ExaPlan: Queueing-Based Data Placement and Provisioning for Large Tiered Storage Systems

Scheduling for efficiency and fairness in systems with redundancy

We consider a simple model of a multiprocessor digital system, with three main ingredients: a load of different transaction types with given arrival statistics, an architecture consisting of a network of loosely coupled nodes with tightly coupled symmetric multiple processors at each node, and a load partitioning or assignment of each step of execution of each transaction type to a given node. In order to quantify this problem, we adopt a Jackson network for ease of analysis. Our goal is to consider tradeoffs between the workload, the architecture, and workload partitioning to maximize performance. We measure performance by two indices, the mean throughput rate of completing each tran saction type, and the mean response time to completely execute each transaction type. We present analytic formulae for the mean response time for a job assuming simple Poisson arrival statistics. To illustrate the many tradeoffs possible, we consider one example in detail. We show that if we fix the total mean instruction rate, the mean response time for the multiprocessor network is always greater than the mean response time for a job submitted to one equivalent fast processor. In the special case of purely random arrivals and equalutilization of all network nodes, the mean response time will be at least N times that of a single equivalent fast processor. Moreover, the optimum assignment of excess processing capacity to each node to minimize mean response time does not obey a simple linear proportionment based on utilization, but rather a square root assignment of excess capacity. This model accounts for randomness in arrival and service patterns. Significant features left out of this analysis are the system overhead in coordinating all the multiprocessor interactions, memory contention, and so forth, which presumably will only further degrade performance. In many applications, other issues, such as flexibility, reliability, cost, and so forth, may play a dominant role in a design decision; here we concentrate solely on the system response time as the performance measure, with a fixed total instruction rate.

We consider finite-capacity Markovian queues with a single hiatus scheme and waiting server. Customers are arriving at a Poisson arrival λ and exponential service distribution, with a mean service rate µ. In which customers join the queue according to the number of customers in the system while the hiatus is in the service-providing process. For the assumed queuing model, steady-state probabilities were derived, and some important performance measures, such as the mean number of customers in the system and mean response time in the system and queue are analysed. The expected expense function is developed and formulated as an optimization problem in order to find the minimum expense. Numerical illustrations are given to show the effect of parameters on the performance measures.

Database systems allow quick creation of performance problems. The goal of database systems is to allow the computer-illiterate to write complex and complete applications. It is the job of the syst...

In this thesis, we present our works on resource management in large scale systems, especially for enterprise cluster and storage systems. Large-scale cluster systems become quite popular among a community of users by offering a variety of resources. Such systems require complex resource management schemes for multi-objective optimizations and should be specific to different system requirements. In addition, burstiness has often been found in enterprise workloads, being a key factor in performance degradation. Therefore, it is an extremely challenging problem of managing heterogeneous resources (e.g., computing, networking and storage) for such a large scale system under bursty conditions while providing performance guarantee and cost efficiency.

Redundancy is an important strategy for reducing response time in multi-server distributed queueing systems. This strategy has been used in a variety of settings, but only recently have researchers begun analytical studies. The idea behind redundancy is that customers can greatly reduce response time by waiting in multiple queues at the same time, thereby experiencing the minimum time across queues. Redundancy has been shown to produce significant response time improvements in applications ranging from organ transplant waitlists to Google’s BigTable service. However, despite the growing body of theoretical and empirical work on the benefits of redundancy, there is little work addressing the questions of how many copies one needs to make to achieve a response time benefit, and the magnitude of the potential gains. In this paper we propose a theoretical model and dispatching policy to evaluate these questions. Our system consists of k servers, each with its own queue. We introduce the Redundancy-d policy, under which each incoming job makes copies at a constant number of servers, d, chosen at random. Under the assumption that a job’s service times are exponential and independent across servers, we derive the first exact expressions for mean response time in Redundancy-d systems with any finite number of servers, as well as expressions for the distribution of response time which are exact as the number of servers approaches infinity. Using our analysis, we show that mean response time decreases as d increases, and that the biggest marginal response time improvement comes from having each job wait in only d = 2 queues. The e-companion is available at https://doi.org/10.1287/opre.2016.1582 .

Over the last few decades several innovative ideas have been explored in the energy storage areas, ranging in size, capacity, design complexity, and targeted applications. Some of them are designed for large scale power system applications, others for smallor medium-scale renewable energy or hybrid power systems, while the others are designed to perform short-term energy storage ride through for critical infrastructure (communication systems, hospitals, military facilities, etc.). Energy storage has become an enabling technology for renewable energy applications, grid integration and enhancing power quality and stability in the power transmission and distribution, having a great potential to improve power grid quality and stability and to provide an alternative to fossil fuel-based energy generation. The major constraints for renewable energy penetration are the availability, intermittency, and variability, which can be addressed through energy storage. The energy storage choice depends on specific usage requirements, often incorporating several energy storage systems in order to increase system reliability, capacity, and supply security. In the electric power system, the renewable energy promise lies in its potential to increase grid efficiency, reliability, or in optimizing power flows and supporting variable power supplies. The parameters used in comparisons of various energy storage technologies include efficiency, energy capacity and density, run time, costs, system's response time, lifetime in years and cycles, self-discharge, and maturity of each energy storage technology. The most common energy storage technologies include compressed air, pumped hydro, batteries, fuel cells, flywheels, and super-capacitors. The last four are suitable for the medium scale applications. The chapter discussed energy storage technologies and gives an up to date comparative summary of their performances. After completing this chapter, the readers are able to understand the role, importance, configurations and topologies of energy storage systems, operation principles, characteristics, performances, and operation of major energy storage systems used in power systems, buildings, and industrial facilities. Another benefit is that readers are able to understand the critical role and necessity of energy storage systems in power and renewable energy systems, the differences between large-, mediumand small-scale energy storage systems, and how a system is selected on specific applications based on system characteristics and performances. Major energy storage technologies discussed in this chapter are compressed air energy storage, pumped hydropower storage systems, batteries, flywheels, hydrogen energy storage, fuel cells, supercapacitors, and superconducting energy storage systems. Thermal energy storage systems are covered in detail in the next chapter. This chapter provides comprehensive reviews of the energy storage technologies and gives an up to date comparative summary of their performances, characteristics, and applications.

