TCP SkyLine:数据中心网络高吞吐率传输

针对数据中心网络在"多对一"并发流量模式下,TCP（Transmission Control Protocol）及其现有改进方案在单轮数据传输和多轮数据传输下吞吐率低下问题,提出了一种通过数据包标记实现丢包快速发现和快速重传并动态调整拥塞窗口初始值的策略,称为TSL（TCP SkyLine）.TSL同时解决了传统TCP Incast问题和多轮数据传输下由遗留窗口引发的TCP Incast问题.实验表明,TSL在单轮数据传输和多轮数据传输下均能获得90%以上的带宽利用率.在10Gbps网络中,其支持的并发连接数与传统TCP和DCTCP相比分别提升了5倍和1倍,有效吞吐率分别提升了18倍和8.6倍;在1Gbps网路中,支持的并发连接数较传统TCP和DCTCP分别提升了5.8倍和1倍.     

A survey on TCP Incast in data center networks

In high bandwidth, low latency data center networks, when multiple data senders simultaneously communicate with a single receiver, namely in many‐to‐one communication pattern, the burst data overload the receiver's switch buffers, which leads to Transmission Control Protocol (TCP) throughput collapse. This is the so‐called TCP Incast problem. It has become a new hot research topic in recent years. Also, many proposals have been put forward from the aspects of multiple layers including link layer, transport layer, and application layer, etc., to mitigate TCP Incast. In this paper, an in‐depth survey on these proposals is given, and the principles, merits, and drawbacks of the main proposals for solving the TCP Incast issue are analyzed and summarized. Copyright © 2012 John Wiley & Sons, Ltd.     

Transport protocols for data center networks: a survey of issues,solutions and challenges

In recent years, data centers play an important role in academia and industry for supporting various services and applications. Compared with other IP networks, data center networks have some special features such as many-to-one communication pattern with high bandwidth, low latency, auto-scaling, shallow buffered switches and multi-rooted tree topology. Owing to these special features of data center networks, traditional TCP suffers from severe performance degradation. For improving the performance of TCP in data center networks, various solutions have been proposed in recent years. This article presents a comprehensive survey of existing transport layer solutions proposed for mitigating the problems of TCP in data center networks. The objective of this article is threefold: to discuss about the issues of TCP in data center networks; to introduce various transport layer solutions and finally to compare and discuss the challenges of existing solutions proposed for improving the performance of TCP in data center networks.     

TCP Incast问题的定量分析

TCP Incast问题是影响现代数据中心性能的典型问题。正确理解TCP Incast问题的解决方案,对提高数据中心的业务能力具有重要意义。通过数学模型,定量分析了数据块大小、并发数量及TCP最小重传超时（RTO_min）等因素对TCP Incast问题的影响,构建并验证了一个预测小数据块Incast问题发生的数学模型。     

On the transmission opportunity and TCP throughput in cognitive radio networks

The characteristics of cognitive radio networks have huge impacts on the end‐to‐end performance of the transmission control protocol (TCP) for secondary users. Thus, the existing TCP throughput expression, widely used in wired and wireless networks, is no longer suitable for cognitive radio networks. In this paper, we derive the transmission opportunity of secondary users, taking into account the dynamics of spectrum availability, the overhead and errors of spectrum sensing, as well as the interaction between TCP and lower layers. The amount of transmission opportunity is expressed in terms of effective data transmission time. On the basis of the analysis of the transmission opportunity, an expression of an effective TCP throughput is then derived. To evaluate this effective TCP throughput expression, two cross‐layer optimization problems are formulated as application examples to maximize the transport layer effective throughput and energy utility, respectively. Simulation results show that our analysis on transmission opportunity is accurate, and the derived effective TCP throughput expression is more precise than existing ones. Copyright © 2012 John Wiley & Sons, Ltd.     

Analysis and improvement of TCP performance in opportunistic networks

Opportunistic networks have attracted attention due to their inherent characteristics, such as long latency, low data rate, and intermittent connectivity. Extensive research has been conducted on opportunistic networks, including the architecture, and routing. However, few in the literature report the performance of TCP in opportunistic networks, especially in the case of Epidemic Routing. In this paper, we first evaluate the performance of TCP in opportunistic networks with Epidemic Routing. Our results show that the Epidemic Routing in opportunistic networks degrades the performance of TCP because multicopy data packets cause duplicate ACKs, and in turn reduce the transmission rate of TCP. Then an enhanced algorithm for TCP, named A-TCP/Reno is proposed to solve the above problem. A-TCP/Reno avoids the duplicate ACK problem caused by Epidemic Routing. The simulation results show that A-TCP/Reno outperforms the TCP/Reno in opportunistic networks with Epidemic Routing protocol.     

