No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
The Performance Impact of Buffer Sizes for Multi-Path TCP in Internet Setups
Abstract
The Multi-Path Transmission Control Protocol (MPTCP) is the new concurrent multi-path transfer extension for the widely-deployed Transmission Control Protocol (TCP). Of course, having multiple and possibly highly dissimilar paths for transmission is a challenge for the management of the send and receive buffers, since optimal throughput is desired with a reasonable allocation of the limited memory resources in MPTCP endpoints. This is particularly important when many MPTCP connections have to be handled simultaneously. This paper measures out the required MPTCP buffer size in the real-world Internet testbed NorNet, comparing theoretical size and real size to analyse MPTCP performance. The experiment shows that multi-path transmission can effectively increase the application payload throughput, and greatly improve the robustness of the data transmission. As an important point of this paper, we can show that appropriate buffer size settings can increase the payload throughput, while not wasting resources. This paper has certain significance for further accurately determining the optimal buffer size settings for multi-path transmission in large-scale Internet setups.
... Effectively utilizing the redundant network resources and improving the network performance -despite the very dissimilar characteristics of the underlying networks -have become a hot issue in recent years. Therefore, the Internet Engineering Task Force (IETF) puts forward the Multi-Path Transmission Control Protocol (MPTCP) [1]- [4], which can aggregate the bandwidths, improve the throughput, as well as enhance the robustness and fast recovery by using the existing network infrastructure in the way of software. ...
... While TCP only establishes a single-path connection for the communication between two hosts, MPTCP can dynamically conduct a multi-path connection consisting of multiple subflows. Subfigure 1(a) illustrates the MPTCP protocol stack [4]. It can be seen that MPTCP provides transparent end-to-end concurrent data transmission services to achieve the purpose of aggregating bandwidth and improving transport performance. ...
... • Ubuntu Linux 16.04 "Xenial Xerus" LTS with Linux kernel version 4.19.128, • Linux MPTCP version 0.95, • Buffer size limit set to 16 MiB, to prevent throughput limitations by lack of buffer space [4]. ...
... MPTCP needs to mainly carry out two tasks [11], [12], as depicted in Figure II-A: one is Packet Scheduling (PS), which is linked with data scheduling, interfacing with the subflows, and congestion control. Another is Path Management (PM) [13], which manages the communication paths, i.e. subflows between source/destination IP address pairs. ...
... leaving the receive buffer). [11] explains that, in order to cover the worst-case retransmission situation, a buffer size B is needed: ...
... For all measurements, we used the following configuration: (a) Linux kernel version 4.4.84, (b) Linux MPTCP [35] version 0.92.1 6 using the "fullmesh" path manager (to use all possible paths [13], [36] Table I THE NORNET CORE TESTBED SITES USED FOR THE MEASUREMENTS • Initial path selection [13], and • Buffer size settings [11]. We designed a Python script that generates data for statistical analysis by measuring throughput at various values of network parameters listed above. ...
... There are several studies on the performance evaluation of MPTCP through heterogeneous paths [5][6][7][8][9][10][11]. Although they give new findings and proposals, they have some problems. ...
... There are some recent studies on MPTCP performance [10,11]. J. Kim et al. [10] proposes a scheduler using the buffer blocking prediction based on receive buffer size and RTT, but they do not care about the send socket buffer size. ...
... J. Kim et al. [10] proposes a scheduler using the buffer blocking prediction based on receive buffer size and RTT, but they do not care about the send socket buffer size. F. Zhou et al. [11] shows the MPTCP performance evaluation over the real Internet by changing socket buffer size, but they assume that the sizes of send and receive socket buffers are the same. ...
Recently, the multipath transport protocol such as Multipath TCP becomes increasingly important. It allows more than one TCP connections via different paths to compose one Multipath TCP communication. However, it has some problems when those paths have different delay. Especially, the limited buffer space at either sender or receiver may degrade the throughput due to head-of-line blocking. Our previous paper pointed out that insufficient send socket buffer and receive socket buffer provide different situations of performance degradation, and that insufficient send socket buffer gives poorer throughput. This paper extends the performance analysis of our previous paper in the conditions with various combinations of send socket buffer size and transmission delay. It gives more detailed analysis using Multipath TCP level sequence number and congestion window size, and suggests the reasons for performance degradation.
... MPTCP [1,2] is an extension of the well-known Transmission Control Protocol (TCP) [3]. It realizes multi-path transmission on the transport layer. ...
... However, MPTCP can dynamically conduct a multi-path connection consisting of multiple flows. Figure 1a illustrates the MPTCP protocol stack [2,4]. MPTCP transfers data simultaneously over different subflows, as depicted in Figure 1a. ...
The transmission performance of a multi-path transmission control protocol (MPTCP) is affected by many parameters, such as path management, congestion control, buffer size, and subflow bandwidth. Most of the previous studies have focused almost exclusively on the improvement of a single parameter, without a holistic view. In this paper, a multi-parameter comprehensive optimized algorithm (MPCOA) is proposed to comprehensively optimize the above parameters. The MPCOA algorithm can find a smaller buffer size and select an appropriate congestion control and path management algorithm on the premise of ensuring better throughput. Experiments in three scenarios show that MPCOA can save buffer space and subflow resources, and achieve a high throughput. Meanwhile, a set of quantitative improvement results given by MPCOA is convenient for us to evaluate the quality of the MPTCP network and provide reference for our ongoing future work.
