Cloud providers offer serverless, pre-trained ML-based Backend-as-a-Service (BaaS) services, for tasks like translation or speech conversion via single API calls. However, for more complex conversions, users compose multiple BaaS services into serverless workflows to overcome individual limitations. This paper investigates the scalability and interoperability of such composite BaaS services within Sky Computing, focusing on dynamic provider selection between AWS and GCP. We develop two scalable composite BaaS services and propose a novel approach to federate BaaS services, offering two interoperability mechanisms: (1) through-workflow and (2) through-function, enabling seamless integration and dynamic selection of cloud providers. Experiments reveal significant runtime variability in Sky Computing, with extreme outliers of up to$12.9 \times $, making naive averaging strategies unsuitable to determine the optimal configuration. In addition to the well-known cold and warm start effects in FaaS performance models, we identify four additional categories of anomalies that, when FaaS and BaaS are combined, generate extreme outliers far beyond baseline performance instability. Finally, we propose a theoretical framework to explain these behaviors and mitigate their impact, paving the way for reliable and scalable composite BaaS workflows.

Extending cloud computing towards fog and edge computing yields a heterogeneous computing environment known as computing continuum. In recent years, increasing demands for scalable, cost-effective, and streamlined maintenance services have led application and service providers to prefer serverless models over monolithic and serverful processing. However, orchestrating the computing continuum in complex application workflows of serverless functions, each with distinct requirements, introduces new resource management and scheduling challenges. This paper introduces an orchestration service for concurrent serverless workflow processing across the computing continuum called HEFTLess. HEFTLess uses two deployment modes tailored to serve each workflow function: predeployed and undeployed. We formulate the problem as a Binary Integer Linear Programming (BLP) optimization model, incorporating multiple groups of constraints to minimize the overall completion time and monetary cost of executing workflow batches. Inspired by the Heterogeneous Earliest Finish Time (HEFT) algorithm, we propose a lightweight serverless workflow scheduling heuristic to cope with the high optimization time complexity in polynomial time. We evaluate HEFTLess using two machine learning-based serverless workflows on a real computing continuum testbed, including AWS Lambda and 325 combined on-promise and cloud instances from Exoscale, distributed across five geographic locations. The experimental results confirm that HEFTLess outperforms state-of-the-art methods in terms of both workflow batch completion time and cost.

Infrastructure-as-Code (IaC) frameworks empower developers to swiftly define and provision their infrastructure with a single click. However, the domain-specific languages (DSLs) utilized for coding the infrastructure often lean towards provider specificity rather than being application-centric. This results in increased developer effort, as they are compelled to duplicate data when deploying serverless functions across diverse regions and providers within federated FaaS environments. To mitigate this challenge, we introduce CODE, a framework engineered to streamline the deployment of functions in federated FaaS settings. CODE facilitates automatic deployment directly from the storage of any provider, eliminating the need for additional development effort to upload or copy deployment packages between disparate providers. Aligned with the guiding principle of “code once, deploy everywhere”, CODE adopts a three-level hierarchy: function → providers → cloud regions. This architectural approach dramatically reduces the lines of code (LoC) in IaC scripts by up to 9.23× when contrasted with prevailing IaC frameworks such as Terraform and Serverless Framework. Additionally, CODE’s unified storage interface slashes LoC by up to 81.8%, both within CODE itself and when coding functions that use storage from providers such as AWS and GCP. In our comprehensive evaluation, we assessed the correlation between deployment package size and deployment time for various functions within a real-world serverless workflow across four regions of AWS and GCP. Our findings indicate that AWS deployment packages are significantly larger, often in the tens of megabytes, compared to GCP. Despite the larger size, AWS deploys these packages up to 6× faster than GCP.

