A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2013; you can also visit the original URL.
The file type is
In data center applications, predictability in service time and controlled latency, especially tail latency, are essential for building performant applications. This is especially true for applications or services built by accessing data across thousands of servers to generate a user response. Current practice has been to run such services at low utilization to rein in latency outliers, which decreases efficiency and limits the number of service invocations developers can issue while stilldoi:10.1145/2391229.2391238 dblp:conf/cloud/KapoorPTVV12 fatcat:5p6aqwgqwfbrtdwtjjclhl5l6i
more »... ng tight latency budgets. In this paper, we analyze three data center applications, Memcached, OpenFlow, and Web search, to measure the effect of 1) kernel socket handling, NIC interaction, and the network stack, 2) application locks contested in the kernel, and 3) application-layer queueing due to requests being stalled behind straggler threads on tail latency. We propose Chronos, a framework to deliver predictable, low latency in data center applications. Chronos uses a combination of existing and new techniques to achieve this end, for example by supporting Memcached at 200,000 requests per second per server at mean latency of 10 µs with a 99 th percentile latency of only 30 µs, a factor of 20 lower than baseline Memcached.
Big Data" computing increasingly utilizes the MapReduce programming model for scalable processing of large data collections. Many MapReduce jobs are I/O-bound, and so minimizing the number of I/O operations is critical to improving their performance. In this work, we present Themis, a MapReduce implementation that reads and writes data records to disk exactly twice, which is the minimum amount possible for data sets that cannot fit in memory. In order to minimize I/O, Themis makes fundamentallydoi:10.1145/2391229.2391242 dblp:conf/cloud/RasmussenLCPKV12 fatcat:mof52f5gcrbergu4rjbo33hmdq
more »... different design decisions from previous MapReduce implementations. Themis performs a wide variety of MapReduce jobs -including click log analysis, DNA read sequence alignment, and PageRank -at nearly the speed of TritonSort's record-setting sort performance  . 1 Themis is a Titan in Greek mythology who is tasked with creating balance, order and harmony.
This paper proposes a combination of an Intrusion Detection System with a routing protocol to strengthen the defense of a Mobile Ad hoc Network. Our system is Socially Inspired, since we use the new paradigm of Reputation inherited from human behavior. The proposed IDS also has a unique characteristic of being Semi-distributed, since it neither distributes its Observation results globally nor keeps them entirely locally; however, managing to communicate this vital information without accretionarXiv:1307.7833v1 fatcat:2vdgrctr3rb3dop4opfhmenu4u
more »... f the network traffic. This innovative approach also avoids void assumptions and complex calculations for calculating and maintaining trust values used to estimate the reliability of other nodes observations. A robust Path Manager and Monitor system and Redemption and Fading concepts are other salient features of this design. The design has shown to outperform normal DSR in terms of Packet Delivery Ratio and Routing Overhead even when up to half of nodes in the network behave as malicious.
While numerous studies have examined the macro-level behavior of traffic in data center networks-overall flow sizes, destination variability, and TCP burstiness-little information is available on the behavior of data center traffic at packet-level timescales. Whereas one might assume that flows from different applications fairly share available link bandwidth, and that packets within a single flow are uniformly paced, the reality is more complex. To meet increasingly high link rates of 10 Gbpsdoi:10.1145/2535372.2535407 dblp:conf/conext/KapoorSVP13 fatcat:7nzjyrytwrdybl6niokg4tjwyu
more »... nd beyond, batching is typically introduced across the network stack-at the application, middleware, OS, transport, and NIC layers. This batching results in short-term packet bursts, which have implications for the design and performance requirements of packet processing devices along the path, including middleboxes, SDN-enabled switches, and virtual machine hypervisors. In this paper, we study the burst behavior of traffic emanating from a 10-Gbps end host across a variety of data center applications. We find that at 10-100 microsecond timescales, the traffic exhibits large bursts (i.e., 10s of packets in length). We further find that the level of this burstiness is largely outside of application control, and independent of the behavior of higher level applications.
Although 'Halcyon' means serene environment which pervasive computing aims at, we have tried to present a different interpretation of this word. Through our approach, we look at it in context of achieving future 'calm technology'. The paper gives a general overview of the state of pervasive computing today, proposes the 'HALCYON Model' and outlines the 'social' challenges faced by system designers.arXiv:1003.2456v1 fatcat:lpyiwlmcxvgnxaynsgprihfgaa
This paper presents the design and implementation of an incrementally scalable architecture for middleboxes based on commodity servers and operating systems. xOMB, the eXtensible Open MiddleBox, employs general programmable network processing pipelines, with user-defined C++ modules responsible for parsing, transforming, and forwarding network flows. We implement three processing pipelines in xOMB, demonstrating good performance for load balancing, protocol acceleration, and applicationdoi:10.1145/2396556.2396566 dblp:conf/ancs/AndersonBKPV12 fatcat:oygtypnsrvbghgzbddm6b2tghm
more »... ion. In particular, our xOMB load balancing switch is able to match or outperform a commercial programmable switch and popular open-source reverse proxy while still providing a more flexible programming model.