... MPTCP uses multiple subflows to implement concurrent multipath transport [42]. Each subflow is defined by a source/destination IP address pair and appears as a regular TCP connection. ...
... In the traditional Internet, timer-based retransmissions should be rare; however, they must be considered due to the characteristics of MANETs. Hence, to utilize a network path and cover fast retransmission and timer-based retransmission, the send/receive buffer size constraint B is shown in the following formula [42]. In the worst case, it takes three times the highest subflow RTT (first transmission, fast retransmission, timer-based retransmission) plus the highest subflow RTO (retransmission timeout). ...
In some special circumstances (e.g., tsunamis, battlefields, and earthquakes), communication infrastructures are damaged or nonexistent. For communication among people, mobile smart devices (MSDs) can be used to construct mobile ad hoc networks (MANETs). This paper focuses on the problem of data delivery in MANETs aiming to improve the quality of service (QoS) and quality of experience (QoE) users receive. MANETs, however, have the well-known problems of frequent disconnections and high rates of failed transmissions as MSDs move in and out of network coverage areas, and the topology constantly changes. To solve these issues, the main contributions of this work are as follows: (1) we provide and investigate the QoE-driven multipath TCP (MPTCP)-based data delivery model in MANETs; (2) we present hidden Markov model-based optimal-start multipath routing, which can effectively predict a mobile node’s near future network connection state according to its past connection state; (3) we leverage MPTCP to simultaneously transmit data via multiple interfaces of MSDs and improve the establishment method for MPTCP subpaths; and (4) we study and improve the algorithm of multihop routing in MANETs. The test results show that our algorithms can offer more efficient use of multiple subpaths and better network traffic load balancing than using standard MPTCP alone.
... Furthermore, whenever bottleneck buffer is not properly adjusted, we can encounter the effect of "bufferbloat". Although it is an inter-related issue with regard to the receiver buffer as already addressed in [12], it was extensively analysed in [13] for wireless, and in [14] for terrestrial networks, where the traffic on links with high propagation delay experienced an RTT exceeding 7 s. In presence of such a large amount of packets in flight, there is a MP-TCP layer performance reduction in terms of aggregated throughput. ...
... In fact assuming a strict configuration may reduce the flexibility and the feasibility of the network configuration, and it is not suitable when the link characteristic changes. Such limit is set using Eq. 2, inferred from [14], where that value was proposed as size for the receiver buffer. ...
... While there are various articles on all aspects of the multipath transport protocols, e.g. congestion control [8]- [10], path management [11], buffer management [12], [13], shared bottleneck detection [14], handling of latency-sensitive traffic [15] and media streaming [16], [17], little work has been done to examine the properties of the underlying networks. In many cases, our Open Source tool NETPERFMETER 2 [18], [19] [1, Section 6.3] has been used to conduct Transport Layer performance measurements. ...
... Moreover, both TCP and MPTCP have more stable throughput performance when there is enough cache space. The buffer required to achieve stable maximum network throughput in MPTCP is close to 3 times that of TCP cache [9], which consumes a lot of cache space. Actually the buffer pool of the commercial Ethernet switch is usually small. ...
The multi-home host will become increasingly popular within IPv6, especially in data center Fat-tree networks. MPTCP protocol is designed to transmit over multipath load, providing users with transparent multi-path utilization and high data transmission rate. As the characteristics of Many-to-one pattern and a large number of short bursts, all of flows collide on the shared link and same output port, creating an inevitable bottleneck collapse during transmission process. In order to avoid the collapse of throughput and improve the throughput of the receiving end, the MPTCP protocol in the data center network should have the characteristics of minimal packet loss and rapid convergence. In this paper, we focus on throughput collapse avoidance and design a MPTCP Transmission Control method based Queuing Cache Balance Factor called QCBF in Multi-homed FatTree topology. This method can realize the sub-flows are allocated by the size of ToRs cluster cache. We first propose a buffer pool balance factor and build a ToRs cluster cache balance queue. Then according to ToRs cluster cache allocation to calculate a sub-flow congestion window value. QCBF not only avoids the MPTCP Incast problem effectively, but also makes full use of the aggregate bandwidth. Based the software NS3, the experiment results of experiment show the QCBF transmission control method could avoid loss of packet, and achieve the load balance of MPTCP data transmission.
... The severer the network load dynamism is, the larger it should be. The work in [43] adopts 3 in the Internet context. The performance of such a solution will be evaluated later in Section V-D2. ...
Bandwidth guarantee is a critical feature to enable performance predictability in cloud datacenters. This process is expected to achieve three requirements: work conservation, fairness, and simplicity. However, the distributed nature of datacenters raises significant challenges to attaining those requirements at the same time. In this paper, we propose an efficient approach that can satisfy the three requirements simultaneously. Our scheme takes advantage of multipath TCP (MPTCP) to generate explicit bandwidth guarantee (BG) traffic and work conservation (WC) traffic. We further prioritize the BG traffic over the WC traffic in the network fabric. Due to the priority setting, work conservation cannot harm bandwidth guarantees and thus is effectively supported. We show that MPTCP fits this direction well but presents some new issues when the WC subflows own a low priority. We thus adapt MPTCP to handle these issues through a customized scheduler (which strictly prioritizes BG subflow during packet scheduling) and adopting a large receive buffer. In addition, we enable tenants to share unused bandwidth fairly by managing the overall aggressiveness ofWCtraffic. The proposed system can be easily implemented with commercial off-the-shelf servers and switches. We have implemented with Linux kernel MPTCP for experiments. Extensive experiments in a small cluster (including one MapReduce experiment) and tracedriven simulations show that our scheme achieves the design goals effectively.