Cross-layer-based adaptive TCP algorithm for cloud computing services in 4G LTE-A relaying communication

Cloud computing provides various diverse services for users accessing big data through high data rate cellular networks, e.g., LTE-A, IEEE 802.11ac, etc. Although LTE-A supports very high data rate, multi-hop relaying, and cooperative transmission, LTE-A suffers from high interference, path loss, high mobility, etc. Additionally, the accesses of cloud computing services need the transport layer protocols (e.g., TCP, UDP, and streaming) for achieving end-to-end transmissions. Clearly, the transmission QoS is significantly degraded when the big data transmissions are done through the TCP protocol over a high interference LTE-A environment. The issue of providing high data rate and high reliability transmissions in cloud computing needs to be addressed completely. Thus, this paper proposes a cross-layer-based adaptive TCP algorithm to gather the LTE-A network states (e.g., AMC, CQI, relay link state, available bandwidth, etc.), and then feeds the state information back to the TCP sender for accurately executing the network congestion control of TCP. As a result, by using the accurate TCP congestion window (cwnd) under a high interference LTE-A, the number of timeouts and packet losses are significantly decreased. Numerical results demonstrate that the proposed approach outperforms the compared approaches in goodput and fairness, especially in high interference environment. Especially, the goodput of the proposed approach is 139.42 % higher than that of NewReno when the wireless loss increases up to 4 %. Furthermore, the throughput and the response functions are mathematically analyzed. The analysis results can justify the claims of the proposed approach.     

DATCP: deadline-aware TCP for the commoditized data centers

Many flows in data centers have deadlines and missing deadlines would hurt application performance such as affecting response quality in Web applications or delaying computing jobs in MapReduce-like systems.However,transmission control protocol(TCP) which is widely used in data centers now cannot provide deadline-aware transmission service.Service differentiation only distinguishes flows with different priority but is unable to guarantee completion time.In this paper,we propose a new protocol named deadline-aware TCP(DATCP) to provide deadline-aware transmission service for the commoditized data centers,which can be used as a flexible method for flow scheduling.DATCP combines flow urgency and importance to calculate precedence.Flow urgency is dynamically adjusted according to the gap between desired rate and actual throughput.Setting importance can avoid starving the important but no-urgent flows.Furthermore,a flow quenching method is presented which allows as many high precedence flows as possible to meet deadlines under heavy network load.By extensive simulations,the performance of DATCP was evaluated.Simulation results show that DATCP can make flows meet deadlines effectively.     

An overview of network virtualization and cloud network as a service

Cloud computing services delivery and consumption model is based on communication infrastructure (network). The network serves as a linkage between the end‐users consuming cloud services and the providers of data centers providing the cloud services. In addition, in large‐scale cloud data centers, tens of thousands of compute and storage nodes are connected by a data center network to deliver a single‐purpose cloud service. To this end, some questions could be raised, such as the following: How do network architectures affect cloud computing? How will network architecture evolve to support better cloud computing and cloud‐based service delivery? What is the network's role in reliability, performance, scalability, and security of cloud computing? Should the network be a dumb transport pipe or an intelligent stack that is cloud workload aware? This paper focuses on the networking aspect in cloud computing and shall provide insights to these questions. Researchers can use this paper to accelerate their research on devising mechanisms for the following: (i) provisioning cloud network as a service and (ii) engineering network of data centers. Copyright © 2014 John Wiley & Sons, Ltd.     

Storage area network extension solutions and their performance assessment

Several solutions are proposed to extend storage area networking solutions over distances of hundreds to thousands of kilometers. Native fibre channel or end-to-end fibre channel-based solutions can be offered over long distances using SONET-based networks. Protocols such as Internet SCSI (iSCSI), Internet fibre channel protocol (iFCP) and fibre channel over TCP/IP (FCIP) are being proposed to enable storage area networking solutions over networks that use IP as their transport protocol. Performance analysis of these solutions in terms of application throughput under variable network conditions of packet loss, bandwidth availability, extension distance, and TCP implementations in the sender and receiver is presented based on analytical modeling of different solutions.     

