The I/O routine in network storage has longer transferring time and overhead than local I/O. To well make use of the storage network bandwidth and meet demand of concurrent I/O requests, we need to find the appropriat...
详细信息
ISBN:
(纸本)0769526519
The I/O routine in network storage has longer transferring time and overhead than local I/O. To well make use of the storage network bandwidth and meet demand of concurrent I/O requests, we need to find the appropriate schedule policy for I/O requests at the level of the network. The focus in the paper is on how to schedule concurrent I/O data transferring in share storage network, and to improve performance such as the response time affected by scheduling method. Our objective is to analysis the response time of concurrent requests as function their schedule order such as the sequenced, parallel and hybrid policies. We then verify our results by both the analytical method and experiment.
Object replication is an important and sophisticated question, it can reduce access latency and bandwidth consumption, replication can also help in load balancing and can improve reliability by creating multiple repli...
详细信息
storagesystems represent a growing market. In recent years there has been an explosion of applications that use enormous amounts of data. Disk arrays are today's standard solution for enterprise class storage sys...
详细信息
storagesystems represent a growing market. In recent years there has been an explosion of applications that use enormous amounts of data. Disk arrays are today's standard solution for enterprise class storagesystems. One of the most crucial parameters in designing a disk array is the size of the striping unit. The paper presents and validates a performance model for RAID level 5. The model is created as a closed fork-join queue network (CFJQN) under synchronous I/O workloads. It can be used to compute the optimal strip unit size for disk arrays
This paper researches a programmable signal device by an improved direct digital frequency synthesis technology on DSP and FPGA to simulate servo signal for read-write channel of micro hard disk. This device has wide ...
详细信息
This paper researches a programmable signal device by an improved direct digital frequency synthesis technology on DSP and FPGA to simulate servo signal for read-write channel of micro hard disk. This device has wide frequency range from 1 Hz to 10MHz, high frequency resolution of 0.001 Hz, high precision of frequency, phase and amplitude of output wave, good reliability, little distortion. Compared with general devices, this device can generate all kinds of servo signal, and circuit is simply, reliable, and programmable
Within high performance backup/restore environment there are cases that require the retrieval of massive data with huge size from tape storage to disks. The data retrieval time affects the system utilization or the sy...
详细信息
Within high performance backup/restore environment there are cases that require the retrieval of massive data with huge size from tape storage to disks. The data retrieval time affects the system utilization or the system recovering speed. Some schemes were/will be proposed to place data across tape drives for minimized data retrieval latency. SimTape, a Java-based tape library simulator, is built to compare above mentioned allocation schemes which minimize data seek time and tape switch time respectively. This paper describes the development and application of SimTape's software simulation model, also, for the design and optimization of the record, readout and data recovery processes common to all existing tape library processes. The simulation model provides an accurate statistical report from which each scenario can be evaluated
This paper researches a programmable signal generator by an high integrated direct digital frequency synthesis (DDFS) technology on FPGA to simulate servo signal for read-write channel of micro hard disk. This device ...
详细信息
This paper researches a programmable signal generator by an high integrated direct digital frequency synthesis (DDFS) technology on FPGA to simulate servo signal for read-write channel of micro hard disk. This device use very large-scale integrated circuit FPGA (field-programmable logic) to integrate PDSP (programmable digital signal processor) for look-up table and integrate DDFS design, use 16 bit high resolution DAC (digital to analog converter) .This device has high precision of frequency, phase and amplitude of output wave, good reliability, little distortion. Compared with general devices, this device can generate all kinds of servo signal, and circuit is simply, reliable, and programmable.
Referring to the concept of "distributed virtual disks" in cluster file system, this paper proposes a new storage architecture for mid/large-scale RAID-ESS, which consists of several different-structure SubR...
详细信息
Referring to the concept of "distributed virtual disks" in cluster file system, this paper proposes a new storage architecture for mid/large-scale RAID-ESS, which consists of several different-structure SubRAIDs. Clients data can migrate into optimal SubRAID dynamically according to their access characteristics; the number of SubRAIDs and the structure of each SubRAID can also be adjusted to the evolution of system load, thus the whole ESS system can always provide a virtual disk for each client to satisfy its I/O need. Simulation results demonstrate significant I/O performance advantage in ESS, compared with the traditional mid/large-scale RAID5 or RAID1 of the same storage space.
With computer application popularizing and storage capacity increasing, research for NAS and SAN storagesystem are gradually turning to OBS. The OBS will become the next wave of network storage technology. In this pa...
详细信息
With computer application popularizing and storage capacity increasing, research for NAS and SAN storagesystem are gradually turning to OBS. The OBS will become the next wave of network storage technology. In this paper a framework for accessing general object storage (GOS) is presented. The GOS system adopts a basic unit called storage device object (SDO) and proposes an access model that contains upper layer, middle layer and lower layer. It not only enriches the concept of object, but also gives a compatible object interface in the network storagesystems. So it allows applications to access to different SDO according their attributes. Therefore it is more universal
Huge object-based storagesystem (HOSS), using objects which possess their own attributes and methods as storage interface, absorbs the advantages of both blocks and files interface. HOSS adopts the interaction among ...
详细信息
Huge object-based storagesystem (HOSS), using objects which possess their own attributes and methods as storage interface, absorbs the advantages of both blocks and files interface. HOSS adopts the interaction among clients, metadata servers and object-based storage device, which has decreased the overload of servers in large measure. After the interactions' flows concerned with I/O operations among the three components described in detail the design and implementation of parallel I/O scheduling is described and discussed. With the buffer, the results show that it enhances the whole system performance efficiently. The performance is better when it's compared with the Lustre system
暂无评论