Comparison of single and double channel I/O-configurations (original) (raw)
Related papers
2000
Enterprise storage systems depend on disk arrays for their capacity and availability needs. To design and main- tain storage systems that efficiently satisfy evolving require- ments, it is critical to be able to evaluate configuration alternatives without having to physically implement them. In this paper, we describe an analytical model to predict disk array throughput, based on a hierarchical decompo-
A Stochastic Analysis of Hard Disk Drives
International Journal of Stochastic Analysis
We provide a stochastic analysis of hard disk performance, including a closed form solution for the average access time of a memory request. The model we use covers a wide range of types and applications of disks, and in particular it captures modern innovations like zone bit recording. The derivation is based on an analytical technique we call "shuffling", which greatly simplifies the analysis relative to previous work and provides a simple, easy-to-use formula for the average access time. Our analysis can predict performance of single disks for a wide range of disk types and workloads. Furthermore, it can predict the performance benefits of several optimizations, including short-stroking and mirroring, which are common in disk arrays.
Measuring and modelling disk I/O subsystems
1989
The large and still increasing gap between the instruction execution time at the processor and the disk access time started new interest in faster I/O subsystems, that use new architectures. Research efforts into reducing the access time gap can be classified into two categories: efforts to achieve faster physical 110s, by the use of faster disk actuators and faster controllers, and efforts to avoid physical 110, by the use of buffering infor mation in semiconductor memory at some point between the processor and the disks. Both categories will be covered in this work. To develop I/O subsystem implementations that provide the best service possible, a detailed understanding of the current I/O subsystem operation and I/O workload is required. For large mainframe systems a considerable number of publications exist that deal with some modelling aspects of these systems, and their I/O equipment. Surpri singly, little research is reported in the I/O workload characterization area. The main objectives of this research project were to provide more insight in the I/O workload parameters, to design a tool that could be used to measure workload parameters, to build a model describing current I/O subsystem implementations, and to evaluate possible future implementations. In addition to the implementations offered by the hardware manufacturers, new controller schemes are proposed. Introduction § 1 2 Measuring and Modelling Disk I/O Subsystems In this work the emphasis will be on the analysis of the two hardware layers: the mechanical disks and their controllers. The analysis of disks and controllers, however, is only possible when an adequate description of the load imposed by the access methods and the operating system is available. The load description is required to estimate both disk and controller performance parameters. The combination of a load and hardware description is used for modelling studies. For example, the time required to retrieve data from a disk depends on the amount of data: large data areas take more time. The delays that occur because controllers are shared between disks depend on the load of the controllers. The fraction of I/O requests that can be served from a buffer in the controller depends on the order in which I/Os are issued to a disk. Such characteristics of the I/O workload are hard to obtain from standard measurement facilities, that usually only collect the service time and I/O rate for disks. Unfortunately, these are essential for a proper analysis. § 1 Introduction Chapter 6 extends the modelling method to other I/O subsystem implementations. The chapter describes various alternative I/O subsystem implementations, that generally give a better performance than the systems described in chapter 5. The performance improvements are obtained by the use of various forms of buffering, and even by the § 1 Introduction The average seek lime is defined as the lime required to cross one-third of the maximum number of cylinders on the disk. For actual workloads, the average seek lime tends to be less.
Heavy-tailed distribution of parallel I/O system response time
Proceedings of the 10th Parallel Data Storage Workshop
Estimating I/O time of applications is critical for computing system research and developments, such as performance tuning and job scheduling. Parallel I/O systems on large-scale HPC systems typically use several I/O servers attached to a number of hard disk drives to read and write data concurrently. As a result, the response time of individual I/O servers affects the overall I/O performance and modeling the response time distribution holds the key to estimate I/O time. Existing studies have generally considered that the response time follows a Uniform or a Normal distribution. However, none of these studies considered supercomputing environments that are actively used by a number of users to verify the existence of Uniform or Normal distributions. In this study, we collected ≈ 2,500,000 measurements on two peta-scale class supercomputers that are actively used by ≈5000 users. These two systems, Hopper and Edison at the National Energy Research Scientific Computing Center (NERSC), typically support hundreds of concurrent jobs. Our performance measurements include the overheads introduced by the entire parallel I/O stack (I/O library, network, parallel file system software, cache and hardware). Our study shows that the response time of parallel I/O system follows a heavy-tailed property, in contrary to the widely accepted Normal or Uniform distributions. In exploring for new models, we identify that a mix of Power Law and Normal distributions is a good fit for the response time of parallel I/O systems that are actively used by hundreds of jobs concurrently.
Statistical link analysis of high-speed memory I/O interfaces during simultaneous switching events
2008 IEEE-EPEP Electrical Performance of Electronic Packaging, 2008
Single-ended signaling systems, popular in memory I/O interfaces, are limited by signal and power integrity issues such as crosstalk and simultaneous switching output noise (SSO). At high data rates, the single-ended systems also suffers from random noise and timing jitter. In this paper, we present an integrated signal and power integrity simulation flow that combines statistical and transient simulation methods to enable the characterization of single-ended systems to account for random timing jitter in addition to the traditional SI issues focused on the deterministic noise such as intersymbol interference (lSI), crosstalk, and SSO noise. To include SSO noise, we co-simulate power distribution network (PDN) and channel models and treat SSO noise as another form of crosstalk. To capture any system nonlinearity, we employ time-domain based multi-edge response (MER) method to characterize the deterministic and passive portion of channels. Then, random noise and timing jitter impact are included via statistical approach. We use GDDR system to demonstrate our simulation flow.
Simulation and modelling of RAID 0 system performance
2008
ABSTRACT RAID systems are fundamental components of modern storage infrastructures. It is therefore important to model their performance effectively. This paper describes a simulation model which predicts the cumulative distribution function of I/O request response time in a RAID 0 system consisting of homogeneous zoned disk drives. The model is constructed in a bottom-up manner, starting by abstracting a single disk drive as an M/G/1 queue.
Design and performance analysis of a disk array system
IEEE Transactions on Computers, 1995
We concentrate on the architectural issues of parallelizing UO access in a disk array system by means of definition of a new, particularly flexible architecture, called partial dynamic declustering, which is fault-tolerant and offers higher levels of performance and reliability than the solutions normally used. A simulation analysis highlights the efficiency of the proposed solution in balancing the file system workload and demonstrates its validity in both cases of unbalanced loads and expansion of the system. Particular attention is also paid to the definition of analytical models, based on stochastic reward nets, in order to analyze the performance and reliability of the system. The response time distribution function is evaluated and a specific performance analysis with varying degrees of declustering and workload is carried out. Index Terms-Parallel UO, disk array system, declustering organization, stochastic reward nets, response time distribution.
Characteristics of the MasPar parallel I/O system
Proceedings Frontiers '95. The Fifth Symposium on the Frontiers of Massively Parallel Computation, 1994
Inputloutput speed continues to present a performance challenge for high-performance computing systems. This is because technology improves processor speed, memory speed and capacity, and disk capacity at a much higher-rate than mass storage latency. Developments in 1 1 0 architecture have been attempting to reduce this performance gap. The MasPar 110 architecture includes many interesting features. This work presents an experimental study of the dynamic characteristics of the MasPar parallel 110. Performance measurements were collected and compared for the MasPar MP-1 and MP-2 testbeds at NASA GSFC. The results have revealed strengths as well as areas for potential improvements and are helpful to software developers, systems managers, and system designers.