A reliable and cost-efficient auto-scaling system for web applications using heterogeneous spot instances

Cloud computing has brought a paradigm shift in computing world. Cloud Service Providers (CSP) use several pricing models for the services they offer. These models can be of short-term or long-term requirements for Cloud Service Users (CSU). Primary objective of our research is to minimize the total cost by finding optimal resource requirements to satisfy CSUs' demand. CSPs like Amazon EC2, a major Infrastructure as a Service (IaaS) provider uses pricing scheme like reserved, on-demand and spot instances for its Virtual Machine (VM) distribution. In this paper, we proposed a QoS aware heuristic approach to minimize job completion time and cost spot instances. In the proposed model, a heuristic approach is used for provisioning spot instances and an artificial neural network (ANN) is employed to predict the pricing spot instance, price which in turn enhances the performance and validate the quality of service. Efficacy of the proposed model is tested using Amazon EC2's real price traces and the total cost of CSUs' are compared with different approaches.

Cloud computing can provide users with basic hardware resources, and there are three instance types: reserved instances, on-demand instances and spot instances. The price of spot instance is lower than others on average, but it fluctuates according to market demand and supply. When a user requests a spot instance, he/she needs to give a bid. Only if the bid is not lower than the spot price, user can obtain the right to use this instance. Thus, it is very important and challenging to predict the price of spot instance. To this end, we take the most popular and representative Amazon EC2 as a testbed, and use the price history of its spot instance to predict future price by building a k-Nearest Neighbors (kNN) regression model, which is based on our mathematical description of spot instance price prediction problem. We compare our model with Linear Regression (LR), Support Vector Machine Regression (SVR), Random Forest (RF), Multi-layer Perception Regression (MLPR), gcForest, and the experiments show that our model outperforms the others.

Cloud providers are auctioning their excess capacity using dynamically priced virtual instances. These spot instances provide significant savings compared to on-demand or fixed price instances. The users willing to use these resources are asked to provide a maximum bid price per hour, and the cloud provider runs the instances as long as the market price is below the user's bid price. By using such resources, the users are exposed explicitly to failures, and need to adapt their applications to provide some level of fault tolerance. In this paper, we expose the effect of bidding in the case of virtual HPC clusters composed of spot instances. We describe the interesting effect of uniform versus non-uniform bidding in terms of both the failure rate and the failure model. We propose an initial attempt to deal with the problem of predicting the runtime of a parallel application under various bidding strategies and various system parameters. We describe the relationship between bidding strategies and programming models, and we build a preliminary optimization model that uses real price traces from Amazon Web Services as inputs, as well as instrumented values related to the processing and network capacities of cluster instances on the EC2 services. Our results show preliminary insights into the relationship between non-uniform bidding and application scaling strategies.

Many cost-conscious public cloud workloads (“tenants”) are turning to Amazon EC2’s spot instances because, on average, these instances offer significantly lower prices (up to 10 times lower) than on-demand and reserved instances of comparable advertised resource capacities. To use spot instances effectively, a tenant must carefully weigh the lower costs of these instances against their poorer availability. Toward this, we empirically study four features of EC2 spot instance operation that a cost-conscious tenant may find useful to model. Using extensive evaluation based on historical spot instance data, we show shortcomings in the state-of-the-art modeling of these features that we overcome. As an extension to our prior work, we conduct data analysis on a rich dataset of the latest spot price traces collected from a variety of EC2 spot markets. Our analysis reveals many novel properties of spot instance operation, some of which offer predictive value whereas others do not. Using these insights, we design predictors for our features that offer a balance between computational efficiency (allowing for online resource procurement) and cost efficacy. We explore “case studies” wherein we implement prototypes of dynamic spot instance procurement advised by our predictors for two types of workloads. Compared to the state of the art, our approach achieves (i) comparable cost but much better performance (fewer bid failures) for a latency-sensitive in-memory Memcached cache and (ii) an additional 18% cost savings with comparable (if not better than) performance for a delay-tolerant batch workload.

Amazon utilizes its free capacity by launching a special type of VMs known as spot instances in Amazon Elastic Compute Cloud (EC2). Users can get these instances by placing bids. These instances are low-cost resources with the risk of reliability. There is a trade-off between cost and reliability in the use of spot instances. So, there is a scope of research in the efficient use of spot instances. The prices of spot instances vary with time, so, one can efficiently utilize spot instances by accurate prediction of prices before bidding. In this article, an algorithm has been proposed for short term prediction of spot prices. The algorithm uses probability between price transitions from the observed history and it can predict the price for next epoch. One can get the short-term price prediction for a particular instance type and product type in a specific region. We have compared the proposed algorithm with other existing prediction algorithms.

