The event extraction problem involves detecting event trigger words and extracting their corresponding event arguments. In contrast to general event extraction, financial event extraction focuses on financial texts, primarily at the document level. Existing methods mainly rely on a uniform sequence labeling model to identify event triggers and event arguments. However, this approach struggles with identifying nested and long entities and faces difficulties in handling complex event arguments. To address these challenges, we propose MRCFEE, a Chinese financial event extraction model based on machine reading comprehension (MRC). In the multi-turn MRC, the input to the model incorporates external knowledge and historical answers based on question templates and previous extraction results. Subsequently, we utilize a large-scale pre-trained language model in the financial field as the embedding layer and employ BiGRU to extract contextual information further. Then, we employ two binary classifiers to identify the probabilities of the starting and ending positions. Finally, we utilize the dynamic thresholding method to assess the rationality of the boundary of the obtained results. Experimental results on the DuEE-Fin dataset demonstrate that our model outperforms the previous methods, obtaining 84.1% and 75.6% F1 values for event detection and event argument extraction, respectively.

Teaching machines to understand human language is one of the most elusive challenges in artificial intelligence. Machine reading comprehension is a crucial task in evaluating how computer systems understand natural language. This study presents a machine reading comprehension model based on external knowledge. We use a framework named K-Adapter to infuse two kinds of external knowledge with two specific adapters. This model can capture richer semantic information, which is more suitable for real application scenarios. The proposed model is evaluated on the COSMOS QA dataset and outperforms the competitive baselines.

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.

As service providers have started deploying SDN in their networks, traditional IP routers are gradually upgraded to SDN enabled switches. In other words, the network will have traditional IP routers and SDN switches coexisting, and it is called a hybrid SDN network. With such a network, we take advantage of SDN and propose an approach to guarantee traffic reachability in the presence of any single link failure. By redirecting traffic on the failed link to SDN switches through pre-configured IP tunnels, the proposed approach is able to react to the failures very fast. With the help of coordination among SDN switches, we are also able to explore multiple backup paths for the failure recovery. This allows the proposed approach to avoid potential congestion in the post-recovery network by choosing proper backup paths. Simulation results show that our proposed scheme requires only a very few number of SDN switches in the hybrid SDN network to achieve fast recovery and guarantee 100% reachability from any single link failure. It also shows that the proposed approach is able to better load-balance the post-recovery network comparing to IP Fast Reroute and shortest path re-calculation.

Intradomain traffic engineering (TE) has become an indispensable tool for Internet service providers (ISPs) to optimize network performance and utilize network resources efficiently. Various explicit routing TE methods were recently proposed and have been able to achieve high network performance. However, explicit routing has high complexity and requires large ternary content addressable memories (TCAMs) in the routers. Moreover, it is costly to deploy explicit routing in IP networks. In this paper, we present an approach, called generalized destination-based multipath routing (GDMR), to achieve the same high performance as explicit routing. The main contribution of this paper is that we prove that an arbitrary explicit routing can be converted to a loop-free destination-based routing without any performance penalty for a given traffic matrix. We present a systematic approach including a heuristic algorithm to realize GDMR. Extensive evaluation demonstrates the effectiveness and robustness of GDMR.

Classical traffic engineering (TE) methods calculate the optimal routing based on a single traffic matrix. However, they are unable to handle unexpected traffic changes. Thus, it is of interest to find a good routing configuration to accommodate multiple possible traffic scenarios. There are two major approaches to achieve load balancing for multiple traffic matrices: destination-based routing and explicit routing. It has been shown that explicit routing performs better than destination-based routing for multiple traffic matrices. However, explicit routing has high complexity and requires large Ternary Content Addressable Memory (TCAM) in the routers. Thus, it is power hungry and unscalable. This paper presents an approach called hybrid routing to achieve load balancing for multiple traffic matrices with low complexity and good scalability. Our basic idea is to complement destination-based routing with a small number of explicit routing forwarding entries to take advantage of both two routing approaches. Hybrid routing greatly reduces the number of forwarding entries compared with pure explicit routing. This has great value for practice in that the scheme requires very small TCAM to implement. Hybrid routing is very suitable for implementation using SDN. A heuristic algorithm is developed to obtain the near-optimal hybrid routing configuration. Extensive evaluation demonstrates the effectiveness of hybrid routing. The results show that hybrid routing achieves near-optimal load balancing compared with pure explicit routing. In particular, hybrid routing saves at least 84.6% TCAM resources in all practical networks used in our evaluation.

