TCP Incast Research Articles

R-AQM: Reverse ACK Active Queue Management in Multitenant Data Centers

TCP incast has become a practical problem for high-bandwidth, low-latency transmissions, resulting in throughput degradation of up to 90% and delays of hundreds of milliseconds, severely impacting application performance. However, in virtualized multi-tenant data centers, host-based advancements in the TCP stack are hard to deploy from the operators’ perspective. Operators only provide infrastructure in the form of virtual machines, in which only tenants can directly modify the end-host TCP stack. In this paper, we present R-AQM, a switch-powered reverse ACK active queue management (R-AQM) mechanism for enhancing ACK-clocking effects through assisting legacy TCP. Specifically, R-AQM proactively intercepts ACKs and paces the ACK-clocked in-flight data packets, preventing TCP from suffering incast collapse. We implement and evaluate R-AQM in NS-3 simulation and NetFPGA-based hardware switch. Both simulation and testbed results show that R-AQM greatly improves TCP performance under heavy incast workloads by significantly lowering packet loss rate, reducing retransmission timeouts, and supporting 16 times (i.e., 60 to 1000) more senders. Meanwhile, the forward queuing delays are also reduced by 4.6 times.

Poche: A Priority-Based Flow-Aware In-Network Caching Scheme in Data Center Networks

Datacenters currently deploy shallow-buffered switches to achieve low latency by avoiding long waiting time in the data plane. However, the limited buffer space in the switch causes the frequent overflow and the notorious TCP incast problem. Moreover, the simple scheduling strategy in buffer deprives the switch of the ability to offer deeply differentiated services. Therefore, we present a novel priority-based flow-aware in-network caching scheme, named Poche, which supplies more control capabilities for the network side through introducing some additional cache resource into switches. Poche classifies network traffic into multiple priorities according to the latency requirements of flows. The end server adds priority tags to packets and sets different RTO values for flows with distinct priorities. The switch monitors the buffer utilization of each port and performs the priority-based flow-aware caching and injecting strategies based on the analysis of the scheduling model between the buffer and cache. We conduct comprehensive experiments to compare Poche with the state-of-the-art traffic optimization schemes. The results demonstrate that Poche can reduce the FCTs of latency-sensitive flows by at least 59.1% and improve the network throughput by at least 54.4%, while ensuring the finite cached volume and effectively addressing the incast problem.

Mitigating TCP incast in data center networks using enhanced application layer technique

Mitigating TCP incast in data center networks using enhanced application layer technique

An application layer technique to overcome TCP incast in data center network using delayed server response

Applications running in data center networks exhibits barrier-synchronized many-to-one communication pattern. Under this pattern, multiple workers synchronously transmit data to single aggregator. Synchronized transmission from multiple workers overload small-sized buffer of switch connected to aggregator. This causes severe packet drops and leads to overall application goodput degradation. This is named as TCP Incast problem in data center network. In this paper, an application layer technique called Delayed Server Response at Application Layer (DSRAL) is proposed to mitigate TCP Incast problem. It is delay-based sequential data transmission technique. In DSRAL, aggregator first determines amount of time required by each worker to finish its data transmission. This can be done by knowing Server Request Unit (SRU) size, average Round Trip Time (RTT) and packet scheduling time. Then along with request packet, aggregator communicates distinct SRU transmission time with each worker and enforces them to transmit SRUs sequentially. Through simulation it is proved that DSRAL avoids TCP Incast problem and provides improved performance than existing application layer serialized method. It supports large number of workers and simultaneously achieves high application goodput without causing any packet loss.

FAMD: A Flow-Aware Marking and Delay-based TCP algorithm for datacenter networks

Datacenter networks have brought new challenges to meet datacenters’ unique performance requirements. Moreover, the many-to-one traffic pattern, which is quite common in datacenter networks, causes heavy network congestion and leads to TCP incast problem. To solve this problem, many TCP congestion control protocols have been proposed to mitigate the TCP incast. However, these protocols have more or less problems, including high computational complexity and implementation complexity. In this paper, we propose a flow-aware marking and delay-based TCP algorithm for datacenters, referred to as FAMD. FAMD is built on the delay-based protocols and leverages a DCTCP-like marking scheme to indicate the network congestion level and regulate the congestion window. Based on above mechanism, FAMD can maintain a low and stable transmission end-to-end delay. Furthermore, because of the disadvantage of fixed marking threshold, this paper shows how to choose an appropriate threshold dynamically and presents a flow-aware threshold adjustment algorithm, referred to FATA. Therefore, our FAMD can achieve a good generality for different scenarios. Our extensive evaluation shows that FAMD can gain a good throughput and low end-to-end delay.

