23 research outputs found
Βελτίωση της απόδοσης του δικτύου αξιοποιώντας πολλαπλά μονοπάτια σε ασύρματα δίκτυα πλέγματος
No full textIn order to meet the increased demand for quality of service over wireless mesh networks, a large number of studies have suggested employingmultiple paths in parallel, in order to aggregate the scarce resources. Multipath utilization comes in many different flavoursincluding schemes that perform routing, resource allocation, flow and congestion control, and opportunistic forwarding.Multipath utilization has been shown to be advantageous in terms of delay, throughput, reliability, and load balancing.Two common performance goals are increased throughput and fault tolerance.For achieving reliability, multipath utilization is combined with redundancy.However, multipath utilization in wireless networks, is more complicated compared to their wired counterparts since transmissionsacross a link interfere with neighbouring links and may result in reduced network performance.In this thesis, static, random access, wireless mesh networks are considered, where receivers have multi-packetreception capabilities. Multiple unicast flows are forwarded to their destinations through node-disjoint paths.In the first part of the thesis, different forwarding schemes, employing multiple paths and different degrees of redundancyare compared in terms of delay and throughput.An analytical framework for expressing the throughput and delay of these schemes is evaluated, through Ns2 simulationsof various scenarios and is also extended for the case where link success probability is captured through the SINR model.In the second part of the thesis, multipath utilization for maximizing average aggregate flow throughput is addressed, for the aforementioned type of networks.A distributed flow rate allocation scheme that maximizes average aggregate flow throughput, while also providing bounded delay is proposed, thatdoes not employ any kind of redundancy.For the purposes of the suggested scheme, flow rate allocation is formulated as an optimization problem. A simple model for the average aggregate flow throughputis employed that captures both intra- and inter-path interference through the SINR model.As far as interference is concerned, two different variants of the suggested scheme are explored.In the first one, interference is approximated by considering only that link's dominant interferers.In the second variant, a simple topology is employed where receivers apply successive interference cancellation, instead oftreating interference as noise.For the evaluation process, Ns2 simulations of some illustrative topologies, along with several random wirelessones are employed. The proposed scheme is compared with three other simple flow allocation schemes both in termsof average aggregate flow throughput and flow delay.Προκειμένου να ικανοποιηθούν οι αυξανόμενες απαιτήσεις για ποιότητα υπηρεσίας πάνω από ασύρματα δίκτυα πλέγματος, ένας μεγάλος αριθμός μελετών προτείνει την αξιοποίηση πολλαπλών μονοπατιών ταυτόχρονα με σκοπό να αθροιστούν οι πόροι τους. Υπάρχουν αρκετές διαφορετικές προσεγγίσεις αξιοποίησης πολλαπλών μονοπατιών οι οποίες περιλαμβάνουν σχήματα που πραγματοποιούν δρομολόγηση, δέσμευση πόρων, έλεγχο ροής και ευκαιριακή προώθηση κίνησης. Η ταυτόχρονη αξιοποίηση πολλαπλών μονοπατιών έχει αποδειχτεί επωφελής ως προς την καθυστέρηση, το ρυθμό διαμεταγωγής δεδομένων, τηναξιοπιστία και την ισοκατανομή φόρτου κίνησης. Δύο διαδεδομένοι στόχοι είναι ο αυξημένος ρυθμός διαμεταγωγής δεδομένων και η αξιοπιστία. Προς επίτευξη της αξιοπιστίας συγκεκριμένα, η αξιοποίηση πολλαπλών μονοπατιών συνδυάζεται με πλεονασμό.Παρ' όλα αυτά, η αξιοποίηση πολλαπλών μονοπατιών στα ασύρματα δίκτυα είναι πιο περίπλοκη, από ότι στα ασύρματα καθώς μεταδόσεις σε μία ζεύξη παρεμβάλλουν γειτονικές ζεύξεις και μπορεί να οδηγήσουν σε μειωμένη δικτυακή απόδοση. Στην παρούσα διδακτορική διατριβή, θεωρούνται στατικά, ασύρματα δίκτυα πλέγματος, τυχαίας προσπέλασης, όπου τα μονοπάτια αποτελούνται από πολλαπλές ζεύξεις. Επιπλέον, οι παραλήπτες έχουν δυνατότητα για λήψη πολλαπλών πακέτων. Πολλαπλές ροές κίνησης, τύπου μονοεκπομπής, προωθούνται προς τους προορισμούς, μέσω μονοπατιών που δεν έχουν κοινούς κόμβους.