: A new wholesale level replenishment model is proposed for managing the Navy's inventories of repairable items. It is a readiness-based model which seeks to determine the depths of items of a weapon system which minimize the system's Mean Supply Response Time subject to budget constraint. The model incorporates both a batch procurement and batch repair of the items. Required inputs to this model are the specified values of each. The model assumes that demand is a Poisson process. The model formulation is presented. The solution procedure, which uses marginal analysis, is described. The budget generation process is also described. The model's performance is illustrated with an example of ten items. The results show that the proposed model provides much better Mean Supply Response Time values than the current Navy model. As an added benefit, it also gives better Supply Material Availability values that the current model. Results are also presented of a study conducted to determine potentially desirable values for the procurement order quantity and repair induction quantity. Finally, the use of a Mean Supply Response Time goal to determine the depths of the items is illustrated.

Locating emergency vehicles with an approximate queuing model and a meta-heuristic solution approach

A real-time, on-line information storage and retrieval system for telephone information service has been designed. The input and output of the system are processed in Chinese via a keyboard and a graphic display unit respectivey. Functions of the system include: sorting, merging, updating, displaying, and printing of data.

An increasingly prevalent technique for improving response time in queueing systems is the use of redundancy. In a system with redundant requests, each job that arrives to the system is copied and dispatched to multiple servers. As soon as the first copy completes service, the job is considered complete, and all remaining copies are deleted. A great deal of empirical work has demonstrated that redundancy can significantly reduce response time in systems ranging from Google's BigTable service to kidney transplant waitlists. We propose a theoretical model of redundancy, the Redundancy-d system, in which each job sends redundant copies to d servers chosen uniformly at random. We derive the first exact expressions for mean response time in Redundancy-d systems with any finite number of servers. We also find asymptotically exact expressions for the distribution of response time as the number of servers approaches infinity.

System response delay has been cited as the single most frustrating aspect of using the Internet and the most worrisomeaspect of Web application design. System response time (SRT) research generally concludes that delay should be eliminated where possible to as little as a few seconds, even though delay reduction is costly. Unfortunately, it is not clear if these conclusions are appropriate outside of the developed world where nearly all of the SRT research has taken place. Cultural effects have been, hence, generally missing from SRT research. The one SRT study to date outside of the developed world did report differences using the theoretical construct of cultural chronism, and this finding could limit the generalizability of SRT research findings from developed countries to many economically developing nations. However, limitations and potential confounds in this single study render those findings tentative. The end of Apartheid in South Africa allowed an opportunity to conduct a longitudinal free simulation experiment that overcomes the critical limitations of this previous research. Subjects were members of historically polychronic and monochronic groups who had been segregated by Apartheid and now live in an integrated society with shared infrastructure and computer access. Results find that members of the historically polychronic group are more accepting of longer delays and are more willing to trade longer delays for improved functionality than are their historically monochronic counterparts. Furthermore, tests find that members of the historically monochromic population that came of age in a desegregated, majority-polychronic culture appear to be polychronic themselves and to differ significantly from the older monochronic generation. Results from this study can be applied to design culturally sensitive applications for users in the developing economies of the world.

Connection and service request scheduling affects the performance of a mobile relay node (MRN) in a relay-assisted long-term evolution network in terms of crucial quality-of-service (QoS) metrics such as the blocking probability and the system's response time. In this paper, a novel Deadline-constrained Connection and service Request Scheduling (DCRS) scheme is proposed with the objective of improving an MRN's performance in terms of the above-mentioned metrics. In this paper, a stochastic queueing model is formulated for the purpose of capturing the MRN's behavioral dynamics and evaluating its performance as it operates under DCRS. Simulations are conducted in order to verify the model's results' validity and accuracy. Moreover, the MRN performance achieved under DCRS is compared to that achieved under the typically adopted first-come-first-service (FCFS) scheme. Results indicate that DCRS remarkably outperforms FCFS.

: The results of our work on a biologically motivated model of reflexive reasoning were used to implement a large scale system for performing rapid reasoning using very large knowledge bases. The system - SHRUTI- cM-5 runs on a 32 node CM-5, and has been tested on a large randomly generated knowledge base contalning over 500,000 items and WordNet, a real-world lexical database. The system's response time is fast enough to support retrieval and inferencing for real-time applications such as speech understanding. We have also made some progress in addressing the catastrophic interference problem.

In this paper we address the problem of building a scalable component-based system by means of dynamic reconfiguration. Specifically, we consider the system response time as the performance metric; we assume that the system components can be dynamically reconfigured to provide a degraded service with lower response time. Each component operating at one of the available quality levels is assigned a utility. Higher quality levels are associated to higher utility. We propose an approach for performance-aware reconfiguration of degradable software systems called PARSY (Performance Aware Reconfiguration of software SYstems). PARSY tunes individual components in order to maximize the system utility with the constraint of keeping the system response time below a pre defined threshold. PARSY uses a closed Queueing Network model to select the components to upgrade or degrade.