Explore chapters and articles related to this topic
Architectural Framework for Asynchronous Transfer Mode Networks: Broadband Network Services
Published in Jerry D. Gibson, The Communications Handbook, 2018
Gerald A. Marin, Raif O. Onvural
followed by a silent period during which the source processes data in its memory and does not generate any network traffic. The high burstiness of the source would mean that the periods of activity persist for considerably less time than the periods of inactivity. On the other hand, consider a PCM voice application generating frames at the constant rate of 64 Kb/s. The challenge is to provide network services such that when the two sources are active simultaneously on one or more physical links, the QoS provided to the voice source does not degrade due to a sudden spurt of cells belonging to the distributed computing application. The problem of dealing with applications with widely varying burstiness in an integrated manner cannot be avoided as we move toward VBR multimedia services; network services that meet this challenge are needed to deploy integrated services.
ATM Network Interfaces and Protocols
Published in P. S. Neelakanta, ATM Telecommunications, 2018
This flexibility comes from the rigid size of the cells that are switched on a hardware basis. Each cell includes a header that identifies which set of cells – which streams of information – it belongs to and, as in a connectionless packet system, where it is going. A stream of cells with the same header constitutes a virtual circuit. The cells from different virtual circuits are interleaved, or multiplexed, by the ATM switching system so that the overall speed and burstiness of each reflect the speed and burstiness of each original data stream that came into the network through a customer-access interface. A continuous-bit-rate video circuit will get its cell into the overall bit stream in a regular pattern, separated by regular numbers of other cells. A bursty circuit for file transfer might be allocated a number of cells in a row, but there would be longer stretches in which none of the file-transfer cells appeared.
Applications in Communications
Published in Phil Mars, J.R. Chen, Raghu Nambiar, Learning Algorithms, 1996
Phil Mars, J.R. Chen, Raghu Nambiar
where Pi is the peak rate of the ith hold on call, and C is the capacity of the output link at the node. This approach is quite similar to bandwidth reservation in STM, but with the added flexibility of being able to reserve any peak rate required rather than a multiple of a base channel rate. The advantages with nonstatistical multiplexing are minimal cell delay and no cell loss due to buffer overflow. However when a large proportion of the traffic flow in the link is bursty, nonstatistical multiplexing can show low efficiency in making use of bandwidth resource. Thus, statistical multiplexing is considered to exploit the burstiness of traffic flow and obtain potential gain in bandwidth efficiency. In statistical multiplexing, the total peak cell transmission rate of all the accepted calls is allowed to exceed the capacity of the link at the expense of cell delay or cell loss. However, under a proper control strategy the cell delay or cell loss can be controlled within a tolerable range.
Unsupervised Event Detection Using Self-learning-based Max-margin Clustering: Analysis on Streaming Tweets
Published in IETE Journal of Research, 2020
Among others, the problem of event extraction from Twitter data streams poses sufficient challenges due to the aforementioned issues. Events can loosely be defined as the depictions of surrounding states of the users at the temporal scale. In short, everything posted by the users on Twitter is largely focused on events taking place around them. In recent past, there are various methods suggested to extract events from Twitter data: some of them consider keyword burstiness [12,13] while others depend on some external thesaurus [3] to predict the event candidates. However, given the noisy contents of the tweets along with the possible heterogeneity in the contents, it is potentially difficult to predict the events from tweet streams without concrete supervision.
