Reducing traffic burstiness for MPTCP in data center networks

The number of mobile terminals that have both a cellular interface and a Wi-Fi interface has been growing in the recent years. Meanwhile, Multipath TCP (MPTCP) that enables the simultaneous use of multiple paths with multiple TCP flows while showing just one regular TCP connection with the application has been standardized by the IETF and has drawn more attention recently. One feature of MPTCP is that it can be realized only by operating system layers supporting multipath communication without any modifications of the application. However, it might be difficult to install MPTCP in all server operating systems. In this paper, we propose a method for applying a proxy that forwards TCP communication from a server to MPTCP communication on a mobile core network, and show the results of evaluation with the existing equipment.

Recent data centers provide dense inter-connectivity between each pair of servers through multiple paths. These data centers offer high aggregate bandwidth and robustness by using multiple paths simultaneously. Multipath TCP (MPTCP) protocol is developed for improving throughput, fairly sharing network link capacity and providing robustness during path failure by utilizing multiple paths over multi-homed data center networks. Running MPTCP protocol for latency-sensitive rack-local short flows with many-to-one communication pattern at the access layer of multi-homed data center networks creates MPTCP incast problem. In this paper, Balanced Multipath TCP (BMPTCP) protocol is proposed to mitigate MPTCP incast problem in multi-homed data center networks. BMPTCP is a window-based congestion control protocol that prevents constant growth of each worker’s subflow congestion window size. BMPTCP computes identical congestion window size for all concurrent subflows by considering bottleneck Top of Rack (ToR) switch buffer size and increasing count of concurrently transmitting workers. This helps BMPTCP to avoid timeout events due to full window loss at ToR switch. Based on current congestion situation at ToR switches, BMPTCP adjust transmission rates of each worker’s subflow so that total amount of data transmitted by all concurrent subflows does not overflow bottleneck ToR switch buffer. Simulation results show that BMPTCP effectively alleviates MPTCP incast. It improves goodput, reduces flow completion time as compared to existing MPTCP and EW-MPTCP protocols.

Today, there is a generalized standard usage of internet for all.The devices via multiple technologies that facilitates to provide few communication methods to scholars to work with. By forming multiple paths in the data center network, latest generation data centers offer maximum bandwidth with robustness. To utilize this bandwidth, it is necessary that different data flows take separate paths. In brief, a single-path transport seems inappropriate for such networks. By using Multipath TCP, we must reconsider data center networks, with a diverse approach as to the association between topology, transport protocols, routing. Multipath TCP allows certain topologies that single path TCP cannot use. In newer generation data centers, Multipath TCP is already deployable using extensively deployed technologies such as Equal-cost multipath routing. But, major benefits will come when data centers are specifically designed for multipath transports. Due to manifold of technologies like Cloud computing, social networking, and information networks there is a need to deploy the number of large data centers. While Transport Control Protocol is the leading Layer-3 transport protocol in data center networks, the operating conditions like high bandwidth, small-buffered switches, and traffic patterns causes poor performance of TCP. Data Center TCP algorithm has newly been anticipated as a TCP option for data centers which address these limitations. It is worth noting that traditional TCP protocol.