Serverless is a popular paradigm for expressing compute-intensive applications as serverless workflows. In practice, a significant portion of the computing is typically offloaded to various Backend-as-a-Service (BaaS) cloud services. There cent rise of federated serverless and Sky computing offers cost and performance advantages for these BaaS-enabled serverless workflows. However, due to vendor lock-in and lack of service interoperability, many challenges remain that impact the development, deployment, and scheduling of BaaS-enabled serverless workflows in federated serverless infrastructures. This paper introduces BAASLESS – a novel platform that delivers global and dynamic federated BaaS to serverless work flows. BAASLESS provides: (i) a novel SDK for uniform and dynamic access to federated BaaS services, reducing the complexity associated with the development of BaaS-enabled serverless workflows, (ii) a novel globally-federated serverless BaaS framework that delivers a suite of BaaS -less ML services, including text-to-speech, speech-to-text, translation, and OCR, together with a globally federated storage infrastructure, comprising AWS and Google cloud providers, and (iii) a novel model and an algorithm for scheduling BaaS-enabled serverless workflows to improve their performance. Experimental results using three complementary BaaS-enabled serverless workflows show that BAASLESS improves workflow execution time by up to 2.95× compared to the state-of-the-art serverless schedulers, often at a lower cost.

Most well-known cloud providers offer advanced support for serverless applications that goes beyond single function invocation by enabling developers to build entire workflows, which are known as serverless function choreographies (FCs). Current support for FCs by many FaaS systems uncovered important problems including maximum number of parallel function executions, unexpected considerable delays, and provider lock-in. These limitations can result in longer execution times or even failure to execute individual functions or entire FCs. To overcome some of these limitations, we introduce a scalable middleware service xAFCL that can schedule and execute different functions of the same FC across multiple FaaS systems (currently supporting all top five providers). In order to support scheduling under xAFCL, we introduce a novel FaaS model which estimates the completion time of functions by considering FaaS system limitations, submission delays, and overheads for executing functions. Experimental results demonstrate that xAFCLs FaaS model shows very low inaccuracy of up to 2.9% for AWS and 20% for IBM for real-life BWA data-bound FC that uses S3. Moreover, xAFCL outperforms an earliest start time (EST) scheduler by up to 43% for makespan and 2.7x for throughput.

Many researchers migrate scientific serverless workflows or function choreographies (FCs) on Function-as-a-Service (FaaS) to benefit from its high scalability and elasticity. Unfortunately, the heterogeneity of federated FaaS hampers decisions on appropriate parameter setup to run FCs. Consequently, scientists must choose between accurate but tedious and expensive experiments or simple but cheap and less accurate simulations. Unfortunately, related works support either simulation models for serverfull workflows running on virtual machines and containers or partial FaaS models for individual serverless functions focused on execution time and neglecting various kinds of federated overheads.

Nowadays, scientists migrate workflow applications on serverless Function-as-a-Service (FaaS) platforms in a form of so called function choreographies (FCs) to benefit from FaaS high elasticity and instantly spawning numerous functions. How-ever, the heterogeneous nature of federated FaaS overburdens decisions for the most appropriate configuration setup. Unfor-tunately, related work mainly support either (i) scheduling of serverful workflow applications that run on virtual machines or (ii) container-based algorithms to schedule individual functions on specific container (executor). Either approach is hard to implement for FCs in federated FaaS; the former due to specifics of the FaaS deployment model, while the latter because they are primarily focused on bag of functions and reducing startup latency down to microseconds. Such optimization is negligible for scientific FCs whose functions may run hundreds of seconds due to enormous compute and I/O operations to distributed cloud storage. Instead, scientific FCs would benefit from schedulers that select the appropriate FaaS provider, cloud region, and memory settings. To bridge this gap in scheduling scientific FCs, this paper introduces FaaSt, a novel list-based FC scheduler that optimizes makespan of an FC that runs functions in federated FaaS. The evaluation with three other schedulers showed that FaaSt overcomes limitations of a single FaaS region and generates speedup of up to 2.82× when running FCs across four cloud regions compared to a single region. Moreover, FaaSt achieves speedup of up to 1.74 × compared to the other state-of-the-art FC schedulers across the same four regions.