Classical TE methods calculate the optimal routing based on a known traffic matrix. However, they are unable to handle unexpected traffic changes. Thus, various methods were proposed in recent years, such as online dynamic TE and robust static routing TE. However, online dynamic TE requires additional overhead on routers for information dissemination and suffers from the transient disruptions during routing protocol convergence, while using one robust static routing to accommodate a wide range of traffic scenarios is unable to ensure near optimality of performance for each individual traffic scenario. This paper presents an approach called dynamic hybrid routing (DHR) to achieve load balancing for a wide range of traffic scenarios. Our basic idea is to configure several routing policies in advance and then dynamically rebalance traffic by applying different preconfigured routing policy to react to traffic fluctuations. Each routing policy composes of a common basic destination-based routing and a few complementary explicit routing forwarding entries for a small set of selected ingress/egress node pairs. We design a method to find the near-optimal dynamic hybrid routing configuration. Extensive evaluation demonstrates the effectiveness of DHR. We show that DHR achieves nearoptimal load balancing and thus obtain about at least 96% throughput compared to optimal routing for each individual traffic scenario with very low overhead.

Enterprise data centers (EDCs) are critical infrastructure to large enterprises, government agencies, research institutions, etc. They are used to support a variety of off-the-shelf and customized services. EDCs are different from cloud data centers (CDCs) in two major aspects. Firstly, an EDC is usually built over time and consists of old and new equipment. Secondly, the type of services and applications in EDCs are quite different from those in CDCs. Therefore, we expect that the traffic characteristics in EDCs would also be different from those in CDCs. While most existing data center measurements were from CDCs, we performed extensive traffic measurement and analysis in an EDC that provided multiple services to over a million users. We present the data center architecture, measurement methodology, measurement results, and analysis. The results include traffic matrix, traffic distribution, flow characteristics, and TCP characteristics. Our research reveals that the traffic characteristics in the EDC are indeed quite different from the reported results in CDCs. For example, the traffic matrix tends to be sparse rather than all-to-all. Based on the analysis we provide a few guidelines for EDC design, optimization, and anomaly detection. As the first most extensive study on EDC traffic, our work provides valuable information to future EDC design and implementation, and also helps researchers develop insights into the differences and similarities between EDCs and CDCs.

An important challenge of running large-scale cloud services in a geo-distributed cloud system is to minimize the overall operating cost. The operating cost of such a system includes two major components: electricity cost and wide-area-network (WAN) communication cost. While the WAN communication cost is minimized when all virtual machines (VMs) are placed in one datacenter, the high workload at one location requires extra power for cooling facility and results in worse power usage effectiveness (PUE). In this paper, we develop a model to capture the intrinsic trade-off between electricity and WAN communication costs, and formulate the optimal VM placement problem, which is NP-hard due to its binary and quadratic nature. While exhaustive search is not feasible for large-scale scenarios, heuristics which only minimize one of the two cost terms yield less optimized results. We propose a cost-aware two-phase metaheuristic algorithm, Cut-and-Search, that approximates the best trade-off point between the two cost terms. We evaluate Cut-and-Search by simulating it over multiple cloud service patterns. The results show that the operating cost has great potential of improvement via optimal VM placement. Cut-and-Search achieves a highly optimized trade-off point within reasonable computation time, and outperforms random placement by 50%, and the partial-optimizing heuristics by 10-20%.

Security is critical to data centers, especially multi-tenant data centers that host a variety of applications in a single facility. Conventional schemes place security devices (middleboxes) at a few choke points (e.g., core routers) and rely on routing policy to guarantee middlebox traversal. Coupling routing and security services together complicates operation and troubleshooting since routing and security are operated by different teams. When a data center scales, the security system needs upgrade accordingly. However, the current approaches are not flexible and incur high cost. Observing that rich computing resources are already available in data centers, we are interested in using a large number of software middleboxes to achieve scalability and cost efficiency. We present Hybrid Security Architecture (HSA), a design to decouple security services from routing and to allow the integration of hardware and software middleboxes in a complementary way. HSA is more cost-effective and flexible compared to the conventional schemes that solely use hardware middleboxes. It allows topology and routing changes with minimal impact to security services, and vice versa. In particular, HSA does not require modification to switches and routers. This paper explains the framework of HSA, describes the key techniques, presents a testbed to validate the design, and discusses future research directions.

Equal-Cost Multipath (ECMP) routing has been widely adopted to perform load balancing. With ECMP, a router can maintain multiple next hops for a destination IP prefix. The most common method used by such routers is to split traffic with per-flow basis evenly among those next hops. This approach, although simple, cannot achieve optimal load balancing. In this paper we study the optimal configuration of weighted ECMP, where traffic splitting among the available paths is based on a set of pre-determined ratios. The contribution of this paper is two-fold. First, we develop a model to obtain the split ratios such that the overall network end-to-end delay is optimized. This is important because better delay performance is a result of better bandwidth allocation and has a direct impact on application, while most existing work tries to minimize the traffic load on the most utilized link. Second, we prove that the problem can be first solved by using a simple flow-based routing model and then converting the results to apply to IP networks, where destination-based forwarding is used. We present a heuristic algorithm to find the near-optimal weight configurations and demonstrate the effectiveness of the algorithm using computer simulations.

