Scale and performance in a distributed file system
ACM Transactions on Computer Systems (TOCS)
Performance Analysis of Mass Storage Service Alternatives for Distributed Systems
IEEE Transactions on Software Engineering
Design and implementation of the Sun network filesystem
Innovations in Internetworking
VISA: Netstation's virtual Internet SCSI adapter
Proceedings of the eighth international conference on Architectural support for programming languages and operating systems
A cost-effective, high-bandwidth storage architecture
Proceedings of the eighth international conference on Architectural support for programming languages and operating systems
A Performance Analysis of the iSCSI Protocol
MSS '03 Proceedings of the 20 th IEEE/11 th NASA Goddard Conference on Mass Storage Systems and Technologies (MSS'03)
MSS '03 Proceedings of the 20 th IEEE/11 th NASA Goddard Conference on Mass Storage Systems and Technologies (MSS'03)
Network-Centric Buffer Cache Organization
ICDCS '05 Proceedings of the 25th IEEE International Conference on Distributed Computing Systems
A Performance Comparison of NFS and iSCSI for IP-Networked Storage
FAST '04 Proceedings of the 3rd USENIX Conference on File and Storage Technologies
Performance of optimized software implementation of the iSCSI protocol
SNAPI '03 Proceedings of the international workshop on Storage network architecture and parallel I/Os
A scalable and high performance software iSCSI implementation
FAST'05 Proceedings of the 4th conference on USENIX Conference on File and Storage Technologies - Volume 4
CA-NFS: A congestion-aware network file system
ACM Transactions on Storage (TOS)
Performance study of iSCSI-based storage subsystems
IEEE Communications Magazine
Hi-index | 0.00 |
With the developing of the architecture of networked storage, a new type of storage servers acting as the data conduits over the network emerge, which is called pass-through servers. A typical example is the NFS servers based on iSCSI whose one end is connected to NFS clients and the other is connected to iSCSI storage device. As a store-and-forward device, the NFS servers experience heavy load, which includes protocols and data copying overhead, so a lot of CPU resource is consumed. In this paper, we build a mathematical model for the flow of data in pass-through servers using queuing theory and put forward a scheme of CPU time distribution. This scheme can allocate time of CPU to the service of iSCSI and NFS reasonably. Consequently, the flow rate of data inside servers is accelerated and the system performance is enhanced. We carry out both simulation experiments and real experiments to prove the conclusions. The results show that, if we properly adjust the CPU time distribution ratio according to different request sizes and different ratios of read/write requests, the system can improve the throughput more than 17% compared to the original one and can greatly reduce the mean response time of the data forwarding tasks.