[J1C2 Presentation Abstract at IEEE SERVICES 2021 for IEEE Transactions on Services Computing DOI 10.1109/TSC.2021.3128137]

Porting existing monoliths to the Function-as-a-Service (FaaS) (FaaSification) can be very challenging for software developers due to different architectural styles. For a successful porting, developers need to resolve various dependencies, such as method invocations of external packages or user-defined codes, as well as global and local variables used in and after the code block that should be faasified. To bridge the gap and automatize FaaSification, this paper introduces M2FaaS, a FaaSifier that automatically converts a Node.js monolith into a hybrid by faasifying annotated code blocks as serverless functions on multiple FaaS providers. M2FaaS is a novel FaaSifier that resolves many challenges for the resulting monolith to work properly after the FaaSification. Developers may annotate all dependencies that need to be resolved for the generated functions to run properly and specify variables that should be returned by the function to the monolith because they are used later in the monolith. Moreover, M2FaaS is the first FaaSifier that faasifies arbitrary code blocks. The current M2FaaS prototype supports FaaSification of individual functions on two FaaS providers, AWS Lambda and IBM Cloud Functions. Finally, M2FaaS introduces an optional annotation for alternative functions to be invoked in case the primary faasified function fails. The resulting hybrid application invokes the automatically deployed serverless functions, while the original code remains executable. Experiments with four complementary monoliths demonstrate that M2FaaS outperforms state-of-the-art FaaSifiers in terms of development effort by up to 73.3%. Moreover, with the fault tolerance support, M2FaaS finishes all submitted functions, thereby achieving by 18.5% higher throughput than the other FaaSifiers.

Cloud applications often benefit from deployment on serverless technology Function-as-a-Service (FaaS), which may instantly spawn numerous functions and charges users for the period when serverless functions are running. Maximum benefit is achieved when functions are orchestrated in a workflow or function choreographies (FCs). However, many provider limitations specific for FaaS, such as maximum concurrency or duration often increase the failure rate, which can severely hamper the execution of entire FCs. Current support for resilience is often limited to function retries or try-catch, which are applicable within the same cloud region only. To overcome these limitations, we introduce rAFCL, a middleware platform that maintains reliability of complex FCs in federated clouds. In order to support resilient FC execution under rAFCL, our model creates an alternative strategy for each function based on the required availability specified by the user. Alternative strategies are not restricted to the same cloud region, but may contain alternative functions across five providers, invoked concurrently in a single alternative plan or executed subsequently in multiple alternative plans. With this approach, rAFCL offers flexibility in terms of cost-performance trade-off. We evaluated rAFCL by running three real-life applications across three cloud providers. Experimental results demonstrated that rAFCL outperforms the resilience of AWS Step Functions, increasing the success rate of entire FC by 53.45%, while invoking only 3.94% more functions with zero wasted function invocations. rAFCL significantly improves availability of entire FCs to almost 1 and survives even after massive failures of alternative functions.

Serverless workflow applications or function choreographies (FCs), which connect serverless functions by data- and control-flow, have gained considerable momentum recently to create more sophisticated applications as part of Function-as-a-Service (FaaS) platforms. Initial experimental analysis of the current support for FCs uncovered important weaknesses, including provider lock-in, and limited support for important data-flow and control-flow constructs. To overcome some of these weaknesses, we introduce the Abstract Function Choreography Language (AFCL) for describing FCs at a high-level of abstraction, which abstracts the function implementations from the developer. AFCL is a YAML-based language that supports a rich set of constructs to express advanced control-flow (e.g. parallelFor loops, parallel sections, dynamic loop iterations counts) and data-flow (e.g multiple input and output parameters of functions, DAG-based data-flow). We introduce data collections which can be distributed to loop iterations and parallel sections that may substantially reduce the delays for function invocations due to reduced data transfers between functions. We also support asynchronous functions to avoid delays due to blocking functions. AFCL supports properties (e.g. expected size of function input data) and constraints (e.g. minimize execution time) for the user to optionally provide hints about the behavior of functions and FCs and to control the optimization by the underlying execution environment. We implemented a prototype AFCL environment that supports AFCL as input language with multiple backends (AWS Lambda and IBM Cloud Functions) thus avoiding provider lock-in which is a common problem in serverless computing. We created two realistic FCs from two different domains and encoded them with AWS Step Functions, IBM Composer and AFCL. Experimental results demonstrate that our current implementation of the AFCL environment substantially outperforms AWS Step Functions and IBM Composer in terms of development effort, economic costs, and makespan.