... Therefore, another milestone is Multi-Path TCP (MPTCP) [16,24], providing transport redundancy by multi-homing without changing the existing network infrastructure. It also supports load balancing over heterogeneous networks connections [15,17,27,30]. But MPTCP is not only useful when an endpoint is connected to multiple Internet connections. ...
Today, a steadily increasing number of users are not just passively consuming Internet content, but also share and publish content. Users publish text, photos and videos. With the availability of 5G high-speed, low-latency mobile broadband networks, real-time video streaming will also be possible. We believe this will become a very popular application in the coming years. But the more popular a service is, the higher the need for resilience. In this paper, we introduce our work-in-progress live video streaming platform for future mobile edge computing scenarios, which makes use of MPTCP+IPv6 to support multi-homing for resilience and multi-path transport for load balancing. As a proof of concept, we will show that the platform is (1) compatible with IPv6, (2) utilizes load balancing when possible and (3) provides robustness by network redundancy.
... (a) Linux kernel version 4.1.27, (b) Linux MPTCP [15] version 0.91 2 , (c) ndiffports=2 (only relevant for "ndiffports" path manager), (d) Explicit Congestion Notification (ECN) support [40], [41] enabled, and (e) TCP (and MPTCP) buffer size limit [42] set to 16 MiB 3 (in order to avoid effects caused by buffer space scarcity). ...
With the rapid development of Internet communications, there is a growing demand to support devices being connected to multiple Internet service providers simultaneously. For example, every modern smartphone already provides at least mobile broadband (UMTS, LTE) as well as Wi-Fi interfaces. This multi-homing property can be used for resilience, but there is also an increasing interest in making use of concurrent multi-path transport. That is, multiple network paths can be utilised simultaneously, in order to improve the payload throughput for applications like big data or cloud computing. In this paper, we examine the performance of multi-path transport in real-world Internet setups, based on Multi-Path TCP (MPTCP) in the NorNet testbed for multi-homed systems. However, systems in such challenging setups need proper configuration. Therefore, we particularly would like to highlight the performance impact of different path management and congestion control settings in such realistic scenarios.
This paper presents a usage-based insurance (UBI) platform that incorporates Internet of Vehicles (IoV) and blockchain technologies, discussing the potential stakeholders, business models, and interaction modes involved in this platform. Existing UBI products mostly use data on the driver’s mileage, driving period, or driving region for more accurate insurance calculations. Automobile UBI encourages customers to continue improving their ability to drive safety and provides a means to smoothly, transparently, and rationally calculate insurance pricing and payout. This paper proposes blockchain architecture to remedy management problems in a UBI environment. A bidding mechanism suitable for the blockchain-based UBI platform was designed to close the information gap between the insurance company and consumer, thus increasing consumer trust in the platform.
Deep learning technology is widely used in medicine. The automation of medical image classification and segmentation is essential and inevitable. This study proposes a transfer learning–based kidney segmentation model with an encoder–decoder architecture. Transfer learning was introduced through the utilization of the parameters from other organ segmentation models as the initial input parameters. The results indicated that the transfer learning–based method outperforms the single-organ segmentation model. Experiments with different encoders, such as ResNet-50 and VGG-16, were implemented under the same Unet structure. The proposed method using transfer learning under the ResNet-50 encoder achieved the best Dice score of 0.9689. The proposed model’s use of two public data sets from online competitions means that it requires fewer computing resources. The difference in Dice scores between our model and 3D Unet (Isensee) was less than 1%. The average difference between the estimated kidney volume and the ground truth was only 1.4%, reflecting a seven times higher accuracy than that of conventional kidney volume estimation in clinical medicine.
Multipath Transmission Control Protocol (MPTCP) is in a rapid development process. MPTCP allows a network node to utilize multiple network interfaces and IP paths at the same time. It can take full advantage of network resources and provide reliable transmission, which brings advantages to users in terms of performance and reliability. In order to study the performance of MPTCP under high load, this paper uses Mininet to create an SDN environment and compare the performance differences between MPTCP and TCP under high load, and simulates a common Web application network architecture. The performance of MPTCP under this architecture is tested and analyzed. The test results show that MPTCP performs better than traditional TCP under high load. Besides, its performance can be optimized furtherly by adjusting related parameters.
This document collects some experiences of Multi-Path TCP (MPTCP) evaluations in the NorNet testbed.
NorNet is an open, international Internet testbed platform for research on multi-homed systems. Multi-homed systems have the property of being connected to multiple Internet Service Providers (ISP) simultaneously, in order to still provide connectivity in case of ISP/network failures. Basis of NorNet is Linux, together with other Open Source software. At the moment, the testbed infrastructure spreads over 21~sites on 4~continents. NorNet makes extensive use of advanced Linux features like virtualisation, file system features, routing rules, SCTP, MPTCP, and more. The global distribution creates further challenges. Goal of this talk is therefore to provide an overview of the problems that occurred when building the testbed, as well as solutions and lessons learned from solving these challenges. The idea is to present guidelines for utilising the advanced Linux features in own projects.