Multipath Transmission Control Protocol (MPTCP) is an approach towards high-throughput and efficient load balancing over multiple paths. Each of paths forms a TCP connection with an IP address, and those can be implemented as multiple network interfaces or multiple ports within a network interface. In this paper, we focus on the multiple network interfaces environment. Each network interface with an IP address is called as a subflow. A subflow is a TCP connection which can have a different internet path identified by IP addresses of source and destination network interfaces. To control these multiple subflows, MPTCP supports many options. Specifically, to establish a new subflow, MPTCP uses an <monospace>ADD&#x005F;ADDR</monospace> option. A host sends <monospace>ADD&#x005F;ADDR</monospace> option to inform another host of its IP address, and then, the host receiving <monospace>ADD&#x005F;ADDR</monospace> option tries to establish a subflow at the address of <monospace>ADD&#x005F;ADDR</monospace> option. However, by forging the <monospace>ADD&#x005F;ADDR</monospace> option, an attacker can create a fake subflow that passes through itself and eventually hijack the connection between both end hosts. In a previous study, Hash-based Message Authentication (HMAC) was added to the <monospace>ADD&#x005F;ADDR</monospace> option, preventing it from being forged. Nevertheless, since the keys for generating HMAC can be leaked during three-way handshake, a variant of the <monospace>ADD&#x005F;ADDR</monospace> attack called the persistent <monospace>ADD&#x005F;ADDR</monospace> attack can be possible. To this end, we propose a protocol that can prevent the <monospace>ADD&#x005F;ADDR</monospace> attacks by backward confirmation of the <monospace>ADD&#x005F;ADDR</monospace> option without encryption. The main idea of our proposal is to apply a digital signature scheme for the backward confirmation. We show security analysis for the proposed protocol and compare with the previous studies in terms of time/space overheads.

Data center networks (DCNs) are at the forefront of a radical transformation, fueled by the rapid rise of artificial intelligence (AI), cloud-native adoption, and enterprise digitalization, which together place unprecedented demands on their design and performance. From healthcare AI and intelligent transportation to business analytics, modern DCNs are expected to support a diverse and evolving mix of bandwidth-demanding applications, each with distinct quality of service (QoS) requirements. Yet, most state-of-the-art DCN solutions rely on oversimplified DCN data traffic models, failing to address the required application-level differentiation in terms of end-to-end (E2E) latency demands. In response, we propose ALAS-DCN, a novel, to our knowledge, application-aware, low-latency, scalable DCN architecture, based on optical packet switching in the intra-rack and inter-rack domains that is both designed and rigorously evaluated using commercial QoS specifications and realistic traffic patterns of different applications coexisting within the same DCN. ALAS-DCN integrates three hierarchical optical sub-networks: intra-rack, inter-rack, and bridge, each governed by a custom wavelength division multiplexing (WDM) medium access control (MAC) protocol that dynamically prioritizes critical traffic packets, according to the specific application E2E latency demands. Unlike prior approaches, our evaluation leverages realistic workload traces generated by established traffic generators, capturing the application diversity within modern data center (DC) environments. Extensive experimental simulations across five representative DC categories, each hosting a different mix of modern applications, demonstrate that ALAS-DCN consistently achieves up to 95% throughput efficiency and sub-µs latency for high priority traffic across a variety of applications, consistently meeting their differentiated latency constraints. Scalable up to 20,000 servers, ALAS-DCN redefines the DCN as not just a transport layer but an intelligent enabler of next-generation application performance.

With the development of the Internet of Things (IoT) and 5G, there are ubiquitous smart devices and network functions providing emerging network services efficiently and optimally through building many network connections based on WiFi, LTE/5G, Ethernet, and etc. The Multipath TCP (MPTCP) protocol that enables these devices to establish multiple paths for simultaneous data transmission, has been a widely used extension of standard TCP in smart devices and network functions. On the other hand, more heavy and time-varying traffic loads are generated in an MPTCP network, so that an efficient congestion control mechanism that schedules the traffic between multiple subflows and avoids congestion is highly required. In this paper, we propose a decentralized learning approach, DeepCC, to adapt to the volatile environments and realize the efficient congestion control. The Multi-Agent Deep Reinforcement Learning (MADRL) is used to learn a policy of congestion control for each subflow according to the real-time network states. To deal with the problem of the fixed state space and slow convergence, we adopt two self-attention mechanisms to receive the states and train the policy, respectively. Due to the asynchronous design of DeepCC, the learning process will not introduce extra delay and overhead on the decision-making process. Experiment results show that DeepCC consistently outperforms the well-known heuristic method and DRL-based MPTCP congestion control method in terms of goodput and jitter. Besides, DeepCC with the attention mechanism reduces convergence time by about 50% and increase goodput by about 80% compared with the commonly used structures of neural networks.