Many researchers rely on simulations to analyze and validate their researched methods on Cloud infrastructures. However, determining relevant simulation parameters and correctly instantiating them to match the real Cloud performance is a difficult and costly operation, as minor configuration changes can easily generate an unreliable inaccurate simulation result. Using legacy values experimentally determined by other researchers can reduce the configuration costs, but is still inaccurate as the underlying public Clouds and the number of active tenants are highly different and dynamic in time. To overcome these deficiencies, we propose a novel model that simulates the dynamic Cloud performance by introducing noise in the computation and communication tasks, determined by a small set of runtime execution data. Although the estimating method is apparently costly, a comprehensive sensitivity analysis shows that the configuration parameters determined for a certain simulation setup can be used for other simulations too, thereby reducing the tuning cost by up to 82.46 percent, while declining the simulation accuracy by only 1.98 percent on average. Extensive evaluation also shows that our novel model outperforms other state-of-the-art dynamic Cloud simulation models, leading up to 22 percent lower makespan inaccuracy.

Amazon EC2's spot instances (SIs) represent a competitive Cloud resource in terms of price compared to reliable and fixed price options. The drawback, however, is that SIs may not always be available and they can be revoked at any given time. In this paper, we describe a comprehensive experimental evaluation for EC2 SIs to characterize their performance and behavior in three different regions each of which in a different continent. We describe the life cycle of SIs with the most important phases of an SI, introduce the most relevant events that can prevent a user from obtaining SIs, and draw important conclusions that can be exploited by the research community to effectively use the spot market. Our results reveal the fulfillment rate of requests for SIs, waiting time until requested SIs become fulfilled, details about the frequency of SI interruption, and how long SIs run before being interrupted. Our study also indicates that the SI frequency of interruption influences the fulfillment rate, SIs are highly reliable in the first 20 to 30 minutes after deployment, and SIs can be reclaimed by EC2 regardless of an SI's bid price and current workload when EC2 lacks resources for On-Demand and Reserved instances.

In this paper, we elaborate a new e-Testing platform based on essay questions and a special grading procedure. Our platform enables anonymity during the grading process, equal treatment of all answered questions, grading strategy for each question, easier and faster grading process. The realized user experience analysis shows that the students are satisfied with this new platform, especially that it enables them to more easily understand what they have missed during the provision of answers, and it enables continual learning even after the evaluation. Almost all students that have failed the course also were satisfied of the system. The system is described with its functions, benefits and impact.

Cloud is an eco-system in which virtual machine instances are starting and terminating asynchronously on user demand or automatically when the load is rapidly increased or decreased. Although this dynamic environment allows to rent computing or storage resources cheaper rather than buying them, still it does not guarantee the stable execution during a period of time as the traditional physical environment. This is emphasised even more for workflows execution, since they consist of many data and control dependencies, which cause the makespan to be instable when a workflow is being executed in different periods of time in Cloud. In this paper we analyse several parameters of workflow and the cloud environment that are expected to impact the workflow execution instability and investigate the correlation between them. The cloud parameters include the number of instances and their type, as well as the correlation with the efficient or inefficient execution of workflow parallel sections. We conduct a series of experiments, repeating each experiment by 30 test cases in order to evaluate instability for different cloud and workflow parameters. The results show a neglectful correlation between each pair of parameters, as well as the tasks and file transfers within the workflow. Oppose to the expectations, the distribution of the makespan per experiment does not always comply with the normal distribution, which is also not correlated to a particular cloud or workflow parameter.

Computer science students use data array processing in many courses. To exploit the full power of caches and obtain higher performance, they mostly use the textbook example of sequential access of data arrays. However, a lot of discrepancies occur and the expected performance is not obtained in real life program executions, mostly due to the existence of several cache levels, with various architectures and organizations, different cache sizes and associativity, and cache line width. In addition, Students must be aware of the cache misses categorized as compulsory, capacity and conflict misses. In this paper we present an extension of our EDUCache simulator, which simulates the work of a cache and gives details on data addresses, cache occupancy and other cache parameters. The extension is realized by integrating the cache parameters with GeoGebra, the open source tool whoose primary goal is for teaching and learning mathematics. In our case, we enhance the usage for novel learning and teaching of the hardware-based courses - the Computer architecture and organization, as well as high performance computing. The extension consists of visualizing the performance behavior when a program accesses huge data arrays, analyzing real three level cache based systems, used in almost all today's conventional processors. The possibility to visualize the cache behavior dependent on various cache parameters is the final benefit of this simulator, since the students can understand the cache parameters and their impact on the overall performance. Students can design their own cache with the parameters they like and then execute some arbitrary length data array access to see the performance of the execution.