Multi-homing denotes the simultaneous connection of endpoints (e.g. cloud servers, smartphones, etc.) to multiple Internet Service Providers (ISP). That is, the endpoints remain reachable even when some of the ISPs have problems (e.g. malfunction of hardware or break of cables). Besides the redundancy aspect, multi-homing can also make load sharing by multi-path transport possible, i.e. increasing the application throughput by utilising multiple paths simultaneously. Multi-path transport can e.g. be realised by Concurrent Multi-Path Transfer for SCTP (CMT-SCTP) and Multi-Path TCP (MPTCP), two protocols that are currently under standardisation in the IETF. The growing need for and deployment of multi-homed applications makes large-scale testing and evaluation in realistic Internet setups necessary. For instance, different paths can have very different characteristics with regard to bandwidth, packet loss rate, congestion, delay and jitter. Therefore, the NorNet project of the Simula Research Laboratory is building up an open platform for such experiments: the NorNet testbed. It provides programmable nodes with multiple ISP connections -- wired as well as wireless -- that are distributed all over Norway as well as some international locations. This talk will give an overview over NorNet.
Multi-path transport has become a hot topic in Internet protocol research with the evolution of emerging technologies, particularly with the market penetration of access terminals having multiple network interfaces (e.g. smartphones with LTE/UMTS and Wi-Fi interfaces). Multi-Path TCP (MPTCP) is an extension of TCP that allows a connection to create several subflows for utilizing multiple network paths. Using multiple end-to-end TCP connections as subflows, MPTCP distributes data to different subflows over multiple ISPs, so as to enhance network robustness and improve throughput.
This paper first presents MPTCP’s architecture and multi-path congestion control algorithm concepts. Then, it examines three test scenarios in the NorNet testbed, particularly highlighting the performance difference between using uncoupled and coupled congestion controls in multi-homed, real-world Internet setups. The results show that MPTCP with coupled CCs gets more benefits than TCP and demonstrates the lower aggressiveness in comparison to MPTCP with uncoupled CCs.
NorNet Core is the world's first, open, large-scale Internet testbed for multi-homed systems and applications. Particularly, it is currently used for research on topics like multi-path transport and resilience. Researchers can run experiments on distributed, programmable nodes that are distributed over various locations and providing access to multiple different Internet service providers (ISP) with different access technologies. Clearly, a key feature of this testbed is to work in the real-world Internet. That is, it is especially desired to expose experiments to real Internet behaviour like background traffic. However, for the researcher, it is necessary to actually know how paths -- being used for an experiment -- actually behave: Are the paths actually working? How are the round-trip times among sites over different ISPs, etc.. How did the behaviour change over time? To provide such information to the researchers, we have designed and developed a maintenance and monitoring infrastructure for the NorNet Core testbed. In this paper, we will first introduce this infrastructure. Furthermore, we will demonstrate its usefulness with some useful, real-world examples. Our infrastructure has now become part of the testbed, and it is therefore available for all users of NorNet Core as well.
Nowadays, cloud applications are becoming more and more popular. However, in order for such applications to work, they need a stable Internet connectivity. To avoid the Internet access becoming a single point of failure, redundancy by multi-homing -- i.e. simultaneous access to multiple Internet service providers (ISP) -- is becoming increasingly common as well. Multi-homing leads to the desire to utilise all network attachment points simultaneously, which is e.g. provided by the Multi-Path TCP (MPTCP) extension for TCP. MPTCP is still under development by researchers and standardisation in the IETF. Particularly, it is necessary to evaluate MPTCP under realistic Internet conditions. NorNet Core is the world's first, large-scale Internet testbed for multi-homed systems and applications. It is therefore a useful platform for evaluating MPTCP. In this paper, we therefore present our NorNet Core extension that adds MPTCP support to the testbed. Particularly, our extension is now available to all users of NorNet Core as well, which significantly reduces the effort of MPTCP researchers to evaluate MPTCP and its improvements. In a proof of concept, we furthermore show the strengths and weaknesses of state-of-the-art MPTCP in NorNet Core, in a configuration covering 29 ISP connections at 14 sites in 5 different countries.
The market penetration of access devices with multiple network interfaces has increased dramatically over the last few years. As a consequence, there is a strong interest to use all of the available interfaces concurrently to improve data throughput. Corresponding extensions of established Transport protocols are receiving considerable attention within research and standardization. Currently two approaches are in the focus of the IETF: The Multipath TCP (MPTCP) extension for TCP and the Concurrent Multipath Transfer extension for SCTP (CMT-SCTP). This paper evaluates and compares implementations of these two loadsharing protocols by using both lab measurements and intercontinental testbed realized via the Internet between Europe and China. The experiments show that some performance critical aspects have not been taken into account in previous studies. Furthermore, they show that the simple scenario with two disjoint paths, which is typically used for evaluation, does not sufficiently cover the real Internet environment. Based on these insights, we highlight that the different path management strategies of the two protocols have a significant impact on their performance in real Internet scenarios.