A solution to MPTCP’s inefficiencies under the incast problem for Data Center Networks

In recent years several multipath data transport mechanisms, such as MPTCP and XMP, have been introduced to effectively exploit the path diversity of data center networks (DCNs). However, these multipath schemes have not been widely deployed in DCNs. We argue that two key factors among others impeded their adoption: TCP incast and minimum window syndrome . First, these mechanisms are ill-suited for workloads with a many-to-one communication pattern, commonly found in DCNs, causing frequent TCP incast collapses. Second, the syndrome we discover for the first time, results in 2-5 times lower throughput for single-path flows than multipath flows, thus severely violating network fairness.To effectively tackle these problems, we propose AMP: an adaptive multipath congestion control mechanism that quickly detects the onset of these problems and transforms its multipath flow into a single-path flow. Once these problems disappear, AMP safely reverses this transformation and continues its data transmission via multiple paths. Our evaluation results under a diverse set of scenarios in a fat-tree topology with realistic workloads demonstrate that AMP is robust to the TCP incast problem and improves network fairness between multipath and single-path flows significantly with little performance loss.

Optimizing data center network throughput by solving TCP Incast problem using k‐means algorithm

SummaryTCP in casting referred to as one of the complex problems in the data center, which receives a huge amount of traffic volume from n number of senders. There are many methods proposed to solve the TCP Incast problem. In this paper, there are two steps proposed: (i) grouping sender nodes using the k‐means algorithm and (ii) handling incoming packets from the senders after grouping. Grouping senders make it easy to handle the incoming packets effectively as the complexity reduces reasonably. This is achieved through the k‐means algorithm that groups the senders and makes a priority queue for each group. The second step can be achieved through Pro‐Ack (Pro‐Acknowledgement) control in one of the methods for handling a huge volume of data at the receiver side. We demonstrated how the incoming packets can be processed and handled effectively through Pro‐Ack control with the help of the k‐means algorithm. Along with the above concepts, the robust algorithm is proposed to make priority queues. The input throughput also out performs with the existing algorithms.

Flow-Aware Adaptive Pacing to Mitigate TCP Incast in Data Center Networks

In data center networks, many network-intensive applications leverage large fan-in and many-to-one communication to achieve high performance. However, the special traffic patterns, such as micro-burst and high concurrency, easily cause TCP Incast problem and seriously degrade the application performance. To address the TCP Incast problem, we first reveal theoretically and empirically that alleviating packet burstiness is much more effective in reducing the Incast probability than controlling the congestion window. Inspired by the findings and insights from our experimental observations, we further propose a general supporting scheme Adaptive Pacing (AP), which dynamically adjusts burstiness according to the flow concurrency without any change on switch. Additionally, a sender-based approach is proposed to estimate the flow concurrency. Another feature of AP is its broad applicability. We integrate AP transparently into different TCP protocols (i.e., DCTCP, L2DCT and D2TCP). Through a series of large-scale NS2 simulations and testbed experiments, we show that AP significantly reduces the Incast probability across different TCP protocols and the network goodput can be increased consistently by on average $7\times $ under severe congestion.

Effective Routing Technique: Augmenting Data Center Switch Fabric Performance

Today, data center networks (DCNs) are built using multi-tier architecture. These large-scale networks face many challenges, such as security, delay, low throughput, loops, link oversubscription, TCP Incast and Outcast, etc. In this paper, a TCAM (Ternary Content Addressable Memory) based routing technique is proposed, augmenting the routing capabilities of multi-tier architectures in large scale networks. The routing complexities in these architectures are rectified and improved by implementing an additional TCAM based routing table in Leaf/ToR switches for a specific number of compute nodes in particular Pods, and it is scalable to whole datacenter nodes. To test the model, we implemented two prototype models: one depicting our proposed TCAM based switch and the other is a typical Top-of-the-Rack (ToR) switch and compared the performance of the proposed model and if any overhead introduced in it. The preliminary results show that our TCAM based routing table technique is fast and it forwards the network packets at line-rate, does not introduce considerable latency, on-chip resources power consumptions is less than 3%, and helps to solve or mitigate the above critical problems that are present in the current large DC's three-tier architecture, especially in Top of the Rack and aggregation layers switches.