Coding (programming) is a skill that nowadays is considered as an important asset. Teaching programming is an activity that becomes more and more popular. In the last few years there are great discussions on inclusion of coding in schools. These initiatives impacted the policy makers in Macedonia to include coding as a part of a new compulsory subject for the pupils at the age of 8. In this paper our focus is on the newest change in the Macedonian curriculum - the introduction of the course "Working with computers and basics of programming", which can be shortly named as "Computing" in 2015. Here, we present the proposed and accepted curriculum with emphasis on the topics about computational thinking and programming. Then we discuss the available software and tools suitable for the implementation of the before mentioned topics, and present a newly developed game. At the end we explain the background of the teachers, give the format of the preliminary training of all primary school teachers in the country and present the first impressions of the trainers that conducted the training, and we elaborate the teachers' feedback.

Computer technology changes rapidly, especially in the last decades. IEEE and ACM have developed curriculum recommendations for computer technology in the last 50 years and always add extensions or modify the content to keep the pace with the ongoing changes. Nowadays, five curricula are defined within the Computing Curricula: Computer Science, Computer Engineering, Information Systems, Information Technology and Software Engineering. Although some of them are updated in a period of 4-5 years, there are examples lasting for quite a long time, such as Computing from 2005 and Computer Engineering curricula from 2004. Still, the latest emerging technologies - Cloud computing, Internet of Things, Internet of Everything, Big Data, Machine to Machine and Human to Machine communications and interaction, software-defined everything, smart cities, high performance scaled computing, etc., raise the challenges if these curricula are ready to cover modern trends. Even more, the real question is whether they should be changed, upgraded or give rise to a new curriculum? This paper analyzes the new emerging trends and technologies and how they are covered in the current curricula that are present at our faculty (Computer Science and Computer Engineering). We present how a track of courses and their syllabuses are adapted towards these new emerging technologies, without changing the whole curriculum. There are multiple results of these changes. Students now can choose a track and learn the courses with increased interest; they can see the "whole picture" after finishing all courses of the track; they prepare more complex projects and they are happier with the changes. Finally, several diploma theses emerged that follow the current trends in the computer technology, which prepare the students to be already good engineers on the labor market. We strongly believe that with our new approach, the motivation for learning the hardware-based courses will be returned to the students, which will facilitate the trend of decreasing interest and number of engineering students.

Teaching a hardware-based course for computer science students is a very challenging task. Students need to do a lot of practical work in the laboratory, usually do some practical project, learn theory and finally take the exam. This is a very hard task for the first hardware-based course - Computer architecture and organization (CAO), especially if it is offered in the first year of studies. Many successful examples report introducing some online learning tools, such as distance laboratories or visual simulators, which help students easier to learn the required learning objectives. In this paper, we introduce a brand new approach of reorganizing the curriculum of the course of Discrete Mathematics (DM). This approach transfers the fundamentals of logic design earlier in the DM course. Since both DM and CAO courses are scheduled in parallel, this transfer and reorganization offers necessary background for topics learned in the CAO course. This change prepares the students with relevant content about digital logic part of the CAO course and results with an improvement of 50% of those who had passed the midterm exam compared to the previous years.

Unlike the network TCP/IP's and OSI's layered structure of protocols, which allows the independence of protocols of different layers, as well as defining the upper layer protocols through the protocols of the lower layers, the cloud service layers are tightly dependent on each other.For example, an application of the SaaS layer can neither communicate nor exchange data with another application found on the same layer.The goal of this paper is to overview the cloud interoperability and to analyze it as a service model perspective.Several aspects and categories of cloud interoperability are analyzed in this paper.