In just about 4 years, IPv6 will celebrate its 20th anniversary. While the protocol itself is already quite old, its deployment has only recently picked up speed. Not so many Internet service providers offer direct IPv6 connectivity to their customers, yet. Clearly, when IPv6 is available to customers, they expect that IPv6 offers at least the same -- or even better -- stability of connections in comparison to IPv4. The main goal of this paper is to investigate whether this is true today. In our paper, we present up-to-date measurement results on the stability of IPv4 and IPv6 paths in the real Internet, based on machines that are distributed over a large geographical area, as part of the NorNet Core testbed infrastructure for multi-homed systems. The measurements not only cover high-speed research networks, but also consumer-grade ADSL connections -- i.e. the ISP connection types of "normal" end-users -- as well as a broad range of different ISPs. The measurements show that IPv6 paths are less stable than corresponding IPv4 paths. We also find that the use of load balancing is more prevalent in IPv6 than in IPv4.
Multi-path transfer protocols such as Concurrent Multi-Path Transfer for SCTP and Multi-Path TCP (MPTCP), are becoming increasingly popular due to widespread deployment of smartphones with multi-homing support. Although the idea of using multiple interfaces simultaneously to improve application throughput is tempting, does transmission over multiple interfaces always provide benefits especially in realistic setup? In this paper, we first show that multi-path transfer might actually have a negative impact in real-world scenarios with mobile broadband and wireless LAN networks. We then introduce our Dynamic Relative Path Scoring (DRePaS) algorithm that continuously evaluates the contribution of paths to the overall performance and dynamically influences the scheduling decisions to make best use of the paths for the overall system performance. We show that DRePaS outperforms the current MPTCP implementation in terms of throughput and application delay, especially when the links are heterogeneous.
Today, most of the smart phones are equipped with two network interfaces: Mobile Broadband (MBB) and Wireless Local Area Network (WLAN). Multi-path transport protocols provide increased throughput or reliability, by utilizing these interfaces simultaneously. However, multi-path transmission over networks with very different QoS characteristics is a challenge. In this paper, we studied Multi-Path TCP (MPTCP) in heterogeneous networks, specifically MBB networks and WLAN. We first investigate the effect of bufferbloat in MBB on MPTCP performance. Then, we propose a bufferbloat mitigation algorithm: Multi-Path Transport Bufferbloat Mitigation (MPT-BM). Using our algorithm, we conduct experiments in real operational networks. The experimental results show that MPT-BM outperforms the current MPTCP implementation by increasing the application goodput quality and decreasing MPTCP's buffer delay, jitter and buffer space requirements.
Networks have become multipath: mobile devices have multiple radio interfaces, datacenters have redundant paths andmultihoming is the normfor big server farms. Meanwhile, TCP is still only single-path. Is it possible to extend TCP to enable it to support multiple paths for current applications on today's Internet? The answer is positive. We carefully review the constraints--partly due to various types of middleboxes-- that influenced the design of Multipath TCP and show how we handled them to achieve its deployability goals. We report our experience in implementing Multipath TCP in the Linux kernel and we evaluate its performance. Our measurements focus on the algorithms needed to efficiently use paths with different characteristics, notably send and receive buffer tuning and segment reordering. We also compare the performance of our implementation with regular TCP on web servers. Finally, we discuss the lessons learned from designing MPTCP.
The Internet has made it possible to communicate and to use services over large geographical distances. While it has originally been built for less critical services like e-mail and file transfer, it is nowadays also increasingly often used for availability-critical services like e.g. e-commerce or healthcare. Clearly, the reach ability of such services must be ensured by so-called multi-homing of endpoints. That is, endpoints are simultaneously connected to multiple Internet Service Providers (ISP) to provide redundancy. If one ISP has problems, it is intended that the connection to another one still works. However, such assumptions have never been verified in real, large-scale setups. The intention of the Nor Net project is to build up a realistic Internet test bed for multi-homing. In this paper, we describe the design of Nor Net with focus on the implementation of its fixed-line part: noun{Nor Net Core}. This paper is intended to give researchers an overview of its mode of operation, its capabilities as well as its interesting feature realisations. The knowledge about these items is very useful to plan own experiments in the Nor Net test bed.
The Stream Control Transmission Protocol (SCTP) as defined in RFC 4960 is an advanced Transport Layer protocol that provides support for multi-homing. That is, SCTP endpoints may simultaneously use multiple Network Layer addresses, which allows to connect the endpoints to multiple networks for redundancy purposes. However, for the transfer of user data, only one of the possible paths is currently used at a time. All other paths remain as backup and are only used for retransmissions.
Clearly, the existence of multiple paths has led to the idea of applying load sharing among the paths. An extension to SCTP -- denoted as Concurrent Multipath Transfer (CMT) -- realises this load sharing functionality. While this approach works well for similar paths, i.e. paths having similar characteristics regarding bandwidths, bit error rates and delays, the use of dissimilar paths does not work that neatly.
In this thesis, the issues of dissimilar paths for CMT-based load sharing will be demonstrated first. The reasons for these issues will be identified and solutions proposed. These solutions will be evaluated in simulations, as well as partially also in a real-world Internet testbed setup, in order to show their effectiveness. In particular, it will be shown that a combination of multiple mechanisms is necessary to make CMT work as expected under a wide range of network and system parameters.
Furthermore, the fairness of CMT-based transport -- in concurrency to classic non-CMT flows -- will be analysed. The usage of plain CMT leads to an overly aggressive bandwidth occupation on so-called shared bottlenecks. As a countermeasure, the idea of Resource Pooling will be utilised. For this purpose, two new and one adapted congestion control approach -- all based on the Resource Pooling principle -- will be introduced and examined in similar as well as dissimilar path setups, in order to show how to fairly deploy CMT transport in the Internet.