TCP Incast in Data Center Networks: Issue and Existing Solutions

Data centers networks supports heterogeneous kind of applications like social networking, e-commerce, web search, video data hosting, computation-intensive and data-storage. It has high-bandwidth links, low propagation delay and commodity switches with small-size buffers. Under cluster-based storage environment, data center supports barrier-synchronized many-to-one communication pattern where multiple worker nodes simultaneously transmit bulk of data to single aggregator node by running standard TCP protocol. This synchronized transmission may overload aggregator’s switch buffer, which leads to severe packet loss and overall throughput fall. This is called as TCP Incast problem. This paper analyses issue of TCP Incast and provides detailed survey about several solutions at link, transport and application layer to mitigate impact of TCP Incast in data center network. Solutions are described with their procedural approach for alleviating Incast. Comparative evaluation between these solutions provides understanding about their merits, demerits and applicability under various implementation circumstances.

PTCP: A priority-based transport control protocol for timeout mitigation in commodity data center

In data centers, the occurrence of timeout for TCP may hurt its data transmission performance dramatically, causing problems like TCP Incast, TCP Outcast and long query completion time. To mitigate timeouts, the transport protocol should try to maintain a small switch queue to avoid the packet loss and recover lost packets quickly. Recent work suggests using Explicit Congestion Notification (ECN), Round Trip Time (RTT) or the in-network signal to achieve that. However, these solutions either still suffer from many timeouts when the number of concurrent flows becomes larger or require the nontrivial hardware support. The limitations motivate us to design a Priority-based Transport Control Protocol termed PTCP to mitigate timeouts as far as possible for commodity data center. The key idea of PTCP is that it inserts a high priority packet following each window of data packets. The key insight is that since the priorities of data packets and the inserted packet are different, they may arrive at the receiver in different sequences depending on the network congestion. By checking the sequences of the received ACKs of the two kinds of packets, PTCP can infer the network congestion to guide the fine adjustment of its sending window such that the switch buffer occupation is kept small. Additionally, by keeping the high priority packet always in flight, PTCP could determine to retransmit the possible lost data packets quickly. With the two mechanisms, PTCP significantly alleviates timeouts even when the number of concurrent flows becomes large. Furthermore, PTCP only requires the priority queuing function of switch, which is available in existing commodity switch hardware. Thus, it does not require the hardware modification. Extensive performance evaluation is conducted to demonstrate that PTCP has zero timeout and better performance for problems like TCP Incast, TCP Outcast and long query completion time compared with several state-of-the-art protocols.

RAPID: Avoiding TCP Incast Throughput Collapse in Public Clouds With Intelligent Packet Discarding

Many applications in public clouds require a high fan-in, many-to-one type of data communication (known as TCP incast) in modern Data Center Networks (DCNs). Such communication could cause severe incast congestion in switches and result in TCP throughput collapse, substantially degrading the application performance. The root cause of throughput collapse is the Retransmission Timeouts (RTO) due to packet losses in congested switches. Tenants in public clouds can opt to use a variety of TCP versions. However, the existing solutions rely on modifications of TCP protocols and specific techniques from switches, and thus these existing solutions are not always feasible for public clouds. In this paper, we are inspired by the emerging virtualization and network softwarization technologies to develop a novel scheme called Retransmission timeout Avoidance by Packet Intelligent Discarding (RAPID) using software switches. RAPID considers the number of packets of each incast flow, buffered in the switch to selectively discard some packets, and ensures that the Fast Retransmission/Fast Recovery rather than RTO is invoked at the sender(s) in response to packet loss. Thus, the long idle period of a timeout and the throughput drop are avoided. We prove that, given a predetermined minimum switch buffer space, dedicated to the incast application, RAPID can prevent RTO in all the incast senders. We also present a low-complexity heuristic version of RAPID named RAPID-ED, which combines the principles of RAPID and early detection and is extremely easy to implement on today’s software switches. We evaluate the two proposed schemes in a data center network testbed built on NS-3 simulator. The simulation results confirm the theoretical expectation, and show that the RAPID and RAPID-ED perform very well to prevent RTO of TCP incast flows and hence the throughput collapse. Compared with other incast solutions, RAPID and RAPID-ED do not modify TCP protocols and therefore are more suitable in public clouds.

