Public clouds like AWS, Azure, GCP, and OCI offer a range of services including Container as a Service (CaaS) and Relational Database as a Service (RDBaaS). From the perspective of an IT system provider there is a notable lack of information on the overall performance that can be achieved when using specific configurations of these interdependent CaaS and RDBaaS services. To address this issue and avoid incorrect architectural assumptions, it was decided to empirically evaluate the combined performance offered by CaaS and RDBaaS services, considering their hardware configurations selected based on a predetermined cost constraint. The experiments were conducted in two stages. The first stage was aimed at narrowing down the set of investigated CaaS and RDBaaS services. Using the measurements collected in the second stage, a statistical analysis was performed, comparing the performance achieved in the cloud environments with the performance obtained in the on-premise environment, taking cost constraints into account. The experiments conducted for the cloud services provided insights into their limitations and performance, enabling informed architectural decisions during the design of complex IT systems. The results of the analysis confirmed that the performance obtained for the studied combinations of services and their associated costs significantly varied among the different cloud providers.

The implementation of parallel applications is always a challenge. It embraces many distinctive design decisions that are to be taken. The paper presents issues of parallel processing with use of .NET applications and popular Database Management Systems (DBMSes). In the paper, four design dilemmas are addressed: how efficient is the auto-parallelism implemented in the .NET TPL library, how do popular DBMSes differ in serving parallel requests, what is the optimal size of data chunks in the data parallelism scheme, and how the TPL auto-parallelism behaves in the public clouds. They are analyzed in the context of the typical and practical business case originated from IT solutions which are dedicated for the energy market participants. The paper presents the results of experiments conducted in a controlled, on-premises and cloud environments. The experiments allowed to compare the performance of the TPL auto-parallelism with a wide range of manually set numbers of worker threads. They also helped to evaluate four DBMSes: Oracle, MySQL, PostgreSQL, and MSSQL in the scenario of serving parallel queries. Finally, they showed the impact of data chunk sizes on the overall performance.

This chapter contains sections titled: Introduction, Replication in Data-centric Service-Oriented Systems, Dependability in Resource-centric Systems, Related Work, Conclusion and Future Work, Acknowledgment, Note, References

Discovery in large, highly dynamic service oriented systems cannot be realized by traditional means, i.e. with a centralized index. Instead, distributed data structures are used, with a distributed hash table deployed on an overlay network being a typical example. The efficiency of overlay networks built on top of the IP network commonly suffers from the mismatch between the topologies of the overlay and the underlying IP network, resulting in unnecessary traffic and increased latencies. Substantial improvement can be achieved by optimizing the logical links between overlay nodes. In this paper, we propose a new method for self-optimization of a pervasive, DHT-based discovery service. Our method has no need for active measurement of inter-node latencies, thus minimizing network traffic costs of node insertion and topology maintenance. We verify our method by means of analysis of a large data sets of latency measurements between arbitrary nodes on the Internet, proving correlation among common IP prefix length of communicating nodes and latency.

In this paper, we evaluate the performance of a P2P- based content distribution system in heterogeneous, wireless networks. The mobile users coordinate each other with cooperation strategies enabled by the multi-source download mechanism, as in eDonkey or BitTorrent. Due to the mobility, vertical handovers between different wireless access technologies are required which may result in transmission delays and IP address changes of the switching peer. Hence, connections among users have to be reestablished and downloading users are requeued at a providing peer's waiting queue. In detail, we investigate the impact of requeueing with each VHO as well as the use of mechanisms that preserve the IP address and connections beyond VHOs, like MobilelP. Another important phenomenon occurring with VHOs is the abrupt change of the available bandwidth, e.g., from a fast WLAN connection to a rather slow UMTS connection. We evaluate the download times for files by means of simulation while considering different load scenarios in today's and future network layouts of the B3G network. As a result of the performance evaluation, we derive a new time-based cooperation strategy that counters the impact of mobility. Instead of downloading individual blocks of a file, a user gets a time slot at a providing peer. We show that this leads to a significant performance gain.