The results of this work have also been contributed to the ongoing IETF standardisation process of SCTP and its extensions.
With the strongly growing popularity of mobile devices like smartphones and tablet computers, the number of end-systems with more than one network access - like UMTS/LTE and WLAN - is also increasing. This so-called multi-homing also leads to the desire of utilising multiple network paths simultaneously, in order to improve application payload throughput. Clearly, this so-called multi-path transfer feature is also very useful for the transport of multimedia contents, particularly when a single network access alone is not fast enough to fulfil the bandwidth requirements of the application. In many cases, multimedia transport is also sensitive for delays and packet losses. However, the focus of the current multi-path transfer approaches has been on bandwidth only. In order to tackle this challenge, our paper introduces two new send strategies to map payload data to different wireless paths. Finally, by using measurements, we show that a significant performance improvement for delay and loss-sensitive applications can be achieved in comparison to the existing approaches.
We present Nornet Edge (NNE), a dedicated infrastructure for measurements and experimentation in mobile broadband networks. NNE is unprecedented in size, consisting of more than 400 measurement nodes geographically distributed all over Norway. Each measurement node is a Linux-based embedded computer, and is connected to multiple mobile broadband providers. In addition, NNE includes an extensive backend system for deploying and managing experiments and collecting data. NNE makes it possible to run long-term measurement experiments to assess and compare quality and performance across different network operators on a national scale. Particular focus is put on allowing experiments to run in parallel on multiple network connections, and on collecting rich context information related to the experiments. In this paper we give a detailed presentation of NNE, and describe three different measurement experiments that illustrate how the infrastructure can be used. We also provide a roadmap for further development of NNE.
Over the last decade, the Internet has grown at a tremendous speed in both size and complexity. Nowadays, a large number of important services – for instance e-commerce, healthcare and many others – depend on the availability of the underlying network. Clearly, service interruptions due to network problems may have a severe impact. On the long way towards the Future Internet, the complexity will grow even further. Therefore, new ideas and concepts must be evaluated thoroughly, and particularly in realistic, real-world Internet scenarios, before they can be deployed for production networks. For this purpose, various testbeds – for instance PlanetLab, GpENI or G-Lab – have been established and are intensively used for research. However, all of these testbeds lack the support for so-called multi-homing.
Multi-homing denotes the connection of a site to multiple Internet service providers, in order to achieve redundancy. Clearly, with the need for network availability, there is a steadily growing demand for multi-homing. The idea of the NorNet Core project is to establish a Future Internet research testbed with multi-homed sites, in order to allow researchers to perform experiments with multi-homed systems. Particular use cases for this testbed include realistic experiments in the areas of multi-path routing, load balancing, multi-path transport protocols, overlay networks and network resilience. In this paper, we introduce the NorNet Core testbed as well as its architecture.
In both TCP and SCTP, selectively acked (SACKed) out-of-order data is implicitly renegable; that is, the receiver can later discard SACKed data. The possibility of reneging forces the transport sender to maintain copies of SACKed data in the send buffer until they are cumulatively acked. In this paper, we investigate the situation where all out-of-order data is non-renegable, such as when the data has been delivered to the application, or when the receiver simply never reneges. Using simulations, we show that SACKs result in inevitable send buffer wastage, which increases as frequency of loss events and loss recovery durations increase. We introduce a fundamentally new ack mechanism, Non-Renegable Selective Acknowledgments (NR-SACKs), for SCTP. Using NR-SACKs, an SCTP receiver can explicitly identify some or all out-of-order data as being non-renegable, allowing the sender to free up send buffer sooner than if the data were only SACKed. Simulation comparisons show that NR-SACKs enable efficient utilization of a transport senderpsilas memory. Further investigations show that NR-SACKs also improve throughput in Concurrent Multipath Transfer (CMT) [4].
Multi-homed Internet sites become more and more widespread, due to the rising dispersal of inexpensive Internet access technologies combined with the growing deployment of resilience-critical applications. Concurrent Multipath Transfer (CMT) denotes the Transport Layer approach to utilise multiple network paths simultaneously, in order to improve application payload throughput. Currently, CMT is a quite hot topic in the IETF - in form of the Multipath TCP (MPTCP) and CMT-SCTP protocol extensions for TCP and SCTP. However, an important issue is still not fully solved: multipath congestion control. In order to support the IETF activities, we have set up a distributed Internet testbed for CMT evaluation. An important tool - which we have developed for multiprotocol Transport Layer performance analysis - is the Open Source NetPerfMeter tool-chain. It supports the parametrisation and processing of measurement runs as well as results collection, post-processing and plotting. However, its key feature is to support multiple Transport Layer protocols, which makes a quantitative comparison of different protocols - including state-of-the-art features like CMT - possible. In this paper, we first introduce NetPerfMeter and then show a proof-of-concept performance evaluation of CMT congestion controls which are currently discussed in the IETF standardisation process of CMT-SCTP.
