We consider the allocation of Virtual Arrays (VAs)with different RAID levels in a Heterogeneous Disk Array (HDA). This is expected to provide significant cost savings by consolidating multiple disk arrays into one. Disk load balancing across disk arrays is a by-product of HDA. We provide an example of potential performance benefits expected from HDA. The number of Virtual Disks (VDs) required to materialize a VA is determined by upper bounds on VD bandwidth and space per physical disk drive. Only RAID1 and RAID5 are considered in this paper, but we develop the analysis for other RAID levels. A VA allocation is successful if disks sharing VAs with a failed disk are not overloaded in degraded mode. Single-pass data allocation methods for HDA are evaluated with a synthetic stream of allocation requests. Experimental results show that allocation methods minimizing the maximum disk bandwidth and capacity utilization or their variance across all disks yield the maximum number of allocated VAs. When the disk bandwidth is the bottleneck resource we utilize clustered RAID5 arrays to increase the number of allocated VAs.

background or methods

Queueing Systems - Vol. 1: Theory

Data Allocation in Heterogeneous Disk Arrays

In this paper, we study the problem of admission control in heterogeneous network environments where users randomly arrive to the system over time and finish service after a random period of time. We formulate the problem as a dynamic and stochastic assignment problem and find optimal control policies using the technique of semi-Markov decision processes. The performance of optimal control policies is compared to the performance of a greedy strategy. Furthermore, we analytically study some structural properties of optimal policies for systems with ordered resource cost values.

methods

Optimal dynamic admission control in heterogeneous networks

This paper studies the impact of long-range-dependent (LRD) traffic on the performance of reassembly and multiplexing queueing. A queueing model characterizing the general reassembly and multiplexing operations performed in packet networks is developed and analyzed. The buffer overflow probabilities for both reassembly and multiplexing queues are derived by extending renewal analysis and Benes fluid queue analysis, respectively. Tight upper and lower bounds of the frame loss probabilities are also analyzed and obtained. Our analysis is not based on existing asymptotic methods, and it provides new insights regarding the practical impact of LRD traffic. For the reassembly queue, the results show that LRD traffic and conventional Markov traffic yield similar queueing behavior. For the multiplexing queue, the results show that the LRD traffic has a significant impact on the buffer requirement when the target loss probability is small, including for practical ranges of buffer size or maximum delay.

Performance Analysis of Reassembly and Multiplexing Queueing with Long-Range-Dependent Input Traffic

A conventional ALOHA satellite link uses a transponder which blindly echoes all up-channel traffic on the down-channel. An ALOHA channel can never be fully utilized, so an intelligent satellite could statistically multiplex the successful packets from several slotted ALOHA up-channels onto a single down-channel to conserve bandwidth, and hence reduce cost. We refer to this as a concentrated ALOHA system. Throughput, delay and stability effects are considered, varying the number of up-channels per down-channel and the satellite buffer size. Up- and down-channel bandwidths are assigned independent linear costs, and all performance comparisons are between constant cost systems. It is shown that the marginal increase in system performance drops off so quickly that a small number of up-channels maximizes throughput if up-channel bandwidth has a non-zero cost. This small number is a function of the buffer size and the relative cost of up- to down-channel bandwidth. It is also shown that, even if satellite buffer space is free, a small buffer minimizes average delay for some previously studied protocols of this type. A new protocol which improves performance and allows a large buffer to be used effectively is introduced and analyzed. Solving for throughput and delay in concentrated ALOHA systems provides new analytic and numeric results for the G/D/I queue with rest period equal to the service time.

/pdf/analysis-of-concentrated-aloha-satellite-links-21kqr4mev7.pdf

Analysis of concentrated ALOHA satellite links

This paper presents an approach utilizing deterministic and stochastic Petri nets (DSPN) to analyze on-chip communication. In order to demonstrate the suitability of this approach, the on-chip communication structure of two examples featuring typical system-on-chip (SoC) communication conflicts like competition for common communication resources have been studied. A state-of-the-art heterogeneous digital signal processor (DSP) and a design example with an on-chip bus have been examined. The results show that sufficient modeling accuracy can be achieved with low modeling effort in terms of computation and implementation time.

Queueing Systems - Vol. 1: Theory

Citations

Data Allocation in Heterogeneous Disk Arrays

Optimal dynamic admission control in heterogeneous networks

Performance Analysis of Reassembly and Multiplexing Queueing with Long-Range-Dependent Input Traffic

Analysis of concentrated ALOHA satellite links

A Case Study for the Application of Deterministic and Stochastic Petri Nets in the SoC Communication Domain

Related Papers (5)

Theory, Volume 1, Queueing Systems

Queueing systems, Vol. I: Theory

Fundamentals of queueing theory

Introduction to matrix analytic methods in stochastic modeling

Queueing Systems, Vol. II: Computer Applications.