It is becoming increasingly commonplace for multiple applications with different quality of service (QoS) requirements to share the resources a distributed system. Within this environment, resource management algorithms must take into account QoS desired by and ability system provide it. We present taxonomy specifying components system, from down resources. specify as combination metrics policies. are used performance parameters, security relative importance work in define three types...

The paper describes two innovative models that facilitate adaptive QoS driven resource management in distributed systems comprising heterogeneous computing, storage, and communication resources. first model, denoted the Logical Application Stream Model (LASM), recursively captures a application's structure, requirements, relevant end to quality of service (QoS) parameters. Upon invocation application by user, manager can use this model initially structure application, allocate resources...

We propose a mechanism for distributed radio resource management using multi-agent deep reinforcement learning to mitigate the interference among concurrent transmissions in wireless networks. equip each transmitter network with RL agent, which receives partial delayed observations from its own associated users, while also exchanging neighboring agents, and decides on user serve what transmit power level use at scheduling interval. scalable agent design, where dimensions of observation...

The recent development of reinforcement learning (RL) has boosted the adoption online RL for wireless radio resource management (RRM). However, algorithms require direct interactions with environment, which may be undesirable given potential performance loss due to unavoidable exploration in RL. In this work, we first explore use <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">offline</i> solving RRM problem. We evaluate several...

We propose a mechanism for distributed resource management and interference mitigation in wireless networks using multi-agent deep reinforcement learning (RL). equip each transmitter the network with RL agent that receives delayed observations from its associated users, while also exchanging neighboring agents, decides on which user to serve what transmit power use at scheduling interval. Our proposed framework enables agents make decisions simultaneously manner, unaware of concurrent other...

In this paper we describe the implementation of our end-to-end QoS-driven resource management scheme, called ERDoS, within a CORBA-compliant ORB that call ERDoS QoS ORB. Unlike other real-time CORBA implementations focus on support for simple client-server applications, provides (i.e., spanning computer network, and storage resources) to while retaining benefits an open distributed object system. Specifically present three contributions. First, model describing applications as combination...

A distributed system that supports applications with varying quality of service (QoS) requirements must use many adaptive mechanisms to guarantee performance the applications. This paper presents mechanism graceful adaptation for multiple levels performance. The different amounts resources. Our ERDoS architecture such in order provide end-to-end guarantees we present is based on market protocols which trade-offs are made between QoS dimensions an application so as satisfy application's needs...

Because the ATM Forum does not standardize ABR flow control algorithm that an switch should run, some networks are likely to contain switches run different algorithms. Even if all within a "cloud" of use same algorithm, individual clouds (each with it own algorithm) will be interconnected by virtual circuits. Virtual circuits which traverse multiple therefore controlled two algorithms concurrently. We explore ramifications mixing in network. identify rate mismatch problem, arises when...

Complex distributed systems require resource management capabilities that can allocate resources to applications, monitor and control the use of resources, reallocate in response anomalies. To utilize efficiently meet application system wide requirements, must consider availability, policies, application's quality service (QoS) requirements when making allocation decisions. We identify several open research areas QoS based briefly describe ongoing Global Resource Management (GRM) project,...

AI deployed in many real-world use cases should be capable of adapting to novelties encountered after deployment. Here, we consider a challenging, under-explored and realistic continual adaptation problem: agent is continuously provided with unlabeled data that may contain not only unseen samples known classes but also from novel (unknown) classes. In such challenging setting, it has tiny labeling budget query the most informative help learn. We present comprehensive solution this complex...

Interference among concurrent transmissions in a wireless network is key factor limiting the system performance. One way to alleviate this problem manage radio resources order maximize either average or worst-case However, joint consideration of both metrics often neglected as they are competing nature. In article, mechanism for resource management using multi-agent deep reinforcement learning (RL) proposed, which strikes right trade-off between maximizing and $5^{th}$ percentile user...

Defining the interoperable interface between clients and end-system devices within a distributed control system has become increasingly more difficult. Because level of intelligence increased, multiple vendor implementations may present widely disparate sets capabilities interfaces single class devices. Our solution establishes an architecture that defines common to superset for device provides well-defined mechanisms understanding subsets capabilities. The abstract model approach, which we...

Resource sharing between multiple workloads has become a prominent practice among cloud service providers, motivated by demand for improved resource utilization and reduced cost of ownership. Effective sharing, however, remains an open challenge due to the adverse effects that contention can have on high-priority, user-facing with strict Quality Service (QoS) requirements. Although recent approaches demonstrated promising results, those works remain largely impractical in public environments...

A growing number of service providers are exploring methods to improve server utilization and reduce power consumption by co-scheduling high-priority latency-critical workloads with best-effort workloads. This practice requires strict resource allocation between contention maintain Quality-of-Service (QoS) guarantees. Prior work demonstrated promising opportunities dynamically allocate resources based on workload demand, but may fail meet QoS objectives in more stringent operating...