The actor programming model is gaining popularity due to the prevalence of multi-core systems along with the rising need for highly scalable and distributed applications. Frameworks such as Akka, Orleans, Pony, and C++ Actor Framework (CAF) have been developed to address these application requirements. Each framework provides a runtime system to schedule and run millions of actors, potentially on multi-socket platforms with non-uniform memory access (NUMA). However, the literature provides only limited research that studies or improves the performance of actor-based applications on NUMA systems. This paper studies the performance penalties that are imposed on actors running on a NUMA system and characterizes applications based on the actor type, behavior, and communication pattern. This information is used to identify workloads that can benefit from improved locality on a NUMA system. In addition, two locality-and NUMA-aware work-stealing schedulers are proposed and their their respective execution overhead in CAF is studied on both AMD and Intel machines. The performance of the proposed work-stealing schedulers is evaluated against the default scheduler in CAF.

Natural short term fluctuations in the load of transactional data systems present an opportunity for power savings. For example, a system handling 1000 requests per second on average can expect more than 1000 requests in some seconds, fewer in others. By quickly adjusting processing capacity to match such fluctuations, power consumption can be reduced. Many systems do this already, using dynamic voltage and frequency scaling (DVFS) to reduce processor performance and power consumption when the load is low. DVFS is typically controlled by frequency governors in the operating system, or by the processor itself. In this paper, we show that transactional database systems can manage DVFS more effectively than the underlying operating system. This is because the database system has more information about the workload, and more control over that workload, than is available to the operating system. We present a technique called POLARIS for reducing the power consumption of transactional database systems. POLARIS directly manages processor DVFS and controls database transaction scheduling. Its goal is to minimize power consumption while ensuring the transactions are completed within a specified latency target. POLARIS is workload-aware, and can accommodate concurrent workloads with different characteristics and latency budgets. We show that POLARIS can simultaneously reduce power consumption and reduce missed latency targets, relative to operating-system-based DVFS governors.

Addressing, routing, and forwarding in the Internet must form a coherent architecture that satisfies user and technical requirements, such as performance, robustness, and efficiency. This paper presents the Core-Rooted Path Addressing (CRPA) architecture, which is a novel combination of mostly known architectural components. CRPA is designed as a network layer for physical topologies and combines a generic rendezvous service with the path addressing principle. The benefits of CRPA include forwarding components that are completely independent of dynamic routing, as well as compact forwarding tables. The key challenge for such a design is the overall practicality of its addressing scheme. This paper investigates the characteristics of CRPA using Internet topology analysis and simulation, and the findings indicate that it is practicable. Forwarding table sizes are quite manageable, and the frequency and scope of dynamic updates are reduced. While actual deployment might be unrealistic in the current Internet ecosystem, its straightforward design can make CRPA a valuable benchmark for systematic reasoning about network architecture.

Adversarial Queueing Theory (AQT) has shown that seemingly innocent traffic injection rates might lead to unbounded queues in packet-switched networks - depending on scheduling strategies as well as topological characteristics. Little attention has been given to quantifying these effects in realistic network configurations. In particular, the existing AQT literature makes two unrealistic assumptions: infinite buffers and perfect synchrony. Because finite buffers inherently limit queue sizes, adversarial effects ultimately lead to packet loss which we address in this work. In addition, we study the effect of imperfect network synchronization under the packet loss metric. Our results, using analysis and simulation, indicate that classical AQT examples appear harmless under realistic assumptions but for a novel class of adversaries considerably higher loss can be observed. We introduce this class by giving examples of two new AQT concepts to construct loss-efficient network adversaries. Our analysis proves the robustness of these new adversaries against randomized de-synchronization effects in terms of variable link delays and nodal processing.

There are several analytical results on distributed hash tables (DHTs) that can tolerate Byzantine faults. Unfortunately, in such systems, operations such as data retrieval and message sending incur significant communication costs. For example, a simple scheme used in many Byzantine fault-tolerant DHT constructions ofnnodes requiresO(log3n) messages; this is likely impractical for real-world applications. The previous best known message complexity isO(log2n) in expectation. However, the corresponding protocol suffers from prohibitive costs owing to hidden constants in the asymptotic notation and setup costs. In this paper, we focus on reducing the communication costs against a computationally bounded adversary. We employ threshold cryptography and distributed key generation to define two protocols, both of which are more efficient than existing solutions. In comparison, our first protocol is deterministic withO(log2n) message complexity, and our second protocol is randomized with expectedO(logn) message complexity. Furthermore, both the hidden constants and setup costs for our protocols are small, and no trusted third party is required. Finally, we present results from microbenchmarks conducted over PlanetLab showing that our protocols are practical for deployment under significant levels of churn and adversarial behavior.