Στο πρώτο τμήμα της διατριβής, διαφορετικά σχήματα προώθησης κίνησης που αξιοποιούν πολλαπλά μονοπάτια και διαφορετικούς βαθμούς πλεονασμού συγκρίνονται ως προς την καθυστέρηση και το ρυθμό διαμεταγωγής δεδομένων. Αξιολογείται, μέσω προσομοιώσεων διαφόρων σεναρίων, στον προσομοιωτή δικτύων Ns2 ένα αναλυτικό πλαίσιο το οποίο εκφράζει την καθυστέρηση και το ρυθμό διαμεταγωγής δεδομένων για τα προαναφερθέντα σχήματα προώθησης κίνησης. Το συγκεκριμένο αναλυτικό πλαίσιο επεκτείνεται και αξιολογείται επίσης για την περίπτωση όπου η πιθανότητα λάθους μίας ζεύξης εκφράζεται μέσω του μοντέλου σηματοθορυβικής σχέσης.Στο δεύτερο μέρος της διατριβής, εξερευνάται το θέμα της αξιοποίησης πολλαπλών μονοπατιών με σκοπό την αύξηση του μέσου αθροιστικού ρυθμού διαμεταγωγής δεδομένων, για τον προαναφερθέντα τύπο δικτύων. Προτείνεται ένα κατανεμημένο σχήμα κατανομής ροής, το οποίο μεγιστοποιεί τον μέσο αθροιστικό ρυθμό διαμεταγωγής δεδομένων για όλες τις ροές, ενώ ταυτόχρονα, παρέχει πεπερασμένη καθυστέρηση. Δεν χρησιμοποιείταικανενός είδους πλεονασμός. Για τους σκοπούς του συγκεκριμένου σχήματος, η κατανομή ροής σε πολλαπλά μονοπάτια διατυπώνεται σαν ένα πρόβλημα βελτιστοποίησης. Το προτεινόμενο σχήμα βασίζεται σε ένα απλό μοντέλο για το μέσο αθροιστικό ρυθμό διαμεταγωγής δεδομένων, το οποίο εκφράζει τις παρεμβολές από γειτονικά μονοπάτια, αλλά και τις παρεμβολές από κόμβους του ίδιου μονοπατιού, μέσω του μοντέλου σηματοθορυβικής σχέσης.Αναφορικά με τις παρεμβολές, εξετάζονται δύο εκδοχές του προτεινόμενου σχήματος κατανομής ροής. Στην πρώτη εκδοχή, η παρεμβολή σε μία ζεύξηπροσεγγίζεται λαμβάνοντας υπόψη μόνο τους σημαντικότερους παρεμβολείς. Στη δεύτερη εκδοχή, οι παραλήπτες εφαρμόζουν την τεχνική της διαδοχικήςαπαλοιφής παρεμβολών. Για την διαδικασία αξιολόγησης, αξιοποιούνται προσομοιώσεις, τόσο ενδεικτικών όσο και τυχαίων ασύρματων σεναρίων, χρησιμοποιώντας τον προσομοιωτή δικτύων Ns2. Το προτεινόμενο σχήμα κατανομής ροής συγκρίνεται με άλλα απλάσχήματα κατανομής ροής, τόσο ως προς την καθυστέρηση όσο και ως προς το ρυθμό διαμεταγωγής δεδομένων
Exploring the impact of node failures on the resource allocation for parallel jobs
No full textIncreasing the size and complexity of modern HPC systemsalso increases the probability of various types of failures. Failures maydisrupt application execution and waste valuable system resources dueto failed executions. In this work, we explore the eect of node failureson the completion times of MPI parallel jobs. We introduce a simulationenvironment that generates synthetic traces of node failures, assumingthat the times between failures for each node are independently dis-tributed, following the same distribution but with dierent parameters.To highlight the importance of failure-awareness for resource allocation,we compare two failure-oblivious resource allocation approaches withone that considers node failure probabilities before assigning a partitionto a job: a heuristic that randomly selects the partition for a job, andSlurm's linear resource allocation policy. We present results for a casestudy that assumes a 4D-torus topology and a Weibull distribution foreach node's time between failures, and considers several dierent tracesof node failures, capturing dierent failure patterns. For the synthetictraces explored, the benet is more prominent for longer jobs, up to82% depending on the trace, when compared with Slurm and a failure-oblivious heuristic. For shorter jobs, benets are noticeable for systemswith more frequent failures
Impact of Cache Coherence on the Performance of Shared-Memory based MPI Primitives: A Case Study for Broadcast on Intel Xeon Scalable Processors
No full text<p>Recent processor advances have made feasible HPC nodes with high core counts, capable of hosting tens or even, hundreds of processes. Therefore, designing MPI collective operations at the intra-node level has received significant attention over the past years. Deriving efficient algorithms for modern HPC nodes, with complex internal topologies and memory hierarchies, is challenging. Moreover, the cache coherency protocol, and its impact on performance, further complicate algorithm design for MPI collectives. This latter concern is often only partially addressed.</p><p>In this work, we demonstrate a particularly challenging performance degradation scenario in the case of shared-memory–based MPI broadcast, on three generations of the Intel Xeon Scalable processor architecture. Based on analysis of hardware-based performance counters, we conclude that the performance degradation observed is attributed to the cache coherency protocol and the multi-socket configuration of the execution platforms examined. We present a number of novel approaches designed to amend this effect, and apply them in a cache coherency aware version of the MPI broadcast implementation. We reduce the overall latency of the broadcast operation by up to 1.5 × and 1.25 × for small and large messages, respectively.</p>