The speedup is usually limited by two main laws in high-performance computing, that is, the Amdahl's and Gustafson's laws.However, the speedup sometimes can reach far beyond the limited linear speedup, known as superlinear speedup, which means that the speedup is greater than the number of processors that are used.Although the superlinear speedup is not a new concept and many authors have already reported its existence, most of them reported it as a side effect, without explaining why and how it is happening.In this paper, we analyze several different superlinear speedup types and define a taxonomy for them.Additionally, we present several explanations and cases of superlinearity existence for different types of granular algorithms (tasks), which means that they can be divided into many sub-tasks and scattered to the processors for execution.Apart from frequent explanation that having more cache memory in parallel execution is the main reason, we summarize other different effects that cause the superlinearity, including the superlinear speedup in cloud virtual environment for both vertical and horizontal scaling.

Courses of computer science undergraduate studies are usually classified in two main groups: core and elective courses. The former are mandatory and students must pass them all so they can achieve Bachelor degree. The number of the latter is usually several times greater than the number that a student should select. This challenges the student which elective courses to select. Some elective courses have prerequisites, so the student has to pass other (elective) course(s), while for many others the student can be enrolled in without any prerequisites. The main dilemma for the student is whether to enroll in either more tightly coupled elective courses, or more loosely coupled elective courses. Choosing the former, the student will be directed in more specific areas of computer science. However, there are students that would like to learn broad areas of computer science. This paper focuses on the emerged group of students that select interdisciplinary courses. The experience presented in this paper shows that these students enroll into three completely divergent courses: Microprocessors and Microcontrollers, Software Architecture and Design and Human-Computer Interaction. The paper presents the examples of interdisciplinary projects completed by these students. These complex projects are proposed in the Human Computer Interaction course for the students that have already successfully taken the Microprocessors and Microcontrollers and/or Software Architecture and Design courses. Some students have accepted the challenges to work on the interdisciplinary projects, which resulted in defining and finishing various diploma theses that integrate knowledge areas of two or even all of the three courses.

Advanced learning technologies integrate questions as a very relevant function in building e-Testing systems and especially in intelligent tutoring systems. Therefore, a huge challenge is the generation of a necessary and relevant assessment content. The existing efforts to realize open source learning materials and establishment of massive open online courses introduce another challenge for realization of a sophisticated system and appropriate knowledge database with huge number of questions reflecting all relevant knowledge items (learning objectives). In this paper we present the experience in realization of a real e-Testing system by building two variations: the first with standard graphical interface, and the other with an enhanced media interaction supporting interactive images. We present several question generation strategies for conventional approach using multiple-choice questions and also for the system with interactive images. These techniques can be used to develop a large set of questions and assessment content.

Hardware-based courses require a lot of practical work during lab exercises, so students can achieve the learning outcomes more easily. Additionally, instructors expect the students to work on practical projects, which are more challenging and where students must use knowledge of several learning areas. Both lab exercises and projects should be interesting for the students, especially for computer science students, which usually do not prefer the hardware-based courses. This paper presents the new approach of the project gamification in the hardware-based course Microprocessors and Microcontrollers for computer science students. This change significantly improved the course - not only that it improved the average grade of the passed students, but it provoked the students to enroll the other hardware courses of the upper semesters. Even more so, some diploma theses involving microcontrollers were developed for the first time.

This paper presents a new curriculum for the Computer Networks Design course intended as advanced course for computer science students. We have developed an adaptive curriculum, which introduces two CISCO CCNA 3 (LAN Switching) and 4 (WAN Technologies) courses instead of the existing (regular) tutorials and laboratory exercises. The students can choose whether they would like to learn the course by a more practical approach using the CISCO CCNA integration, rather than by a more theoretical approach. The evaluation show much better results for the students that follow the course using the integration of the CISCO practical exercises, rather than those that follow the course regularly only by theoretical approach.

Recently, we have implemented three e-Learning tools in the Computer Architecture and Organization course with more than 800 students, by enabling access to two visual simulators in the laboratory exercises, as well as to an interactive online learning tool. We set a hypothesis that visual simulators will be evaluated by the students as preferred tool for learning in comparison to the online learning tool realized to assess the student knowledge. In this paper, we present the results of a survey to analyze the student satisfaction of introducing these e-learning tools. Our goal was to evaluate the usability and benefits of implemented interactive tools by analyzing anonymous and realistic student feedback. It will give a better understanding about the type of course organization and enable us to make proper decisions about the best learning model suited to the students need and requirements, especially in the case when the course is in the first study year.