To meet the high performance requirement of real time and critical applications in industrial Internet of Things, modern lossless ethernet data center networks (DCNs) deployed with remote direct memory access and priority-based flow control (PFC) are dedicated to delivering low latency and high bandwidth. However, existing load balancing schemes either lack sub-round-trip-time congestion sensing or fail to accurately detect and reroute flows that cause congestion in PFC-enabled lossless DCNs. Therefore, we propose LBoDSN, an in-network load balancing for lossless DCNs using direct switch notification (DSN) for fast congestion flow detection, to address above challenges. LBoDSN tracks the evolution of ingress queue lengths at destination switches to anticipate the initiation of PFC pause, precisely identifies congested flows before PFC pause, and then sends DSNs to source switches to perform rerouting. After rerouting, the congestion notification packet associated with the previous path is selectively discarded to improve transmission performance. Experiments under realistic workloads reveal that, LBoDSN outperforms CONGA by 13%–65%, and 25%–80% in average and tail Flow Completion Times (FCTs), respectively. Compared to ConWeave, LBoDSN achieves approximately 9% improvement in both average and tail FCTs, while reducing switch queue consumption for reordering.

It is common to find multiple network interfaces connected to different Internet Service Providers (ISPs) in devices such as smartphones. Multipath TCP (MPTCP) enables TCP connections to use all these network interfaces in a single TCP connection in an application transparent manner. MPTCP schedules traffic of one TCP connection over subflows created over these network interfaces. It is evident that this requires some scheduling policy. There have been some attempts to allow applications to decide on the scheduling policy. However, this violates the application transparency of MPTCP, and applications do not have all the information required to decide on such a policy. In addition, this allows the applications to monopolize the network connection thus posing a security threat as well. We argue that only the owner of the device (the user) has the right to make that policy decision and only the user can make an informed decision on the scheduling policy. For example, the user has the information on the monetary cost of the connections through different interfaces. In this paper we present a mechanism that allows the user to provide hints to the TCP scheduler to alter its scheduling policy. While this is not a mechanism to implement generic scheduling policies, it demonstrates how a user can guide the scheduling policies. As a proof of the concept, we demonstrate how MPTCP scheduler can be influenced to select a less stable and lossy path over a stable path based on a user preference.

PTP: Path-specified transport protocol for concurrent multipath transmission in named data networks

Nearly 50% of the energy overhead in today's datacenters comes from host-to-host data transfers, which largely depend on the transport layer performance. Multipath TCP (MPTCP) has recently been suggested as a promising transport protocol to improve datacenter network throughput, yet it also increases the host CPU power consumption. It remains unclear whether datacenters can indeed benefit from using MPTCP from the perspective of energy efficiency. By analyzing the performance of MPTCP, we show that (1) despite consuming higher host CPU power than TCP, MPTCP can largely reduce the long flow completion time and thus save the aggregated energy; (2) link-sharing subflows in MPTCP not only has negative impact on both throughput-sensitive long flows and latency-sensitive short flows, but also noticeably increases the host CPU power, especially for short flows. We present MPTCP-D, an energy-efficient variant of multipath TCP for datacenters. MPTCP-D incorporates a novel congestion control algorithm that can provide energy efficiency by minimizing the flow completion time, and an extra subflow elimination mechanism that can preclude link-sharing subflows from increasing the host CPU power. We implement MPTCP-D in the Linux kernel and study its performance by experiments on Amazon EC2. Our results show that, without degrading the performance of the long flow throughput and short flow completion time, MPTCP-D reduces the long flow energy consumption by up to 72% compared to DCTCP for data transfers, and reduces the short flow power consumption by up to 46% compared to MPTCP with linksharing subflows.