File Semantic Aware Primary Storage Deduplication System
Published in IETE Journal of Research, 2022
Amdewar Godavari, Chapram Sudhakar, T. Ramesh
iDedup [8] is considered to be a state-of-the-art work among the primary storage deduplication systems. It is a file type oblivious deduplication system and assumes a strong workload locality. iDedup maintains locality preserving fingerprint caching and applies selective deduplication on large sized files. Performance Oriented data Deduplication (POD) [7] is a block-level deduplication system based on strong workload locality assumption. POD applies fixed-size chunking and deduplication on file irrespective of size. It maintains a locality-based fingerprint cache to accelerate deduplication. In addition to this, adaptive cache sizing between data cache and index cache is proposed to tackle I/O burstiness. FADD considers file size as well as file type and proposes similarity-based indexing. Partially Deduped File System (PDFS) [6] proposes file type oblivious primary storage deduplication. In this system, sub-file level chunking is applied to generate segments. Similar segment identification is a heavy computation process. Similar segments are indexed using a similarity-based index table. This approach allows partial lookup of metadata. FADD is applying file type-specific chunking. It uses the less resource-intensive process for identifying similar segments with the aim of solving the data fragmentation problem. ProSy [21] is a file-level but file type oblivious, primary storage deduplication system. It performs variable sized chunking of the file to get segments. By performing byte by byte comparison, similar segments are grouped together into one category. It has more computation overhead. Heuristically Arranged Non-Backup Deduplication System (HANDS) [17] is a block-level deduplication system and it tries to solve the disk-bottleneck problem by identifying working sets of fingerprints using a heuristic method. FADD system is a file type aware deduplication system with less computation overhead. Hybrid deduplication System (HDS) [22] is file type oblivious block level deduplication system. Irrespective of file type, HDS considers request size for deduplicaton. HDS maintains indexing information separate for small and large size requests. Small size request deduplication is accelerated with the help of hash table and large size request deduplication is accelerated by leveraging similarity based indexing. Apart from this, the selective deduplication is applied on large request to reduce the data fragmentation. Selective deduplication means incoming large request is deduplicated if three consecutive data blocks matches with already existing data blocks. Otherwise, request is not deduplicated.
Modelling and analysis of D-BMAP/D-MSP/1 queue using RG-factorization
Published in Quality Technology & Quantitative Management, 2021
The discrete-time batch Markovian arrival process (D-BMAP) generalizes the discrete-time Markovian arrival process (D-MAP) by allowing batch arrivals. It can capture the correlation that exists among successive inter-batch arrival times. The D-BMAP which is proposed by Blondia (1993) is arguably the most versatile and tractable generalization of batch Bernoulli process. The discrete-time Markovian service process (D-MSP) is independent of the arrival process. It generates real service completions only when the server is busy. In this regard, readers are referred to Blondia (1993), Wang et al. (2011), Frigui et al. (1997), Herrmann (2001), Samanta and Zhang (2012), and Samanta and Nandi (2020) to have the basic comprehensions on D-BMAP and D-MSP. Several connections (data, voice, video, etc.) generate traffic streams with very different characteristics (required bandwidth, burstiness, correlation, etc.). The Markov models characterized by non-correlated arrival and service processes can not adequately as well as definitely adopt the property of complex multimedia traffic such as Web browsing, teleconferencing, IEEE 802.16m WiMAX protocol, MPEG-4 video transmission over wireless networks, fourth-generation (4G) or fifth-generation (5G) cellular network and voice over internet protocol (VoIP). Traffic with certain bursty characteristics can be qualitatively modelled by above non-renewal arrival and service processes. Batch arrival in queueing phenomenon is one of the most important attribute for its wide range of applications such as letters arriving at a post office, ships arriving at a port in convoy, people going to a hospital, theatre as well as restaurants, packet aggregation in optical burst switched (OBS) networks and IEEE 802.11n WLANs, modelling IP traffic, telecommunication system and many more. For more details on application of D-BMAP, the readers are alluded to Klemm et al. (2003), Hofkens et al. (2004), Choi et al. (2015), Vuyst et al. (2009), Turck et al. (2007), Turck et al. (2013), and Zhao et al. (2004). Moreover, the discrete-time queueing system has far better efficiency than continuous ones in modelling the digital transmission systems. Hence, during the last few decades, the researchers and experts are gaining more eagerness to investigate the discrete-time queues due to its prevalent applications to the slotted digital computer, communication system and telecommunication networks based on the cellular network (4G/5G), wireless network, pattern recognition, reliability engineering, ecological studies, automation, geological studies, health studies, communications and electrical engineering. More information about these applications can be found in Mushtaq et al. (2016), Alfa (2016), Dai et al. (1991), Platis et al. (1996), Gyasi-Agyei (2001), Mullubhatla and Pattipati (2000), Norberg et al. (2002), McDonnell et al. (2002), Bruneel and Kim (1993), and San-Qi and Hong-Dah (1994).