This paper presents a novel approach to minimally invasive service differentiation in packet-switched networks. Instead of actively managing service allocation, a simple differentiated queueing algorithm provides traffic classes with essentially the same best-effort service that would result from plain FIFO service using a single queue for all traffic. However, each class is served from a separate virtual queue, which is configured with an individual deterministic delay bound that is enforced in the presence of dynamically varying packet arrival rates. The main advantage of such a scheme for service differentiation is administrative simplicity, because it only needs minimal configuration by a network operator and does not necessarily require control plane functionality. Further, it does not inherently prefer some traffic classes over others and thus satisfies even the most radical definitions of network neutrality. In the paper, the basic approach is motivated with the help of various use case scenarios. A fairly simple and efficient algorithm is presented to implement the differentiated queueing scheme. Finally, a number of simulation experiments and results are shown that confirm the intuitive functionality of the algorithm.

There are several analytical results on distributed hash tables (DHTs) that can tolerate Byzantine faults. Unfortunately, in such systems, operations such as data retrieval and message sending incur significant communication costs. For example, a simple scheme used in many Byzantine fault-tolerant DHT constructions of n nodes requires O(log3n) messages, this is likely impractical for real-world applications. The previous best known message complexity is O(log2n) in expectation, however, the corresponding protocol suffers from prohibitive costs owing to hidden constants in the asymptotic notation and setup costs. In this paper, we focus on reducing the communication costs against a computationally bounded adversary. We employ threshold cryptography and distributed key generation to define two protocols both of which are more efficient than existing solutions. In comparison, our first protocol is deterministic with O(log3n) message complexity and our second protocol is randomized with expected O(log n) message complexity. Further, both the hidden constants and setup costs for our protocols are small and no trusted third party is required. Finally, we present results from micro benchmarks conducted over PlanetLab showing that our protocols are practical for deployment under significant levels of churn and adversarial behaviour.

This paper presents Interleaved Stratified Timer Wheels as a novel priority queue data structure for traffic shaping and scheduling in packet-switched networks. The data structure is used to construct an efficient packet approximation of general processor sharing (GPS). This scheduler is the first of its kind by combining all desirable properties without any residual catch. In contrast to previous work, the scheduler presented here has constant and near-optimal delay and fairness properties, and can be implemented with O(1) algorithmic complexity, and has a low absolute execution overhead. The paper presents the priority queue data structure and the basic scheduling algorithm, along with several versions with different cost-performance trade-offs. A generalized analytical model for rate-controlled rounded timestamp schedulers is developed and used to assess the scheduling properties of the different scheduler versions.

Kernel-based group unicast has been suggested as an efficient mechanism for transmitting the same data to multiple recipients. In this paper, we present a new system call, sendgroup(), which also supports per-recipient private data, but only uses a single inkernel copy of the shared data. We assess the performance of the new system call using micro-benchmarks on three different operating systems. Further, we incorporate sendgroup() into a popular multimedia server and demonstrate an efficiency improvement of ~45% in a representative live-broadcasting scenario. These results show that the new system call is applicable in real-world scenarios, and that its usage can lead to significant performance improvements. Moreover, we demonstrate how Amdahl's Law, when applied to the results of the micro-benchmarks, along with precise analysis of the cost of sending packets, can be used to accurately predict the impact of sendgroup() on this server.

The de facto service architecture of today's communication networks, in particular the Internet, is heterogeneous, complex, ad hoc, and not particularly well understood. With layering as the only means for functional abstraction, and even this violated by middle-boxes, the diversity of current technologies can barely be expressed, let alone analyzed. As a first step to remedying this problem, we present an axiomatic formulation of fundamental forwarding mechanisms in communication networks. This formulation allows us to express precisely and abstractly the concepts of naming and addressing and to specify a consistent set of control patterns and operational primitives, from which a variety of communication services can be composed. Importantly, this framework can be used to (1) formally analyze network protocols based on structural properties, and also to (2) derive working prototype implementations of these protocols. The prototype is implemented as a universal forwarding engine, a general framework and runtime environment based on the Click router.

virtual environment (DVE) have become increasingly popular. Many DVE implementations use a client-server architecture that requires the server to send the same data to all members of a collaborating or interacting group. This type of group communication operation is often implemented by sending data from the server to each recipient in a unicast fashion. The problem with this approach is that the cost of communication at the server does not scale very well with the number of participants because the application requires significant interaction with the operating system, network stack and drivers for each individual send. In this paper, we first propose a general analytic framework for predicting how group communication performance impacts DVE server capacity. We then conduct an experimental evaluation to determine the extent to which using a kernel-based group communication mechanism reduces the cost of group send operations. Lastly, we use the measurements obtained from these experiments to demonstrate how to apply the analytic framework by determining the extent to which the kernel-based group communication mechanism permits example applications to scale to more users.