Optimal channel access for TCP performance improvement in cognitive radio networks

Cognitive radio (CR) is a promising technology to improve spectrum utilization. Most of previous work on CR networks concentrates on maximizing transmission rate in the physical layer. However, the end-to-end transmission control protocol (TCP) performance perceived by secondary users is also a very important factor in CR networks. In this paper, we propose a novel multi-channel access scheme in CR networks, where the channel access is based on the TCP throughput in the transport layer. Specifically, we formulate the channel access process in CR network as a restless bandit system. With this stochastic optimization formulation, the optimal channel access policy is indexable, meaning that the channels with highest indices should be selected to transmit TCP traffic. In addition, we exploit cross-layer design methodology to improve TCP throughput, where modulation and coding at the physical layer and frame size at the data-link layer are considered together with TCP throughput in the transport layer to improve TCP performance. Simulation results show the effectiveness of the proposed scheme.     

Optical interconnection reverse data vortex network: performance analysis

Multistage interconnection networks are employed in data centres to interchange information between the processors and memory elements. Data Vortex (DV) is a multistage minimum logic network which can be used in data centres. DV satisfies the requirements of the interconnection networks such as scalability and throughput. However, the latency is on higher side, and reduction in latency can lead to higher throughput. In the present paper, we describe the feasibility and performance analysis of DV architecture in reverse direction. The routing and the possible hardware model of the node switch have been discussed. We present the performance analysis of Reverse Data Vortex (RDV) architecture in terms of throughput, latency and latency distribution. A comparative study with DV on throughput, latency and latency distribution is also presented. The simulation result shows that the decrease in latency of RDV is about 50 % that of DV and this leads to an increase in injection rate of RDV to values more than two times that of DV.     

Seamless integration of cloud and fog networks

A way to merge cloud computing infrastructures with traditional or legacy network deployments, leveraging the best in both worlds and enabling a logically centralized control. A solution is proposed to extend existing cloud computing software stacks so they are able to manage networks outside the cloud computing infrastructure, the fog, by extending the internal, virtualized network segments. This is useful in a variety of use cases such as incremental legacy to cloud network migration, hybrid virtual/traditional networking, centralized control of existing networks, bare metal provisioning, and even offloading of advanced services from typical home gateways into the operator. Any organization can develop different ‘drivers’ to support new, specific networking equipment, not necessarily tied to a protocol or vendor, and leverage the fog network. Our conceptual solution is prototyped on top of OpenStack, including changes to the API, command‐line interface (CLI), and other mechanisms. Test results indicate that there are low penalties on latency and throughput, and provisioning times are reduced in comparison with similar maintenance operations on traditional computer networks. Copyright © 2016 John Wiley & Sons, Ltd.     

Dynamics of TCP traffic over ATM networks

Investigates the performance of transport control protocol (TCP) connections over ATM networks without ATM-level congestion control and compares it to the performance of TCP over packet-based networks. For simulations of congested networks, the effective throughput of TCP over ATM can be quite low when cells are dropped at the congested ATM switch. The low throughput is due to wasted bandwidth as the congested link transmits cells from “corrupted” packets, i.e., packets in which at least one cell is dropped by the switch. The authors investigate two packet-discard strategies that alleviate the effects of fragmentation. Partial packet discard, in which remaining cells are discarded after one cell has been dropped from a packet, somewhat improves throughput. They introduce early packet discard, a strategy in which the switch drops whole packets prior to buffer overflow. This mechanism prevents fragmentation and restores throughput to maximal levels     

TCP performance over mobile ad hoc networks: a quantitative study

In this paper, we study the performance of the transmission control protocol (TCP) over mobile ad‐hoc networks. We present a comprehensive set of simulation results and identify the key factors that impact TCP's performance over ad‐hoc networks. We use a variety of parameters including link failure detection latency, route computation latency, packet level route unavailability index, and flow level route unavailability index to capture the impact of mobility. We relate the impact of mobility on the different parameters to TCP's performance by studying the throughput, loss‐rate and retransmission timeout values at the TCP layer. We conclude from our results that existing approaches to improve TCP performance over mobile ad‐hoc networks have identified and hence focused only on a subset of the affecting factors. In the process, we identify a comprehensive set of factors influencing TCP performance. Finally, using the insights gained through the performance evaluations, we propose a framework called Atra consisting of three simple and easily implementable mechanisms at the MAC and routing layers to improve TCP's performance over ad‐hoc networks. We demonstrate that Atra improves on the throughput performance of a default protocol stack by 50%–100%. Copyright © 2004 John Wiley & Sons, Ltd.     

