Exploiting regularities in Web traffic patterns for cache replacement

In this paper, a two-unit system with failure interactions is studied. The system is subject to two types of shocks (I and II) and the probabilities of these two shock types are age-dependent. A type I shock, rectified by a minimal repair, causes a minor failure of unit A and type II shock causes a complete system failure that calls for a replacement. Each unit A minor failure also results in an amount of damage to unit B. The damages to unit B caused by type I shocks can be accumulated to trigger a preventive replacement or a corrective replacement action. Besides, unit B with cumulative damage of level z may become minor failed with probability \(\pi (z)\) at each unit A minor failure and rectified by a minimal repair. We consider a more general replacement policy. Under this policy, the system is preventively replaced at the Nth type I shock, or at the time when the total damage to unit B exceeds a pre-specified level Z (but less than a failure level K where \(K>Z\)) or is replaced correctively at first type II shock or when the total damage to unit B exceeding a failure level K, whichever occurs first. To minimize the expected cost per unit time, the optimal replacement policy \((N^{*}\), \(Z^{*})\) is derived analytically and determined numerically. We also show that several previous maintenance models in the literature are special cases of our model.

This paper investigates the effects of unsuccessful repairs on the optimal periodic replacement policy for a repairable product with multiple attempts of minimal repair under the free-repair warranty. The motivation for this study is based on the novel notion of ‘multi-attempt minimal repair’, first introduced by Cha and Finkelstein (2024a, 2024b), in which they highlighted various practical reasons and examples to illustrate that minimal repair attempts may sometimes fail and need to be repeated until success is achieved. Their repair model and useful properties served as the motivation for this study. For both the warranted and non-warranted products, cost models are developed from the user’s perspective, and the corresponding optimal replacement policies are derived to minimize the long-run expected cost rates. By comparing these optimal policies, we show that the optimal planned time for preventive replacement should be adjusted toward the end of the warranty period, and the effects of unsuccessful minimal repair produce different degrees of impact, respectively, on the warranted and non-warranted product. Finally, detailed numerical examples and computation results illustrate our findings, and some insightful observations about practical and managerial implications are concluded.

Traditionally, file transfer protocol (FTP) servers are major archive providers and users apply the Archie server to locate FTP archives. With the extreme popularity of the WWW, Web servers are now important archive providers and users download archives via HTTP. To reduce HTTP traffic, proxy cache servers are deployed on the Internet. However, we find the hit rate of archives in cache servers is quite low. This study proposes a combination of caching and better searching mechanisms to alleviate the problem. We enable a proxy server to automatically collect WWW and FTP archives from its cache, organize them in the form of an FTP directory, and then offer the directory list to the Archie. Accordingly, users can find archives on WWW and FTP servers through the Archie, and they can directly download archives from the proxy server, thus improving the reuse of cached archives. A system was implemented and operated in a real environment to evaluate the approach and results are discussed.

The increasing diversity of Internet appliances calls for an architecture for performance differentiation on information access. The World Wide Web is the dominant interface for information access today. Web proxy caching is the key performance accelerator in the Web infrastructure. While many research efforts have addressed performance differentiation in the network and on Web servers, providing multiple levels of service in proxy caches has received much less attention. This paper has two main contributions. First, we describe, implement and evaluate an architecture for differentiated content caching services as a key element of the Internet infrastructure. Second, we describe a control-theoretical approach that lays well-understood theoretical foundations for resource management to achieve performance differentiation in proxy caches. We describe our experiences with implementing the differentiated caching services scheme in Squid, a popular proxy cache used by many Internet service providers today. Experimental studies and analyses prove that differentiated caching services provide significant better performance to the premium content classes.