The Stream Control Transmission Protocol (SCTP) is a general-purpose transport layer protocol providing a service similar to TCP - plus a set of advanced features to utilize the enhanced capabilities of modern IP networks and to support increased application requirements. Nowadays, there are SCTP implementations for all major operating systems. While SCTP was standardized as an RFC several years ago, there is still significant ongoing work within the IETF to discuss and standardize further features in the form of protocol extensions. In this article, we first introduce the SCTP base protocol and already standardized extensions. After that, we focus on the ongoing SCTP standardization progress in the IETF and give an overview of activities and challenges in the areas of security and concurrent multipath transport.
With the deployment of more and more resilience-critical Internet applications, there is a rising demand for multi-homed network sites. This leads to the desire for simultaneously utilising all available access paths to improve application data throughput. This is commonly known as Concurrent Multipath Transfer (CMT); approaches for several Transport Layer protocols have been proposed. Combined with Resource Pooling (RP), CMT can also fairly coexist with concurrent non-CMT flows. Current approaches focus on symmetric paths (i.e. similar bandwidth, delay and error rate). However, asymmetric paths are much more likely - particularly for realistic Internet setups - and efficient CMT usage on such paths is therefore crucial. In this paper, we first show the challenges of plain as well as RP-aware CMT data transport over asymmetric paths. After that, we introduce mechanisms for efficient transport over such paths. Finally, we analyse the performance of our approaches by using simulations.
The steadily growing deployment of resilience- critical Internet services is leading to an increasing number of Multi-Homed network sites. Asymmetric Digital Subscriber Lines (ADSL) are an inexpensive way to add a secondary Internet access connection. With the development of Multi-Path Transport Layer protocols - like Multipath TCP (MPTCP) and the Stream Control Transmission Protocol (SCTP) furnished by a Concur- rent Multipath Transfer (CMT-SCTP) extension - there is also a strong interest in utilising all access connections simultaneously to improve the data throughput of the applications. However, combining network paths over ADSL with paths over other access technologies like fibre optic links implies highly dissimilar paths with significantly different bandwidths, delays and queuing behaviours. Efficient Multi-Path transport over such dissimilar paths is a challenging task for the new Transport Layer protocols under development. In this paper, we show the difficulties of Multi-Path transport in a real-world dissimilar path setup which consists of a high- speed fibre optic link and an ADSL connection. After that, we present an optimised buffer handling technique which solves the transport efficiency issues in this setup. Our optimisation is first analysed by simulations. Finally, we also show the usefulness of our approach by experimental evaluation in a real Multi-Homed Internet setup. 1234
Recently new data center topologies have been proposed that offer higher aggregate bandwidth and location independence by creating multiple paths in the core of the network. To effectively use this bandwidth requires ensuring different flows take different paths, which poses a challenge. Plainly put, there is a mismatch between single-path transport and the multitude of available network paths. We propose a natural evolution of data center transport from TCP to multipath TCP. We show that multipath TCP can effectively and seamlessly use available bandwidth, providing improved throughput and better fairness in these new topologies when compared to single path TCP and randomized flow-level load balancing. We also show that multipath TCP outperforms laggy centralized flow scheduling without needing centralized control or additional infrastructure.
This paper presents a new TCP variant, called CUBIC, for high-speed network environments. CUBIC is an enhanced version of BIC: it simplifies the BIC window control and improves its TCP-friendliness and RTT-fairness. The window growth function of CUBIC is governed by a cubic function in terms of the elapsed time since the last loss event. Our experience indicates that the cubic function provides a good stability and scalability. Furthermore, the real-time nature of the protocol keeps the window growth rate independent of RTT, which keeps the protocol TCP friendly under both short and long RTT paths. Index Terms—Congestion Control, High-Speed TCP, TCP Friendliness
With the rapid development of Internet communications, there is a growing demand to support devices being connected to multiple Internet service providers simultaneously. For example, every modern smartphone already provides at least mobile broadband (UMTS, LTE) as well as Wi-Fi interfaces. This multi-homing property can be used for resilience, but there is also an increasing interest in making use of concurrent multi-path transport. That is, multiple network paths can be utilised simultaneously, in order to improve the payload throughput for applications like big data or cloud computing. In this paper, we examine the performance of multi-path transport in real-world Internet setups, based on Multi-Path TCP (MPTCP) in the NorNet testbed for multi-homed systems. However, systems in such challenging setups need proper configuration. Therefore, we particularly would like to highlight the performance impact of different path management and congestion control settings in such realistic scenarios.
NetPerfMeter: A Network Performance Metering Tool
- T Dreibholz
T. Dreibholz, "NetPerfMeter: A Network Performance Metering Tool,"
Multipath TCP Blog, Sep. 2015.
Dynamic Reservation Data Scheduling Mechanism for MPTCP
- Qing Hu
- Zou Ran
- Liu Peng
Qing Hu and Zou Ran and Liu Peng, "Dynamic Reservation Data
Scheduling Mechanism for MPTCP," Journal of Chongqing University
of Posts and Telecommunications, vol. 25, no. 6, 2013.
The Benefits of using Explicit Congestion Notification (ECN), " IETF, Internet Draft draft-ietf-aqm-ecn-benefits-08
- G Fairhurst
- M Welzl
G. Fairhurst and M. Welzl, " The Benefits of using Explicit Congestion
Notification (ECN), " IETF, Internet Draft draft-ietf-aqm-ecn-benefits-08,
Nov. 2015.