Performance of Flow Allocation with Successive Interference Cancelation for Random Access WMNs
No full textIn this study, we explore the gain that can be achieved by jointly allocating flow on multiple paths and applying successive interference cancelation (SIC), for random access wireless mesh networks with multi-packet reception capabilities. We explore a distributed flow allocation scheme aimed at maximizing average aggregate flow throughput, while also providing bounded delay when SIC is employed. The aforementioned scheme is evaluated both in terms of delay and throughput, and is also compared with other simple flow allocation schemes. We present simulation results from three illustrative topologies. Our results show that the gain for the scheme with SIC, when compared with a variant that treats interference as noise (IAN), can be up to 15.2%, for an SINR threshold value equal to 0.5. For SINR threshold values as high as 2.0 however, SIC does not always result in higher throughput. In some scenarios, the gain of SIC over IAN is insignificant, while in some others treating interference as noise proves to be better. The reason is that, although SIC improves the throughput on a specific link, it also increases the interference imposed on neighboring receivers. We also show that the gain from applying SIC is more profound in cases of a large degree of asymmetry among interfering links
Flow Allocation for Maximum Throughput and Bounded Delay on Multiple Disjoint Paths for Random Access Wireless Multihop Networks
No full textIn this paper, we consider random-access wireless networks, withmultipacket reception capabilities, where multiple flows are forwarded to the gateways through node disjoint paths. We explore the issue of allocating flow on multiple paths, exhibiting both intra-and interpath interference, in order to maximize average aggregate flow throughput (AAT) and provide bounded packet delay. A distributed flow allocation scheme is proposed where allocation of flow on paths is formulated as an optimization problem. Through an illustrative topology, it is shown that the corresponding problem is nonconvex. Furthermore, a simple but accurate model is employed for the AAT achieved by all flows, which captures both intra-and interpath interference through the signal-to-interference-plus-noise ratio (SINR) model. The proposed scheme is evaluated through Ns2 simulations of several random wireless scenarios. Simulation results reveal that the model employed accurately captures the AAT observed in the simulated scenarios, even when the assumption of saturated queues is removed. Simulation results also show that the proposed scheme achieves significantly higher AAT for the vast majority of the wireless scenarios explored than for the following flow allocation schemes: one that assigns flows on paths on a round-robin fashion, one that optimally utilizes the best path (BP) only, and another one that assigns the maximum possible flow on each path. Finally, a variant of the proposed scheme is explored, where interference for each link is approximated by considering its dominant interfering nodes only.Funding Agencies|People Programme (Marie Curie Actions) of the European Unions Seventh Framework Programme FP7/under REA Grant [612361]; Heracleitus II-University of Crete, National Strategic Research Framework (ESPA); European Union</p
Performance of Flow Allocation with Successive Interference Cancelation for Random Access WMNs
No full textIn this study we explore the gain that can be achieved by jointly allocating flow on multiple paths and employing successive interference cancelation (SIC), for random access wireless mesh networks with multi-packet reception capabilities. More precisely, we explore a distributed flow allocation scheme aimed at maximizing average aggregate flow throughput, while also providing bounded delay, when SIC is employed. Simulation results derived from three illustrative topologies show that the gain over treating interference as noise (IAN) for this scheme can be up to 15.2%, for an SINR threshold value equal to 0.5. For SINR threshold values as high as 2.0 however, SIC does not always result in higher throughput. In some scenarios explored, the gain of SIC over IAN is insignificant, while in some others treating interference as noise proves better. The reason is that, although SIC improves the throughput on a specific link, it also increases the interference imposed on neighbouring receivers. We also show that, the gain of applying SIC is more profound in cases of a large degree of asymmetry among interfering links.</p