In this paper, we propose a new, dual method for self-optimization of a pervasive, DHT-based discovery service. This method addresses the topology mismatch problem. On one hand, it selects close neighbours based on static, readily available information (the IP addresses of the nodes) and thus does not require costly periodic probing of many nodes. On the other hand, it enables the overlay network to optimize its topology in run-time in a cost-effective manner. We prove the effectiveness of our method by statistical and experimental verification.

The efficiency of overlay networks built on top of the IP network is often threatened by the mismatch between the topologies of the overlay and the underlying IP network, resulting in unnecessary traffic and increased latencies. Substantial improvement can be achieved by optimizing the logical links between overlay nodes to better match the IP network topology. In this paper, we propose a new method for selfoptimization of a DHT-based peer-to-peer overlay. Our method has no need for active measurement of inter-node latencies, thus minimizing network traffic costs of node insertion and topology maintenance. We verify our method by means of analysis of large data sets of latency measurements between arbitrary nodes on the Internet, proving correlation among common IP prefix length of communicating nodes and latency.

A notion of a Dependable Information Service (DIS) is advocated as the means for applications in large scale distributed systems to select service instances meeting their requirements, functional as well as non-functional. Applications that need to provide certain correctness of results to the end user can only do so by requiring the services they use to guarantee particular level of correctness. When no such guarantees can be made, applications cannot provide the service and thus become unavailable. A system model is proposed to let applications temporarily accept lower correctness of used services in order to increase the availability caused by application accepted tradeoff. The proposed Dependable Information Service provides applications with a set of services appropriate for their demands and informs them of changes that can affect their quality level.

Discovery of service providers that are useful to service requesters is a common problem in service oriented systems, such as grids and ever more the whole World Wide Web itself. Current solutions do not address this problem adequately in very large, highly dynamic systems. Being usually designed for systems with predominantly static and relatively small set of information on existing service providers, they can not cope with large numbers of advertised services, service providers arbitrarily joining and leaving the system, and the constantly changing properties of the advertised services. This paper presents an architecture and a proof-ofconcept implementation of a discovery service for large, highly dynamic service oriented systems, based on the peerto- peer paradigm. Our discovery service on one hand allows for dynamic publication of service advertisements by very large numbers of service providers that arbitrarily join and leave the system. On the other hand it provides means for service requesters to dependably discover service providers that fullfil complex conditions specified by the requester. Also, the system allows the requester to dynamically establish a contract with the discovered service provider that describes the conditions under which the service will be provided to the requester.

Discovery of service providers that are useful to service requesters is a common problem in large, dynamic service-oriented systems, such as Grids. Simple service indices do not address this problem adequately when the service requesters need to describe complex conditions to be met by the service providers. Furthermore, the problem of sharing services among multiple service providers in order to enable the system to serve as many independent requesters as possible while still meeting their conditions is not addressed by such systems. This paper presents an architecture of a service discovery system for Grids, that allows dynamic registration of services and their provided capabilities. Discovery is based on matching of requested and provided capabilities. The matching process enables trading the degree of satisfaction of requested capabilities for availability of the system to multiple requesters. The discovery service seamlessly integrates with existing Grid infrastructure and specifications.

Unstructured peer-to-peer systems proved to be one of the most popular means of utilisation of computer systems in society. Their typical feature is enabling users to share resources without central administration. Consequently, service availability of P2P systems is not only influenced by resource availability, but - even more significantly - by resource discovery. In this paper, we are therefore identifying and classifying those properties of peer-to-peer systems that, if utilized.. can significantly improve the resource discovery probability, which in turn improves service availability. In other words, we are looking for properties that could help to obtain the best search results at the minimal cost. Further, we are mapping them to the set of design decisions for unstructured peer-to-peer systems. Finally, we are presenting an example how Our work can be used in practice.

Recently, peer-to-peer systems have become widely accepted and are probably the most recognizable examples of distributed applications. As they are maturing and their functionalities are becoming increasingly complex, the need for dependable solutions arises. In particular peer-to-peer systems' dependability is immensely influenced by their virtual overlay networks. The paper presents the results of analysis of diverse overlay networks with respect to their support for dependability.