Consumers can realize significant cost savings by procuring resources from computational spot markets such as Amazon Elastic Compute Cloud (EC2) Spot Instances. They can take advantage of the price differentials across time slots, regions, and instance types to minimize the total cost of running their applications on the cloud. However, Spot markets are inherently volatile and dynamic, as a consequence of which Spot prices change continuously. As such, prospective bidders can benefit from intelligent insights into the Spot market dynamics that can help them make more informed bidding decisions. To enable this, we propose a descriptive statistics approach for the analysis of Amazon EC2 Spot markets to detect typical pricing patterns including the presence of seasonal components, extremes and trends. We use three statistical measures -- the Gini coefficient, the Theil index, and the exponential weighted moving average. We also devise a model for estimating minimum bids such that the Spot instances will run for specified durations with a probability greater than a set value based on different look back periods. Experimental results show that our estimation yields on average a bidding strategy that can reliably secure an instance at least 80% of the time at minimum target guarantee between 50% and 95%.

Infrastructure-as-a-Service (IaaS) cloud providers offer many elasticities and flexibilities for users to run their systems in the cloud. The monetary cost issues of running those systems in the cloud are hardly ignored and there is less work discussing improving the monetary efficiency of running large scale systems in dynamic cloud environments. In this paper, we focus on improving the monetary efficiency of running Hadoop systems in the dynamic public cloud. In particular, we carry out detailed study on improving the monetary efficiency by leveraging spot instances. From a cloud broker's perspective, we propose a price-aware virtual machine auto-scaling with migration algorithm to improve the monetary efficiency of running Hadoop in the cloud using spot instances. We evaluate our proposed algorithm through simulation using Amazon EC2 spot price traces and real world workload traces. Compared with other baseline algorithms, our approach can improve the monetary efficiency by up to 9.3x.

This study aims to reduce the cost of using computing services on AWS. Cost reduction is needed because there is a possibility that the total cost of using cloud services exceeds the estimated budget. One type of EC2 that offers a large discount is the Spot Instance. The downside of this type of EC2 is that AWS reserves the right to stop it at any time. The proposed solution is an automation system to select and run EC2 Spot Instance types based on price, discount, amount of memory, and vCPU usage from previous instances. The automation system is built with the implementation of fuzzy inference system and Best-Worst Method (BWM). All input data is obtained using the Boto3 SDK. System deployment is done in Lambda functions. This Lambda function is automatically executed whenever a Spot Instance is terminated by AWS. The EventBridge service will catch the event and then trigger the Lambda to run. System testing was run for 4 (four) days with event simulation using the Send Events feature. From these tests it is known that the automation system can select the appropriate instance and generate a total cost of $3.85 (USD). After calculating the total cost with regular EC2 estimation (On Demand), the cost is reduced by 71.28%. This number proved to be 4.28% greater than previous similar studies.

This study compared the prices of garri in both rural and urban markets of Edo State, Nigeria covering the period 1990 to 2005. Specifically, the study examined the nature and trend of price movement of garri in both markets (rural and urban), determined the annual rate of increase in the price of garri in the markets and tested if the differences in the price of garri in the rural and urban markets were significant. The prices used for the study were derived from the daily prices of garri collected by staff of the Edo State Agricultural Development Programme (EADP, 1990) from the various markets in Edo State, Nigeria. The data were analyzed using simple descriptive statistics (such as means, standard deviations, tables, graphs, etc) and the Student t-test. The study revealed that the price of garri for both markets were cheapest in the third quarter (months of July to September) and most expensive from the 4th to the 1st quarters (from December to around April). Generally, urban market prices for garri were higher than that of the rural markets for the period considered. The observed mean annual price increases (that is, growth rate) was 20.50 and 43.02% for rural and urban markets, respectively. When the differences in price of garri for both markets were tested on a four yearly basis, they were significantly different from each other implying a wide gap between the rural and urban markets occasioned by rising fuel prices over the years. The significant t-values obtained for the period on a four-yearly basis at the 5% level of significance were 1.342, 1.382, 2.250, and 2.321, respectively. Due to the disparity in prices between the rural and urban markets, it is recommended that government should direct her efforts towards stabilizing the price of fuel and create motorable roads in the rural areas where garri is produced in order to close up the wide price margin between the rural and urban markets. Key words: Garri, market prices, comparative analysis, food security.

Typical web applications employ a multi-tier architecture. Traditionally, a pool of physical servers is used to host web applications. To handle the dynamic workloads which characterize today’s web applications, several authors have proposed schemes for dynamic resource provisioning. Such schemes add more servers during peak loads and remove servers during other times. Advances in cloud computing technologies have created new perspectives for real-time dynamic provisioning. The elastic nature of cloud computing systems allows system administrators to quickly scale resources to respond to unexpected load changes. In such systems, dynamic provisioning is not only concerned with meeting Service Level agreements, but also must take into account monetary costs. In this paper, we exploit performance modeling in the context of cloud computing (Amazon EC2). Having such performance models enables understanding the trade-off between performance and cost, a cornerstone in developing dynamic provisioning performance management schemes.KeywordsAmazon EC2MVAperformance modelsweb applicationsmulti-tier