Combining low-complexity packet marking at internal nodes with path-based load observation at edge gateways enables a rich set of control mechanisms in a packet-switched network. In this paper, we propose a method to estimate the relative per-node load at internal nodes, based on only the aggregated marking signal available at edge nodes, without any knowledge of internal capacities. We present the design of an integrated network system that supports both admission control and per-node load estimation using only two bits in the packet header, which requires a specific encoding of the two distinct marking signals. Finally, we describe a prototype implementation and a few simulation and lab experiments as a proof-of-concept for this approach.

A common requirement of many Internet services is to send exactly the same data to a number of hosts at the same time. Without IP-level multicast, this form of group communication is realized by unicasting the data to each desired host. Although this approach is portable and easy to implement, it is extremely inefficient for the sending host. In this paper, we propose a kernel-based technique to efficiently facilitate unicast send operations for group communication with only minimal additions to the sending operating system interface and implementation. We present the design and prototype implementation of our approach and experimentally demonstrate the significant performance improvements it provides. Additionally, we conduct experiments to decompose the processing costs in the network stack and show that the biggest cost reductions are not necessarily due to reduced memory copying.

Multi-hop wireless networks have the potential to dramatically reduce the cost of deploying communication infrastructure. However, the nature of this technology limits the capacity of radio links. Thus, it is important to utilize them as efficiently as possible. In this paper, we investigate load balancing across multiple paths as a possible mechanism to improve performance in multi-hop wireless networks. Given the inherent interference of multi-hop transmissions in a single radio channel, it is generally assumed that single-channel multipath routing cannot provide any benefits, but in fact would have detrimental effects on resource efficiency. However, a careful investigation of the issue reveals that under certain theoretic conditions, significant gains are possible. In fact, we show throughput improvements of 80-100% in some scenarios. We present a novel interference metric to assess the quality of a set of disjoint paths. We further present a heuristic path selection algorithm to find appropriate routing paths in structured networks, which is a first step towards the application of our basic results in realistic scenarios.

We describe implementation aspects and performance results of an innovative and publicly available RSVP implementation. Much debate exists about the applicability of RSVP as a signalling protocol in the Internet, particularly for a large number of unicast flows. While there has been a significant amount of work published on the theoretical concepts of RSVP signalling and conjectures about its presumed shortcomings, rather little attention has been paid to the implementation details of the core protocol engine. With our work, in spite of being still far from a final judgement, we try to shed light on this issue by presenting certain design details of a new implementation and a study about its performance. One particular result is given by the observation that a relatively cheap router based on PC hardware can sustain the signalling for more than 50,000 unicast flows.

Within a global networked environment, security aspects have become more and more important and access control at network borders is considered essential. For this purpose firewall systems are used which provide a well-established security mechanism to restrict the exchanged traffic to a certain subset of users and applications. In order to cope with the increasing demand for new applications, a firewall must be flexible and extensible to support such new applications and their protocols. RSVP is a dynamic signalling protocol, which has been invented to negotiate resource requirements between end systems and a packet-based communication network. We investigate the interoperation of RSVP with a firewall system in order to support new applications in a generic way. We show how the resulting system flexibility allows for a variety of employment scenarios and incremental deployment of such a technology. We back up our claims by describing a prototype that we have implemented.

The support of a single signalling protocol for all components on the data path of a QoS-based transmission cannot necessarily be assumed. We investigate the issues surrounding hierarchically layered network QoS signalling configurations, as e.g. can be found in RSVP over ATM signalling. After introducing and discussing these issues we conclude that many of the decisions involved require understanding the performance characteristics of such layered signalling configurations. We therefore describe the implementation of an RSVP/ATM edge device, which we then use to conduct measurements in a configuration that involves in effect three layered signalling protocols: RSVP and ATM's UNI and PNNI. Using these measurements we review the issues in layering signalling protocols and reinforce design decisions being taken for the edge device mediating between the different mechanisms of the two network QoS architectures.