Besides the fact that e-Assessment systems can efficiently conduct all paper based tests to evaluate knowledge and skills, they can offer a lot of new features via sophisticated information and communication technologies, including adaptive testing, immediate evaluation, etc. Most of the realised e-Assessment systems use pictures in the realisation of e-Testing, but the pictures are mainly used as supported media enhancement of the multiple choice questions. We introduce a brand new idea to use interactive images, where the user can navigate and zoom the picture and provide answers by clicking on an appropriate graphical object, mark a region, annotate, set an answer/comment on a given position, etc. The application domain of this innovation is huge, including e-Assessment for those sciences, where image analysis is essential, such as analysis of medical images, gathering map selective user's opinion etc. We have developed three new question types based on interactive images that offer these innovations. The impact is not just in enhancement of offered technology, but also on preventing various cheating methods, such as memorising, guessing, etc. These innovations can improve the assessment results, by a more correct evaluation and knowledge assessment.

Nowadays, both wired and wireless computer networks have significant importance. In addition, we are entering the world of big data analysis, where a lot of data is transferred from the sources to given computing centers for further processing. This trend requires changes in the computer science' computer networking curriculum in order to prepare the students with market opportunities and challenges after graduating. The Computer Networks (and / or data communications) course, or the whole knowledge area of networking and communication in general, are supposed to be a core part of computer science and net centric computing. Given the fact that these students prefer to learn software oriented courses, the Universities have to make the course more interesting and sophisticated enough to follow today's trends. In this paper, we present a new adaptive curriculum for the Computer Networks course. The students have the opportunity to choose between a practically or theoretically oriented course. Our intention is to make the most of the learning objectives in the course more practical and thus initiate increased interest of the students. However, the core part of theoretical lectures about low level reliable data communication is obligatory for both approaches.

This paper describes the development of a scalable and elastic e-Assessment solution deployed on the OpenStack open source cloud. The solution activates virtual machine instances with a particular size for each assessment according to the number of students that should be assessed, in order to achieve sustainable performance.

We have recently developed and implemented an E-learning and Benchmarking Platform for the Parallel and Distributed Computing Course. Although this platform reduced the students' problems, we faced an additional problem with the lack of resources expressed in a specific timeframe just before homework deadline. In this paper we propose using a cloud based architecture of an e-Elearning system. It is intended to upgrade the e-Learning and benchmarking platform prototype into a scalable and elastic platform, where the system will send the execution on the virtual machine instance hosted on the cloud node with available resources. Additionally, we propose a strategy for efficient utilisation of cloud resources is proposed.

The main objective of this paper is introducing intelligence in the e-Learning and e-Assessment processes. Therefore, we present an existing adaptive e-Learning and e-Assessment strategies, verify them with machine learning (ML) algorithms, build students Profile and eventually, we present our new model that will be able to estimate the final result of the overall students' work during the semester, taking into account all the learning objectives that the students have passed. Thus, our idea is creating an intelligent agent that will simulate the behavior of a real professor as much as possible.

Cloud architecture has the ability of sharing hardware resources and services among multiple tenants.In this paper we measure the performance for the multi-VM (multiple virtual machines) cloud architecture and compare it with the single-VM architecture.Renting resources on a cloud usually comes with a variety of options, such as use of more and smaller virtual machines or use of less and bigger virtual machines.The objective of this research is to find out which scenario gives better performance for the same price of rented resources.This will be done by comparing the following attributes: Average response time, Pages per second, Average page time, Requests per second, CPU time.We setup a hypothesis that the multi-VM approach would be better, and the best architecture is the one offering the highest number of small virtual machines, predicting that the computational demands will spread to different virtual machines in a balanced manner.The results confirm the hypothesis and lead to a recommendation for an optimal architecture of a cloud based solution for a common transactional web solution.