Burn-in is a widely used engineering method which is adopted to eliminate defective items before they are shipped to customers or put into field operation. In order to shorten the burn-in process, burn-in is most often accomplished in an accelerated environment. However, there have been few probabilistic or stochastic models for the burn-in procedures in accelerated environment. In this article, under a new stochastic model for accelerated burn-in procedure, the problems of determining both optimal accelerated burn-in time and optimal replacement policy are considered. Components are burned-in under an accelerated environment, then those surviving the burn-in procedure are put into field operation and they are maintained under a replacement policy. The properties of the optimal accelerated burn-in time and optimal replacement policy are obtained and a numerical example which illustrates the usage of obtained results will be presented.

Structural properties of optimal tool replacement policy in a machining center

In this paper, a cold standby repairable system consisting of two dissimilar components and one repairman is studied. When failures occur, the repair of both component 1 and component 2 are not ‘as good as new’. The consecutive operating times of component 1 after repair constitute a decreasing geometric process, while the repair times of component 1 are independent and identically distributed. For component 2, its failure is rectified by minimal repair, and the repair time is negligible. Component 1 has priority in use when both components are good. The replacement policy N is based on the failure number of component 1. Under policy N, we derive the explicit expression of the long-run average cost rate C(N) as well as the average number of repairs of component 2 before the system replaced. The optimal replacement policy N*, which minimises the long-run average cost rate C(N), is obtained theoretically. If the failure rate r(t) of component 2 is increasing, the existence and uniqueness of the optimal policy N* is also proved. Finally, a numerical example is given to validate the developed theoretical model. Some sensitivity analyses are provided to show the influence of some parameters, such as the costs for replacement and repair, and the parameters of the lifetime and repair time distributions of both components, to the optimal replacement policy N* and corresponding average cost rate C(N*).

Optimal replacement policy with replacement last under cumulative damage models

The conception rate in dairy cows is dependent on a number of cow factors such as days in milk and insemination number. Unfortunately, some of these factors were not accounted for in optimal insemination and replacement decision models. By using wrong estimates of the conception rate, the calculated optimal insemination and replacement policy might differ from the real optimal insemination and replacement policy. The objective of this study was to evaluate different sets of conception rates with an increasing level of accuracy to determine the insemination policy. An existing dynamic program for optimal insemination and replacement was used to compare three different scenarios in the estimation of conception rates, based on the reproductive performance of Dutch dairy cattle: (i) constant conception rate throughout lactation, (ii) conception rate dependent on parity and months in milk, and (iii) conception rate dependent on parity, months in milk and insemination number. The time step of the model was 30.4 days (1 month). The discounted future cash flow of culling a cow at each time step (replace a heifer immediately) was compared with keeping that cow under optimal future decisions. The difference between immediate culling and optimal decisions is defined as the retention pay-off. The insemination value was calculated as the difference between the future cash flow between immediate insemination of a cow and waiting one time-step. The results show that the difference in the insemination values and the optimal time to stop insemination depend on parity, lactation stage and the relative milk yield. In older cows with equal milk yields and at the same months in milk, the insemination value was lower than in younger cows. Within a parity, the insemination value was higher for cows with a higher milk yield. On individual cow level, using more accurate conception rate as input in the optimal insemination and replacement model might reduce miscalculation of the economic consequences for at least of €20-€38 per cow per year. Basing insemination decisions on less accurate input of the probabilities of conception, however, did not have an economic consequence at the herd level. In conclusion, using the appropriate conception rate as input in the optimal insemination and replacement model would increase the precise decision for the optimal time to stop insemination and hence improve the reproductive management efficacy.

Failure of many engineering systems usually results from a gradual and irreversible accumulation of damage, a degradation process. Most degradation processes can be monitored using sensor technology. The resulting degradation signals are usually correlated with the degradation process. A system is considered to have failed once its degradation signal reaches a prespecified failure threshold. This paper considers a replacement problem for components whose degradation process can be monitored using dedicated sensors. First, we present a stochastic degradation modeling framework that characterizes, in real time, the path of a component's degradation signal. These signals are used to predict the evolution of the component's degradation state. Next, we formulate a single-unit replacement problem as a Markov decision process and utilize the real-time signal observations to determine a replacement policy. We focus on exponentially increasing degradation signals and show that the optimal replacement policy for this class of problems is a monotonically nondecreasing control limit policy. Finally, the model is used to determine an optimal replacement policy by utilizing vibration-based degradation signals from a rotating machinery application.