A Hybrid Framework for TCP Incast Congestion Control in Data Center Networks

Cloud data centers with large bandwidth and low latency networks experiences many-to-one traffic pattern called TCP-incast. It occurs in the partition-aggregate architecture and causes emergence congestion at the network wharfconnected to the parent server, overcoming the port emergence buffer. The ending packet loss requires nodes to encounter loss, retransmit data and slowly rise up throughput per definitive TCP behavior. This paper proposes Receiver-oriented Congestion Control with Edge computing approach (RCCE) for enhancing the speed, nature and firmness of traffic performance. Receiver-oriented Congestion Control (RCC) combines both closed and open loop congestion controls at receiverwhereas edge computing involves localization of traffic management in the middle-tier aggregator for reducing Flow Completion Times (FCT) and latency for the entire application processing deployments. In addition, the centralized controller at the edge balances the load during incast by using spanning trees in a well-made manner by implementing multi-stage Clos networks. The entire prototype is implemented in ns3 and simulation results demonstrates that RCCE has an average decrease of 60.2 % in the 99th percentage latency and 50.4 % of mean queue size in the heavy traffic over TCP.

Queue length feedback-based solution of TCP Incast in data center networks

The Internet offers a large number of applications and services that we use on a daily basis. These widely used applications are hosted on large-scale, high-performance computing systems called data centers. The performance of TCP is inefficient in many-to-one communication, which is a common traffic pattern in data center networks. This many-to-one communication causes significant packet losses followed by timeouts, which consequently results in throughput collapse in data center networks; this problem is known as TCP Incast. In this paper, we present a queue length feedback-based solution to mitigate TCP Incast. The scheme has two parts: i) a novel queue length-based congestion parameter, which accurately measures congestion along the path from source to destination, and ii) a congestion control scheme that effectively uses the new congestion parameter to prevent throughput collapse due to Incast traffic patterns. Results are compared with TCP and DCTCP, the two most common transport protocols deployed in data center networks. The results show that the proposed scheme minimizes packet drops and achieves high utilization and burst tolerance.

Datacentre TCP Protocol of Centralized Window Control

There is an important class of applications in data centre called online dataintensive (OLDI) application, such as search engine, electronic shopping, and advertising, which has low latency, high throughput, and high burstiness. For better user experience, the query response delay in the data centre is controlled under 300 milliseconds. DCTCP can better meet the requirements of OLDI, but all the switches in the data centre are required to support RED. For data centre that does not fully support RED switch, TCP protocol can only be used. TCP Incast will appear in the application, which is far from the requirements of OLDI application. This paper proposes the TCP protocol of data centre based on window control. It is in existing switches and network devices,and centralized control makes the send window match with the switch queue cache. In the case of ensuring low latency, it keeps the switch queue in a small range of fluctuation, enabling the data centre to meet low latency, high throughput and high burstiness application requirements. Simulation experiments verify that WCTCP congestion control method can effectively solve the above problems. It is 5 times lower than TCPNewReno packet loss rate, 6 times higher than the network utilization rate, and the network performance is close to the DCTCP protocol.

Adjusting Packet Size to Mitigate TCP Incast in Data Center Networks with COTS Switches

In data center networks, a large number of concurrent TCP connections suffer the TCP Incast throughput collapse due to packet drops in shallow-buffered Ethernet switches. In this work, we first reveal theoretically and empirically that controlling the IP packet size is much more effective in avoiding Incast than cutting congestion window under severe congestion. We further design a general supporting scheme Packet Slicing, which adjusts the IP packet on widely used commodity switches. The design uses standard ICMP signaling, which makes no modification on TCP protocols and can be transparently utilized by various TCP protocols. To alleviate the impact of micro-burst caused by high flow concurrency, we utilize the TCP Pacing scheme to disperse packets over the round trip time, helping Packet Slicing to support more concurrent TCP flows. We integrate Packet Slicing with three state-of-the-art data center TCP protocols on NS2 simulation and a physical testbed. The experimental results show that Packet Slicing broadly improves the goodput of different data center TCP protocols by average 26x, while having almost no effect on the I/O performance of switches and end hosts.