Cloud service providers offer their customers to rent or release hardware resources (CPU, RAM, HDD), which are isolated in virtual machine instances, on demand.Increased load on customer applications or web services require more resources than a physical server can supply, which enforces the cloud provider to implement some load balancing technique in order to scatter the load among several virtual or physical servers.Many load balancers exist, both centralized and distributed, with various techniques.In this paper we present a new solution for a low level load balancer (L3B), working on a network level of OSI model.When a network packet arrives, its header is altered in order to forward to some end-point server.After the server replies, the packet's header is also changed using the previously stored mapping and forwarded to the client.Unfortunately, the results of the experiments showed that this implementation did not provide the expected results, i.e., to achieve linear speedup when more server nodes are added.

SUMMARY The realization of modern processors is based on a multicore architecture with increasing number of cores per processor. Multicore processors are often designed such that some level of the cache hierarchy is shared among cores. Usually, last level cache is shared among several or all cores (e.g., L3 cache) and each core possesses private low level caches (e.g., L1 and L2 caches). Superlinear speedup is possible for matrix multiplication algorithm executed in a shared memory multiprocessor due to the existence of a superlinear region. It is a region where cache requirements for matrix storage of the sequential execution incur more cache misses than in parallel execution. This paper shows theoretically and experimentally that there is a region, where the superlinear speedup can be achieved. We provide a theoretical proof of existence of a superlinear speedup and determine boundaries of the region where it can be achieved. The experiments confirm our theoretical results. Therefore, these results will have impact on future software development and exploitation of parallel hardware on the basis of a shared memory multiprocessor architecture. Copyright © 2013 John Wiley & Sons, Ltd.

In this paper, we analyze the performance of computation intensive and memory demanding web services hosted in different environments with the same amount of resources, but orchestrated differently. A single-VM addresses the environment where all the resources are allocated in one huge virtual machine instance (VMI), while a multi-VM environment uses several smaller VMIs, each allocated with only one CPU core, and the load is balanced among them. We realize series of experiments with different server loads by changing the message size and the number of concurrent messages to analyze the optimal resource allocation to host web services in order to achieve maximum performance from the same resources in the cloud, i.e., for the same price. Despite the hypothesis that the single-VM environment provides better performance than the multi-VM environment, the results show totally opposite for almost all test cases. We achieve maximal relative speedup of 9.83 comparing the multi-VM environment to the single-VM.

Teaching computer science students how hardware devices work is often a very difficult process. It requires huge efforts both from the instructors and students. Although teaching the theoretical lessons and exercises is important process, the most important part of teaching hardware based courses is hand-on exercises since they can easily awake or asleep computer science student curiosity for hardware courses. This paper presents the continuum in evolution of a hardware based course “Microprocessors and Microcontrollers” in a software oriented curriculum. It presents the improvements realized in hands-on laboratory exercises and new teaching methodology for them. The results show that realized changes have improved the quality of grade distribution even more, and not only in hands-on lab exercises, but also in mandatory practical projects and course overall grade distribution.

Teaching computer architecture requires a lot of effort by the instructor. Introduction of simulators can improve the teaching process and increases student willingness and easier ability to learn the material. There are many visual simulators that cover courses about computer architecture and design. In this paper we present our EDUCache simulator as a supporting tool in the process of understanding the concepts of both computer architecture and computer organization. It focuses on understanding modern multi-layer, multi-cache and multi-core multi-processors. Apart of EDUCache's features to teach the students about the fundamentals of computer architecture and organization, it can be also used for performance engineering of software systems, i.e. the students will also discover the importance of computer architecture which will increase their curiosity for hardware courses in general.

All e-Assessment systems have several challenges, such as accurate evaluation, security and data privacy, performance, scalability etc. In this paper we focus on e-Assessment scalability and performance. We propose a SOA architecture of a cloud hosted e-Assessment system which uses scalability and elasticity in order to achieve sustainable performance. Our solution consists of three subsystems, the first for management, the second for reports, and the third for on-demand activities during the assessments. It reduces the overall costs since it uses minimum resources utilized only during the e-Assessment. Better performance is expected to be achieved since the active subsystem for each assessment works with much smaller data compared to the centralized one.