A (τ, T) replacement policy is commonly used to prevent system failure. In previous research, it was assumed that the maintenance time can be ignored since it is negligibly short compared with operation time. However, neglecting maintenance time when optimizing maintenance policies for systems for which the maintenance time is not relatively short can lead to sub-optimal or even incorrect solutions to the problem, resulting in higher maintenance costs. An optimal (τ, T) replacement policy is presented for series systems in which the units deteriorate following different stochastic processes independently. The failure of each unit in the time interval [τ, T) occurs in accordance with a nonhomogeneous Poisson process, the failure is immediately removed by a minimal repair and the system continues operating. The failure of any unit in time interval [τ, T) causes the entire system to fail, which is immediately detected, and corrective maintenance is performed on the whole system. If a failure does not occur during time interval [τ, T), the system receives preventive replacement at T A (τ, T) replacement policy was formulated that considers both preventive and corrective maintenance times and that determines an optimal preventive replacement interval that minimizes the expected long-run cost per unit time. A numerical example illustrates that implementing the proposed (τ, T) replacement policy can effectively reduce maintenance costs.

We look at a number of policies for managing a file system cache in a Web server. Traces from NCSA World Wide Web (WWW) server are used in this study. We study several caching policies for improving the overall performance of such a server and show that request response time can be improved by some of the replacement policies that take size of the request into account. It is shown that least frequently used (LFU) performs well when cache sizes are small (<16 MB) and SpacexAge replacement policy performs consistently better than LRU. We show that previously proposed cache sharing policies improve performance significantly in clustered Web servers.

Video on demand (VOD) is expected to become one of the most important and successful services to be offered on emerging technologies. There is not just an interest in the delivery of digital video for home entertainment purposes, but there are also several educational and commercial benefits from this service. However, for this service to become viable, it is important that the video servers meet the demanding data rates and real time performance requirements imposed by video. Interval caching has been proposed as a cost effective way of improving the throughput of a server, but all previous studies on interval caching have used simulation. With numerous parameters and their complex interplay affecting the performance of interval caching, it is infeasible to consider a full factorial experiment with simulation. The paper presents an analytical model for interval caching on video servers. The model has been extensively validated over a range of client requests, video data and server parameters. In addition, the model has been generalized to accommodate variable lengths of video data stored at the server. Using this model, the impact of different parameters has been studied on the performance of interval caching scheme.

LofoSwitch: An online policy for concerted server and disk power control in content distribution networks

By rapid growth of the Internet users and devices, thenumber of servers also increase simultaneously which causes the exponentialincrement of the internet traffic and static data. Handling these huge amountsof user requests and efficiently responding to them require high bandwidthlinks, powerful servers and robust equipment, which despite the availability ofthese requirements getting the full user satisfaction is extremely difficultand a tough challenge. In order to overcome the mentioned problem, the cacheservers are being used as a suitable solution. The performance of web cacheserver directly depends on its replacement policies. Several cache replacementpolicies have been proposed in literature each having varied hit rate (HR) andbyte hit rate (BHR) performances on different networks. The replacement policyproposed in this paper is a dynamic cache replacement policy which trainsitself utilizing previous network logs and by exploiting the data miningclustering algorithm. Once the training step is completed, the proposed policyutilizes the normalization formulas to score each metric of the enquiriesincluding recency, frequency, size and delay. Simulation results showed thatthe proposed policy has the optimum performance on different networks and itnot only improved the performance of web cache server in term of HR and BHR,but also decreases the data retrieval time (Delay Ratio (DR)) of the cacheservers.