On the delay of a throughput optimal flow allocation scheme for random access WMNs
No full textIn this paper we consider random access wireless mesh networks with multi-packet reception capabilities where multiple flows are forwarded to the gateways through node disjoint paths. We address the issue of aggregate throughput optimal flow rate allocation with bounded delay guarantees for flows exhibiting both intra- and inter-path interference. In our prior work we have suggested a distributed flow allocation scheme that maximizes the average aggregate throughput of all flows also providing bounded delay guarantees. In this work we evaluate this scheme in terms of delay and compare its performance with the following schemes: a scheme that utilizes paths on a round-robin fashion, one that optimally utilizes the best path only and another one that assigns the maximum possible flow on each path. For the evaluation process Ns2 simulations of several random wireless scenarios are employed.</p
Throughput optimal flow allocation on multiple paths for random access wireless multi-hop networks
No full text"Implementation Notes for the Storage and Data Access Infrastructure", ExaNeSt project Deliverable D4.3
No full textIn this deliverable of WP4, D4.3 “Implementation Notes for the Storage and Data Access Infrastructure”, we describe the implementation of each software component and tool that was specified in deliverable D4.2:
Extensions and enhancements to the common I/O path in Linux with focus on two key areas: (i) supporting protected access to storage devices from user space, i.e. direct access to storage with minimal kernel related overheads; and (ii) enhancements to the access path for memory-mapped file access by, amongst others, making use of NVM devices. [FORTH]
Extensions to the BeeGFS parallel file system: Metadata replication mechanisms to handle consistency management and resilience to failures, as well as incorporating those mechanisms in a caching extension. [FHG]
Acceleration mechanisms for Host-to-VM and VM-to-VM interactions that take into account properties of the hardware platform and the unified interconnect. Two main technologies for hardware-assisted virtualisation are considered: (i) use of RDMA capabilities for accessing remote memory from within a virtual machine, and (ii) HPC API remoting to improve the performance of HPC APIs (specifically MPI) for applications that execute as an ensemble of virtual machines. [VOSYS]
The design of a new replication mechanism for analytical databases, and its implementation in MonetDB. This replication scheme has been designed in such a way that it will not only help improving the availability, reliability, performance and scalability of a database server in general, but also help exploiting the ExaNeSt storage infrastructure and the ExaNeSt platform in particular. [MDBS]
Several monitoring and testing tools: the ExaNeSt storage administration and monitoring [INFN]; experiment automation, stress-load and fault injection tools [FORTH]; tests simulating HPC application I/O behaviours [INFN]; and a database profiler [MDBS].
In the storage infrastructure, resilience is mainly addressed by replication mechanisms in the file system. The system’s resilience will be stressed by the stress-load and fault injection tools, while its health is monitored by the monitoring tools. In addition, the system’s resilience level will be evaluated by the HPC and DBMS applications. To achieve the objectives of WP4, the storage and data access infrastructure proposed by ExaNeSt is built upon the following main components: (i) the distributed file system BeeGFS, (ii) Linux data access technologies, (iii) KVM based virtualisation, and (iv) storage system monitoring and administration tools. Moreover, we will use the state-of-the-art HPC applications and analytical databases to guide the design of the envisioned infrastructure and to evaluate and showcase the result.ExaNeSt project Deliverable D4.
