability on Chained Declustering Parallel Systems

A Compound Parallel Btree for High Scalability and Availability on Chained Declustering Parallel Systems

A fragmented data-declustering strategy for high skew tolerance and efficient failure recovery

Design and Implementation of a Random Data-Placement System with High Scalability, Reliability and Performance

Contact Info

Product

Resources

About