Hierarchical cache design for enhancing TCP over heterogeneous networks with wired and wireless links

TCP is a reliable transport protocol tuned to perform well in traditional networks made up of links with low bit-error rates. Networks with higher bit-error rates, such as those with wireless links and mobile hosts, violate many of the assumptions made by the transmission control protocol (TCP), causing degraded end-to-end performance. We propose a two-layer hierarchical cache architecture for enhancing TCP performance over heterogeneous networks with both wired and wireless links. A new network-layer protocol, called new snoop (NS), is designed. The main idea is to cache the unacknowledged packets at both the mobile switch center (MSC) and base station (BS), to form a two-layer cache hierarchy. If a packet is lost due to transmission errors in the wireless link, the BS takes the responsibility to recover the loss. When a handoff occurs, the packets cached at the MSC can help to minimize the latency of retransmissions due to temporal disconnection. NS can preserve the end-to-end TCP semantics and is compatible with existing TCP applications. Its implementation only requires code modification at the BS and MSC. Simulation results show that NS is significantly more robust in dealing with unreliable wireless links and handoffs as compared with the original snoop scheme, as well as some other existing TCP enhancements.     

Q-learning-based task offloading strategy for satellite edge computing

In this paper, we study the task offloading optimization problem in satellite edge computing environments to reduce the whole communication latency and energy consumption so as to enhance the offloading success rate. A three-tier machine learning framework consisting of collaborative edge devices, edge data centers, and cloud data centers has been proposed to ensure an efficient task execution. To accomplish this goal, we also propose a Q-learning-based reinforcement learning offloading strategy in which both the time-sensitive constraints and data requirements of the computation-intensive tasks are taken into account. It enables various types of tasks to select the most suitable satellite nodes for the computing deployment. Simulation results show that our algorithm outperforms other baseline algorithms in terms of latency, energy consumption, and successful execution efficiency.     

ATP: a reliable transport protocol for ad hoc networks

Existing works have approached the problem of reliable transport in ad hoc networks by proposing mechanisms to improve TCP's performance over such networks, In this paper, we show through detailed arguments and simulations that several of the design elements in TCP are fundamentally inappropriate for the unique characteristics of ad hoc networks. Given that ad hoc networks are typically stand-alone, we approach the problem of reliable transport from the perspective that it is justifiable to develop an entirely new transport protocol that is not a variant of TCP. Toward this end, we present a new reliable transport layer protocol for ad hoc networks called ATP (ad hoc transport protocol). We show through ns2-based simulations that ATP outperforms default TCP as well as TCP-ELFN and ATCP.     

Performance of TCP protocols in deep space communication networks

The communication requirements for space missions necessitate to address the problems due to deep space communication networks. In this letter, the effects of slow start algorithm, propagation delay and the link errors on the throughput performance of transport layer protocols are investigated in deep space communication networks. The objective of this letter is to demonstrate through experimental results that existing TCP protocols are far from satisfying the deep space communication requirements and point out the urgent need for new TCP solutions.     

RELD, RTT ECN Loss Differentiation to optimize the performance of transport protocols on wireless networks

One major yet unsolved problem in wired-cum-wireless networks is the classification of losses, which might result from wireless temporary interferences or from network congestion. The transport protocol response to losses should be different for these two cases. If the transmission uses existing protocols like TCP, the losses are always classified as congestion losses by sender, causing reduced throughput. In wired networks, ECN (Explicit Congestion Notification) can be used to control the congestion through active queue management such as RED (Random Early Detection). It can also be used to solve the transport protocol misreaction over wireless networks. This paper proposes a loss differentiation method (RELD), based on ECN signaling and RTT (Round Trip Time), and applied to TCPlike. TCPlike is one of the three current congestion controls present in the new transport protocol DCCP (Datagram Congestion Control Protocol). Our simulations, using a more realistic simulated loss error model for wireless networks, show that RELD optimizes congestion control and therefore increases the performance of transport protocols over wireless networks, leading to an average performance gain ranging from 10% to 15%.