Urban sprawl and home-delivered services: An exploratory analysis of spatial variations of cost and quality in Flanders (Belgium)

Performance is one of the most important business requirements for web based applications. The suitability of better platform for any web based application depends on various factors attributable to processor architecture, memory, local/ global network, input/output characteristics etc. Performance testing is an important part of any distributed or web application testing plan. Inclusion of performance estimates into planning and development cycles ensures that the application delivered to the user satisfies sustained performance even at high load, availability and scalability requirements. In this paper we have thrown light on performance comparison of web applications on single and multi core servers. The tests have been carried out on single processor single core server and also on dual processor quad core Xeon based server. The performance comparison was carried out based on response time, throughput (No. of hits v/s response time) and connection pooling v/s no connection pooling. It was found that performance of Java applications running on the multi-core servers is far better than those running on single core servers. Web applications on quad-core server give better performance, high availability and scalability. Apache JMeter was used to load test the functional behaviour and measure performance based on throughput (No of hits v/s response time) for dynamic querying of data using Java servlets on Tomcat web server. Based on the analysis of performance of web server under varying load of no. of users and threads, it was found that applications on quad core server give more throughput and less response time as compared to single core server. Connection pooling increases the performance of web applications by reusing active database connections instead of creating a new connection with every request. Performance was compared by running the tests using Java servlets for connection pooling and also by acquiring connections directly from the JDBC driver without connection pooling. Based on the test results it was found that in case of quad core server, throughput i.e the number of requests served per second is much higher and reuse of connections from pool of connections showed significant improvement in application performance. This paper describes the results of tests carried out for comparing performance (based on response time) between web applications running on single core server and on quad core server. Tests have also been carried out to analyze advantage of connection pooling over no connection pooling. The test results have been shown graphically.

Service-oriented computing (SOC) promises a world of cooperating services loosely connected, constructing agile Web applications in heterogeneous environments conveniently. Web application interface (API) as an emerging technique attracts more and more enterprises and organizations to publish their deep computing functionalities and big data on the Internet, Web API has become the backbone to promote the development of SOC, thus forming the prosperous Web API economy. However, the number of available Web APIs on the Internet is massive and growing constantly, which causes the Web API overload problem. Quality of service (QoS) as an indicator is able to well differentiate the quality of Web APIs and has been widely applied for high quality Web API selection. Since testing QoS for massive Web APIs is resource-consuming, and the QoS performance depends on contextual information such as network and location, hence accurate QoS prediction has become very crucial for personalized Web API recommendation and high quality Web application construction. To address the above issue, this paper presents a context aware deep factorization machine model (CADFM for short) for accurate Web API QoS prediction. Specifically, we first carry out detailed data analysis using real-world QoS dataset and discover a positive relationship between QoS and contextual information, which motivates us to incorporate beneficial contexts for enhancing QoS prediction accuracy. Then, we treat QoS prediction as a regression problem and propose a context aware CADFM framework that integrates the contextual information via embedding technique. Particularly, we adopt MF and MLP for high-order and nonlinear interaction modeling, so as to learn the complex interaction between users and Web APIs accurately. Finally, the experimental results on real-world QoS dataset demonstrate that CADFM outperforms the classic and the state-of-the-art baselines, thereby generating the most accurate QoS predictions and increasing the revenue of Web APIs recommendation.

The main goal of this work is to show the application of an econometric model called Logistic Regression (LR) in a study of the bidding prices in the Colombian electricity market. The LR model is used to calculate the statistical significance of some variables of the market and the influence of changes in these variables on the estimated bidding price of Generation Companies GENCOs. The methodology is used to fit the bidding prices behavior of a hydraulic agent considering some exogenous variables such as: hydrological conditions, energy demand, spot prices and competitors behavior. The model obtained with LR shows a good agreement with the real data.

Optimum Bid price estimation is crucial for Amazon Elastic Compute Cloud (EC2) consumers if they want to secure uninterrupted access to Spot instances at reduced costs. We recently reported that Bid price estimation is an implicit function of seasonal components and extreme spikes in the Spot price history. In this paper we apply time series forecasting to further substantiate this claim. In particular, we benchmark a number of standard forecasting techniques including Naive, Seasonal Naive, ARIMA, ETS, STL, and TBATS against Spot markets belonging to different market types based on pricing patterns including the presence of seasonal components, extremes, and trends. We run experiments using different look back and forecast horizons, and evaluate the forecasting techniques using three measures, namely Bid Success Rate (BSR), Bid Price Over/Underestimation (BPO/UE), and Root Mean Squared Error (RMSE). Experimental results confirm that successful estimation of Bid prices in EC2 Spot markets is indeed an implicit function of seasonal components and extreme spikes in the Spot price history. Furthermore, our experiments also indicate that for certain types of markets, it is possible to significantly improve BSR by applying a small correction to the estimated Bid price without causing any major disruptions to the market.