Takaki Nakamura

REC2: Restoration Method Using Combination of Replication and Erasure Coding

When large-scale disasters occur, such as the Great East Earthquake Japan occurred in March 2011, servers located in the affected area are sometimes disrupted. If the servers has important information, such as medical information, it is crucial to restore the servers because the information is utilized by medical activities. Therefore, the servers must be restored as soon as possible. We propose a restoration method using the combination of replication method and erasure coding method, called REstoration method using Combination of Replication and Erasure Coding (REC2.) REC2 applies the replication method to the meta-data of a file, and it also applies the erasure coding method to the user-data of the file. The combination achieves the both the high restoration throughput with the amount of the backup data reduced. In this paper, we describe the implementation details of REC2 and evaluation results. In the evaluation, we compare REC2 with the conventional restoration method about the amount of backup data and the restoration throughput. From the results, we found out that REC2 provides higher restoration throughput than the conventional restoration method while REC2 reduces the amount of backup data.

A Guideline for Data Placement in Heterogeneous Distributed Storage Systems.

We propose a guideline for data placement in heterogeneous distributed storage systems. Heterogeneous distributed storage systems sometimes cause degradation of aggregate data throughput. The proposed guideline is that data accessed by a client should be placed equally to all servers. We evaluate a storage system configured by the proposed guideline by using sysstat and compare read/write data throughput between typical data placement and proposed data placement. Then we conclude that the proposed guideline improves the rate of aggregate data throughput while increasing the number of access streams.

Redundancy-based iterative method to select multiple safe replication sites for risk-aware data replication: REDUNDANCY-BASED ITERATIVE METHOD

This paper presents a method to solve the ‘replication site decision problem’ (RSDP) in a short computation time in the case of multiple replicas. RSDP is a problem of finding which combination with pairs of primary-replication sites is the safest when an assumed disaster such as an earthquake affects hundreds or thousands of sites. The existing representation of RSDP is solvable, but it frequently takes much computation time to seek an optimal solution because numerous replicas cause a rapid increase in the number of primary-replication site combinations. The proposed heuristic method, derived from redundancy-based problem partitioning and iterative parameter update techniques, reduces the number of combinations at the slight cost of data availability in the disaster-affected area. Computation time evaluation shows that the proposed method with two or three replicas costs at most twice or thrice, respectively, as much time as that of the original RSDP with one replica, independently of the number of sites. However, the original RSDP with two replicas costs 5 times as much time as that of the original RSDP at 10 sites and 3036 times at 80 sites. Moreover, the data availability cost of the proposed method is only 0.1%.

Redundancy‐based iterative method to select multiple safe replication sites for risk‐aware data replication

This paper presents a method to solve the ‘replication site decision problem’ (RSDP) in a short computation time in the case of multiple replicas. RSDP is a problem of finding which combination with pairs of primary-replication sites is the safest when an assumed disaster such as an earthquake affects hundreds or thousands of sites. The existing representation of RSDP is solvable, but it frequently takes much computation time to seek an optimal solution because numerous replicas cause a rapid increase in the number of primary-replication site combinations. The proposed heuristic method, derived from redundancy-based problem partitioning and iterative parameter update techniques, reduces the number of combinations at the slight cost of data availability in the disaster-affected area. Computation time evaluation shows that the proposed method with two or three replicas costs at most twice or thrice, respectively, as much time as that of the original RSDP with one replica, independently of the number of sites. However, the original RSDP with two replicas costs 5 times as much time as that of the original RSDP at 10 sites and 3036 times at 80 sites. Moreover, the data availability cost of the proposed method is only 0.1%.

A Method for Eliminating Metadata Cache Deallocation Latency in Enterprise File Servers

This paper presents an analysis of a performance bottleneck in enterprise file servers using Linux and proposes a modification to this operation system for avoiding the bottleneck. The analysis shows that metadata cache deallocation of current Linux causes large latency in file-request processing when the operational throughput of a file server becomes large. To eliminate the latency caused by metadata cache deallocation, a new method, called “split reclaim,” which divides metadata cache deallocation from conventional cache deallocation, is proposed. It is experimentally shown that the split-reclaim method reduces the worst response time by more than 95% and achieves three times higher throughput under a metadata-intensive workload. The split-reclaim method also reduces latency caused by cache deallocation under a general file-server workload by more than 99%. These results indicate that the split-reclaim method can eliminate metadata cache deallocation latency and make possible the use of commodity servers as enterprise file servers.