doi:10.1007/s12098-014-1509-1 pmid:24944148 fatcat:lqmk6kebqbd3rdie3zjh33x5ba
Engineering large-scale data center applications built from thousands of commodity nodes requires both an underlying network that supports a wide variety of traffic demands, and low latency at microsecond timescales. Many ideas for adding innovative functionality to networks, especially active queue management strategies, require either modifying packets or performing alternative queuing to packets in-flight on the data plane. However, configuring packet queuing, marking, and dropping isdoi:10.1145/2396556.2396567 dblp:conf/ancs/Al-FaresKPDWPV12 fatcat:jci7irpqc5e57ibctwpcqjvvdi
more »... ging, since buffering in commercial switches and routers is not programmable. In this work, we present NetBump, a platform for experimenting with, evaluating, and deploying a wide variety of active queue management strategies to network data planes with minimal intrusiveness and at low latency. NetBump leaves existing switches and endhosts unmodified by acting as a "bump on the wire, " examining, marking, and forwarding packets at line rate in tens of microseconds to implement a variety of virtual active queuing disciplines and congestion control mechanisms. We describe the design of NetBump, and use it to implement several network functions and congestion control protocols including DCTCP and 802.1Qau quantized congestion notification.
Mandavia 1 , Karan Kapoor 2 , Vikram Dhar 3 , Anastasia Rachmanidou 4 . 1 Kings College London School Of Medicine, London, UK; 1 University Hospital Lewisham, London, UK; 1 University Hospital Lewisham ... ABSTRACTS ONE STOP VASCULAR CLINICS WOULD IMPROVE WAITING TIMES FOR PATIENTS REQUIRING VENOUS SURGERY 0205 QUALITY OF LIFE ASSESSMENT FOLLOWING ADENOTON-SILLECTOMY FOR OBSTRUCTIVE SLEEP APNOEA IN CHILDREN Rishi ...doi:10.1016/j.ijsu.2011.07.096 fatcat:thzmg4pvsjb4pjlxtk2nilupfi
Design and development of paclitaxel-loaded bovine serum albumin nanoparticles for brain targeting Bovine serum albumin (BSA) nanoparticles loaded with paclitaxel (PTX) were prepared using a desolvation technique. A 32 full factorial design (FFD) was employed to formulate nanoparticles. Nanoparticles were characterized for particle size by photon correlation spectroscopy and surface morphology by scanning electron microscopy (SEM) and transmission electron microscopy (TEM). Encapsulationdoi:10.2478/v10007-011-0012-8 pmid:21684843 fatcat:tbdbxc5fpbgtzpnffcge2znlwu
more »... ncy, zeta potential and particle yield were also determined. Response surface linear modelling (RSLM) was used to predict the optimal formulation. Various models were applied to determine the release mechanism from PTX nanoparticles. The effect of drug-polymer ratio on the release profile of formulations was observed and was applied to determine the suitability of the predicted optimal formulation. A preliminary study to determine the feasibility of targeting the prepared nanoparticles to brain was also carried out using mice as in vivo models.