Cloud computing is an emerging area where all the network services are provided in the form of pay per use. Utilizing the available resources efficiently is one of the challenge in the data center network. Cloud resources are given as virtual machine to the end users. Sometimes, Virtual Machines (VMs) are overloaded or underloaded. The overloaded VMs degrade the performance of applications and underloaded VMs lead to inefficient resource utilization. To effectively utilize the available resources, we have proposed Dynamic Agent-Based’Load Balancing (DA-LB) on Software-Defined Networking (SDN) using Back Propagation Artificial Neural Network (BPANN). This load balancing algorithm uses the global visibility of SDN to efficiently migrate virtual machines in the data center network. The results show that our SDN load balancing approach improves the overall network efficiency and works well on data migration.The proposed algorithm helps to increase the processing speed and predict the loaded VM in heavy load conditions and optimizes the resource utilization. The migration process results are compared with Multi-Path TCP (MPTCP) and Heuristic algorithm (HA) and it is seen that our proposed algorithm takes less migration time compared to MPTCP and HA.

MP-TCP (MultiPath-Transmission Control Protocol) was designed to use multiple interfaces in hosts; it uses several subflows (a set of regular TCP connections) to increase the performance compared to TCP. However, when a host has only one network interface but several subflows are created, they follow the same path to their destination generating bottlenecks and MP-TCP advantages are lost. This paper presents a solution to the shared-bottleneck problem using SDN (Software Defined Networking). A plugin for OpenDayLight is developed for implementing Dijkstra's Algorithm to find paths and assign them to new subflows avoiding bottlenecks. The solution supports concurrent MP-TCP connections. Results of tests obtained using the implemented plugin in several topologies are presented.

A survey on architectures and energy efficiency in Data Center Networks

We present different design options to implement Augmented Multipath Transmission Control Protocol (A- MPTCP) communications via a Locator/Identifier Separation Protocol (LISP) Traffic Engineering (TE) overlay network. MPTCP allows a TCP connection using multiple subflows to maximize resource usage. LISP is a routing and addressing architecture that provides new semantics for IP communications, by separating the device identity (endpoint identifier) from its location (routing locator) using two different numbering spaces. Our proposition is to adopt a LISP overlay network with traffic engineering capabilities to steer MPTCP subflows across wide-area Internet networks. The resulting augmentation consists of a subflow forwarding that can reach edge bottleneck capacity and surround inter-domain transit bottlenecks and inefficient paths. It can be particularly useful for cases where, even if endpoints are single-homed, inter-domain path diversity can be grasped by the LISP-TE network overlay. We specify the different modes at which this augmentation can take place, from stateless and light modes with very limited management in the network, to stateful and advanced modes implementable by a network provider desiring a higher control on the network. Based on extensive experimentation on the worldwide LISP testbed, we show that the achievable gains up 25% in throughput, while identifying required further improvements.

Network functions virtualization (NFV) offers an alternative way to design, deploy, and manage networking services. This paper presents the implementation and deployment of a virtualized network function to solve MP-TCP’ shared bottlenecks problem. MP-TCP (MultiPath-TCP) is a protocol that allows hosts that have several interfaces to send data through multiple paths in the network. An MP-TCP connection is divided into several TCP connections called subflows that may follow different paths, in this way effective connection rates and resilience can be increased. However, in a scenario where MP-TCP is used with hosts having a single network interface, created subflows follow the same path even if several are available; this generates the problem known as a shared bottleneck. A solution to this problem is implemented as a SDN (Software Defined Networking) module that is integrated into OpenDayLight’s framework which is used as the SDN controller. This solution is then deployed as a NFV over a cloud infrastructure implemented with OpenStack over a high-end physical server. The deployment of the NFV is automated to a certain degree using Ansible, a DevOps tool. Finally, results of the tests of the SDN solution to MP-TCP’s shared bottlenecks and its operation as a NFV using a completely virtualized network are reported.