Forward Prediction Data Scheduling Mechanism for MPTCP
- Q Hu
- R Zhou
- L Zhou
Q. Hu, R. Zhou, and L. Zhou, "Forward Prediction Data Scheduling
Mechanism for MPTCP," Application Research of Computers, vol. 30,
no. 2, pp. 560-561, Feb. 2013, ISSN 1001-3695.
Load Sharing for the Stream Control Transmission Protocol (SCTP), " IETF, Individual Submission, Internet Draft draft-tuexen-tsvwg-sctp-multipath-13
- P D Amer
- M Becke
- T Dreibholz
- N Ekiz
- J R Iyengar
- P Natarajan
- R R Stewart
- M Tüxen
P. D. Amer, M. Becke, T. Dreibholz, N. Ekiz, J. R. Iyengar, P. Natarajan,
R. R. Stewart, and M. Tüxen, " Load Sharing for the Stream Control
Transmission Protocol (SCTP), " IETF, Individual Submission, Internet
Draft draft-tuexen-tsvwg-sctp-multipath-13, Dec. 2016.
TCP Extensions for Multipath Operation with Multiple Addresses
- A Ford
- C Raiciu
- M Handley
- O Bonaventure
A. Ford, C. Raiciu, M. Handley, and O. Bonaventure, "TCP Extensions
for Multipath Operation with Multiple Addresses," IETF, RFC 6824,
Jan. 2013, ISSN 2070-1721.
Architectural Guidelines for Multipath TCP Development
- A Ford
- C Raiciu
- M Handley
- S Barré
- J R Iyengar
A. Ford, C. Raiciu, M. Handley, S. Barré, and J. R. Iyengar, "Architectural Guidelines for Multipath TCP Development," IETF, Informational
RFC 6182, Mar. 2011, ISSN 2070-1721.
Non-Renegable Selective Acknowledgements (NR-SACKs) for MPTCP IETF, Individual Submission, Internet Draft draft-deng- mptcp-nrsack-00
- Z Deng
Z. Deng, " Non-Renegable Selective Acknowledgements (NR-SACKs)
for MPTCP, " IETF, Individual Submission, Internet Draft draft-deng-
mptcp-nrsack-00, Dec. 2013.
Performance Analysis of MPTCP and CMT-SCTP Multi-Path Transport Protocols
- F Fu
- Z Xing
- Y Xiong
- H Adhari
- E P Rathgeb
F. Fu, Z. Xing, Y. Xiong, H. Adhari, and E. P. Rathgeb, "Performance
Analysis of MPTCP and CMT-SCTP Multi-Path Transport Protocols,"
Computer Engineering and Applications, vol. 49, no. 21, pp. 79-82,
Oct. 2013.
- M Allman
- V Paxson
- E Blanton
M. Allman, V. Paxson, and E. Blanton, " TCP Congestion Control, " IETF,
Standards Track RFC 5681, Sep. 2009, ISSN 2070-1721.
Multipath TCP Middlebox Behavior IETF, Individual Submission, Internet Draft draft-lopez-mptcp-middlebox-00
- E Lopez
E. Lopez, " Multipath TCP Middlebox Behavior, " IETF, Individual
Submission, Internet Draft draft-lopez-mptcp-middlebox-00, Nov. 2014.
- C Raiciu
- C Pluntke
- S Barré
- A Greenhalgh
- D Wischik
- M Handley
C. Raiciu, C. Pluntke, S. Barré, A. Greenhalgh, D. Wischik, and M. Handley, "Data Center Networking with Multipath TCP," in Proceedings
of the 9th ACM SIGCOMM Workshop on Hot Topics in Networks,
Monterey, California/U.S.A., Oct. 2010, pp. 1-6, ISBN 978-1-4503-0409-2.
IETF, Individual Submission, Internet Draft draft-lopez-mptcp-middlebox-00
- E Lopez
E. Lopez, "Multipath TCP Middlebox Behavior," IETF, Individual
Submission, Internet Draft draft-lopez-mptcp-middlebox-00, Nov. 2014.
TCP Congestion Control
- M Allman
- V Paxson
- E Blanton
M. Allman, V. Paxson, and E. Blanton, "TCP Congestion Control," IETF,
Standards Track RFC 5681, Sep. 2009, ISSN 2070-1721.
Load Sharing for the Stream Control Transmission Protocol (SCTP)
- P D Amer
- M Becke
- T Dreibholz
- N Ekiz
- J R Iyengar
- P Natarajan
- R R Stewart
- M Tüxen
P. D. Amer, M. Becke, T. Dreibholz, N. Ekiz, J. R. Iyengar, P. Natarajan,
R. R. Stewart, and M. Tüxen, "Load Sharing for the Stream Control
Transmission Protocol (SCTP)," IETF, Individual Submission, Internet
Draft draft-tuexen-tsvwg-sctp-multipath-13, Dec. 2016.
Non-Renegable Selective Acknowledgements (NR-SACKs) for MPTCP
- Z Deng
Z. Deng, "Non-Renegable Selective Acknowledgements (NR-SACKs)
for MPTCP," IETF, Individual Submission, Internet Draft draft-dengmptcp-nrsack-00, Dec. 2013.
The Benefits of using Explicit Congestion Notification (ECN)
- G Fairhurst
- M Welzl
G. Fairhurst and M. Welzl, "The Benefits of using Explicit Congestion
Notification (ECN)," IETF, Internet Draft draft-ietf-aqm-ecn-benefits-08,
Nov. 2015.