Solving "Big Data" problems requires bridging massive quantities of compute, memory, and storage, which requires a very high bandwidth network. Recently proposed direct connect networks like Hy-perX  and Flattened Butterfly  offer large capacity through paths of varying lengths between servers, and are highly cost effective for common data center workloads. However data center deployments are constrained to multi-rooted tree topologies like Fat-tree  and VL2  due to shortest pathdoi:10.1109/ancs.2013.6665176 dblp:conf/ancs/RadhakrishnanTKPV13 fatcat:km7tyz5xmbgotnjehcpkanno2y
more »... outing and the limitations of commodity data center switch silicon. In this work we present Dahu 1 , simple enhancements to commodity Ethernet switches to support direct connect networks in data centers. Dahu avoids congestion hot-spots by dynamically spreading traffic uniformly across links, and forwarding traffic over non-minimal paths where possible. By performing load balancing primarily using local information, Dahu can act more quickly than centralized approaches, and responds to failure gracefully. Our evaluation shows that Dahu delivers up to 500% improvement in throughput over ECMP in large scale HyperX networks with over 130,000 servers, and up to 50% higher throughput in an 8,192 server Fat-tree network. Data center network; Direct connect network Dahu makes the following contributions: (1) Novel hardware primitives to efficiently utilize non-minimal paths in different topologies with a modest increase in switch state, while preventing persistent forwarding loops, (2) A virtual port abstraction that enables dynamic multipath traffic engineering, (3) A decentralized load balancing algorithm and heuristic, (4) Minimal hardware modifications for easy deployability, and (5) Large scale simulations on networks with over 130K servers to evaluate Dahu's performance. Our evaluation shows that Dahu delivers up to 50% higher throughput relative to ECMP in an 8,192 server Fat-tree network and up to 500% throughput improvement in large HyperX networks with over 130,000 servers. We are encouraged by these results, and believe that they are a concrete step toward our goal of combining the benefits of HPC and data center network topologies. MOTIVATION AND REQUIREMENTS Fully-provisioned multi-rooted tree topologies are ideal for targeting worst case communication patterns-where all hosts in the network simultaneously try to communicate at access link speeds. However, common communication patterns have only few network hot-spots and overprovisioning the topology for worst-case traffic results in high CAPEX. Oversubscribing the multi-rooted tree topology would reduce CAPEX, but network performance would also suffer in the common case since the oversubscribed layers of the tree have even lower capacity to tolerate hot-spots. Direct networks provide an interesting point in the design space of network topologies, since they provide good performance for most realistic traffic patterns, at much lower cost than fullyprovisioned Clos networks [1, pg.8-9][20, pg.6-8]. There are two defining characteristics of direct networks which distinguish them from tree based topologies. (1) Hosts are embedded throughout the structure of the network. Each switch has some hosts connected to it. (2) There are many network paths between any pair of serversbut they are of varying length. These properties of direct networks allow more flexible use of overall network capacity, with slack bandwidth in one portion of the network available to be leveraged by other congested parts of the network by forwarding traffic along longer less congested paths. In a sense the oversubscription is "spread throughout the network" rather than at specific stages or points in the topology. Direct networks are very popular in HPC-Titan, the world's fastest supercomputer, uses a 3D torus, a direct connect topology  . However, data centers have been largely constrained to multi-rooted trees due to commodity switch silicon and shortest path based routing protocols. Direct networks have significant potential in meeting the bandwidth requirements for data centers. Dahu presents simple enhancements to commodity Ethernet switches (both hardware and software) to support direct connect topologies in data centers.
A Mobile Adhoc Network (MANET) is a cooperative engagement of a collection of mobile nodes without any centralized access point or infrastructure to coordinate among the peers. The underlying concept of coordination among nodes in a cooperative MANET has induced in them a vulnerability to attacks due to issues like lack of fixed infrastructure, dynamically changing network topology, cooperative algorithms, lack of centralized monitoring and management point, and lack of a clear line of defense.arXiv:1006.1956v2 fatcat:ctmnps6u6rdg5o3wdl45e35vhu
more »... We propose a semi-distributed approach towards Reputation Based Intrusion Detection System (IDS) that combines with the DSR routing protocol for strengthening the defense of a MANET. Our system inherits the features of reputation from human behavior, hence making the IDS socially inspired. It has a semi-distributed architecture as the critical observation results of the system are neither spread globally nor restricted locally. The system assigns maximum weightage to self observation by nodes for updating any reputation values, thus avoiding the need of a trust relationship between nodes. Our system is also unique in the sense that it features the concepts of Redemption and Fading with a robust Path Manager and Monitor system. Simulation studies show that DSR fortified with our system outperforms normal DSR in terms of the packet delivery ratio and routing overhead even when up to half of nodes in the network behave as malicious. Various parameters introduced such as timing window size, reputation update values, congestion parameter and other thresholds have been optimized over several simulation test runs of the system. By combining the semi-distributed architecture and other design essentials like path manager, monitor module, redemption and fading concepts; Our system proves to be robust enough to counter most common attacks in MANETs.
The network architecture plays an important role in performance and speed of a deep network to classify images. In this paper we study the different architecture schemes and the variants proposed in GoogLeNet and inception networks. These variants are analyzed in terms of their computation efficiency and the network features and performances are juxtaposed on ImageNet 2012 dataset and critical review on inception networks is provided.doi:10.13140/rg.2.2.16212.35204 fatcat:vwmfgildwjc7lcieyus2qa5y5a
Contemporary convolutional neural network-based image classification models are empirical and lack generalizability, i.e. models tend to perform well only on a given dataset and under strict circumstances. In this paper, we carry out experiments to form a basis for using various variants of Inception blocks, traditional networks, and ResNets networks. The performance of these variants is evaluated on image recognition datasets CIFAR 10, CIFAR 100 and FER 2013.doi:10.13140/rg.2.2.32989.56806 fatcat:3gevj6qw4ffknael3n3ezpmizm
« Previous Showing results 1 — 15 out of 741 results