CCMA: A Dynamical Concurrent-Connection Management Agent to Mitigate TCP Incast in Datacenters

While aggregator applications with many-to-one TCP traffic pattern prevail in datacenter networks, they often suffer from catastrophic goodput degradation due to their highly bursty traffic overwhelming a single bottleneck link, which is known as the TCP Incast issue. In this paper, we first derive three analytical results that mitigate Incast via managing connections, i.e., equalizing the in-flight packet amounts of concurrent connections, closing all the connections with higher timeout possibilities, and closing some of the connections with identical timeout possibilities. Afterward, we design an Incast mitigation application, i.e., concurrent-connection management agent (CCMA). Unlike previous Incast solutions, CCMA is an application-layer agent on the receiver, which manages concurrent TCP connections on behalf of aggregator applications. CCMA requires no modification to existing communication hardware or software. For multiple coexisting aggregators, CCMA sequentially schedules them to avoid traffic collision. For each aggregator, CCMA dynamically adjusts the advertised window sizes and the concurrency number of each aggregator following our analytical results to achieve low Incast probability and high goodput. The extensive real-world experiments and NS3 simulations show that CCMA maintains zero Incast probability while greatly improving the goodput of TCP by 7.8–13.5 times.

PTCP: A Priority-Driven Congestion Control Algorithm to Tame TCP Incast in Data Centers

Data centers have become a prevalent infrastructure to host a large number of services, such as social networking and Web search, which generally have rigorous delay requirements. Even a slight increase in delay can seriously affect the performance of applications. Therefore, there have been several efforts focusing on minimizing the flow completion time (FCT) to meet the delay requirements, such as ICTCP, which adjusts the receive window according to the available bandwidth and DCTCP that uses ECN to provide the feedback to end hosts in mixing workloads. However, both approaches are only for a specific scene and cannot effectively solve the problem that packets losses lead to timeouts due to TCP incast. Moreover, they all use the fair-share mechanism, which causes the increasing of the FCT for small flows, when background flows pour into the network. For example, background flows cause the increasing(≈ 5ms) of the FCT for the ICTCP in an incast experiment. To solve these problems, we propose a priority-driven congestion control algorithm, PTCP, which can effectively avoid the incast problem and improve the FCT in mixing workloads. The PTCP leverages the priority to adjust the receive window and control the interval time of ACK. To evaluate the PTCP, we conduct extensive experiments in NS-2. The simulation results show that the PTCP outperforms greatly previous schemes both in the incast scenario and mixing workloads.

Dynamic ECN marking threshold algorithm for TCP congestion control in data center networks

Recent years have witnessed the rapid development in data center, which hosts a wide variety of applications and services with diverse requirements for network performance. Data center networks (DCNs) have the characteristics of high bandwidth and low latency and TCP incast usually happens when multiple senders simultaneously communicate with a single receiver. In this paper, we first investigate the DCTCP, which is widely deployed in DCNs and uses Explicit Congestion Notification (ECN) as a congestion signal when queue length is larger than a fixed threshold. We demonstrate that the queuing delay may be high or the bottleneck link utilization can be insufficient due to inappropriate fixed threshold. Then we propose a dynamic ECN marking threshold algorithm based on the number of concurrent flows, referred to as DEMT, to improve the network performance. After that, an enhanced DEMT algorithm, referred to as EDEMT, is proposed by using the ratio of the queue length over the marking threshold to perceive the degree of network congestion. We also integrate DEMT with DCTCP to observe its behavior and theoretically analyze the throughput. Finally, simulation results demonstrate that the DEMT and EDEMT algorithms can effectively improve the network throughput while they can maintain a small queue length as well as the short flow completion time.

Empirical Investigation of TCP Incast Congestion in Wireless Cloud Computing Networks

The promising services offered by cloud computing environments have led to huge amount of data that need to be processed and stored. Wireless cloud networks rely on Transmission Control Protocol/Internet Protocol (TCP/IP) for reliable transfer of data traffic between the cloud end-users and servers and vise-versa. Even though TCP has been successful for several applications, it, however, does not perform well in wireless cloud environments. The many-to-one communication pattern used in such environments with such huge amount of data resulted in TCP incast problem. Transmission Control Protocol incast problem happens in cluster based storage workloads where a lot of end-users communicate simultaneously to a server in the cloud through a bottleneck router, creating buffers overflows which lead to high packet loss. This paper presents an empirical study on TCP incast in current wireless cloud networks and how it is caused. It evaluates TCP-Vegas and TCP-Sack to examine their behaviors and suitability for short-lived connections in terms of queue occupancy level, packet drops, throughput, link utilization and bandwidth unfairness between the TCP connections. It was found that both protocols suffer from high packet loss and link underutilization with comparable throughput.