Incast applications have grown in popularity with the advancement of data center technology. It is found that the TCP incast may suffer from the throughput collapse problem, as a consequence of TCP retransmission timeouts when the bottleneck buffer is overwhelmed and causes the packet losses. This is critical to the Quality of Service of cloud computing applications. While some previous literature has proposed solutions, we still see the problem not completely solved. In this paper, we investigate the three root causes for the poor performance of TCP incast flows and propose three solutions, one for each at the beginning, the middle and the end of a TCP connection. The three solutions are: admission control to TCP flows so that the flow population would not exceed the network's capacity; retransmission based on timestamp to detect loss of retransmitted packets; and reiterated FIN packets to keep the TCP connection active until the the termination of a session is acknowledged. The orchestration of these solutions prevents the throughput collapse. The main idea of these solutions is to ensure all the on-going TCP incast flows can maintain the self-clocking, thus eliminates the need to resort to retransmission timeout for recovery. We evaluate these solutions and find them work well in preventing the retransmission timeout of TCP incast flows, hence also preventing the throughput collapse.

The recent literature on data center networks often propose the use of centralized control server to manage resources or for coordination. These controllers are always a single omniscient device. While we see no problems in practice yet as it works for small scale networks, there is a scalability concern. This paper proposes the idea of devolved controllers, namely, a number of controllers that each with only limited information about the network, but they together can replace a omniscient controller. Thus it permits the controllers' workload to scale up. We use dynamic flow reroute as an example, to study how we can build a network with devolved controllers, how to configure them, how they operate, and especially how they provide resilience. We show that these devolved controllers can remove the scalability concern and provide redundancy to each other; externally, they are as easy to use as a single omniscient controller. We describe how these controllers are prepared and used to prove that devolved controllers is feasible idea, and provide a precursor to a new direction on the control aspect of networking.

Failure recovery in IP networks is critical to high-quality service provisioning. In IP over wavelength division multiplexing (WDM) networks, a fiber carries multiple IP logical links. When a fiber fails, all the logical links it carries are disconnected simultaneously. This is called a shared risk link group (SRLG) failure. Recovery from SRLG failures using route recalculation could lead to long service disruption. In this paper, we present a scheme called multi-section shortest path first (MSSPF) that achieves ultra fast recovery from SRLG failures. MSSPF performs all the recovery related calculations in advance. On the detection of an SRLG failure, the affected IP packets are detoured to their destinations through pre-calculated paths to avoid failed links. We prove that MSSPF guarantees 100% recovery from SRLG failures and causes no permanent loops. In particular, the scheme has low complexity and can be implemented in today's networks running link-state routing protocols, e.g., open shortest path first (OSPF). The performance of our scheme is validated with a variety of practical and randomly generated topologies.

We propose a fast reroute scheme for the single link failure protection of IP multicast network. A multicast network connects nodes as a spanning tree, which is vulnerable to link failures. Our scheme uses ring topologies to ensure a backup path always exists to any single link failure and only a few nodes in the neighborhood of the failed link have to react to the failure. Compare to other solutions, we have four advantages: (1) The multicast tree can be arbitrary. (2) There is only a minimal disruption and there would be no packet loss. (3) Failure protection is available to a wider set of network topologies. (4) Only a small number of multicast nodes are involved and the number of spare links used are also small.

Failure recovery using IP fast reroute (IPFRR) has gained much attention recently. The basic idea is to find backup paths and configure the routing tables in advance. After a failure is detected, the pre-determined backup paths are used immediately to forward the affected packets. Since the calculation and configuration are performed in advance, the recovery can be completed very quickly. IPFFR is considered as a promising approach to enhance the survivability of IP networks. While single failure recovery has been extensively researched, using IPFRR for double-link failure recovery remains as a great challenge. We propose a solution for this issue called Efficient SCan for Alternate Paths for double-link failure recovery (ESCAP-DL). ESCAP-DL guarantees 100% coverage from both single and double-link failures and has the advantages of low complexity and resource requirement. The scheme resumes packet forwarding immediately after failures are detected and does not require failure advertising throughout the network.

Failure recovery in IP networks is critical to high quality service provisioning. The main challenge is how to achieve fast recovery without introducing high complexity and resource usage. Today’s networks mainly use route recalculation and lower layer protection. However, route recalculation could take as long as seconds to complete; while lower layer protection usually requires considerable bandwidth redundancy. We present two fast rerouting algorithms to achieve recovery from single-link and single-node failures, respectively. The idea is to calculated backup paths in advance. When a failure is detected, the affected packets are immediately forwarded through backup paths to shorten the service disruption. This paper answers the following questions: 1. How to find backup paths? 2. How to coordinate routers during the rerouting without explicit signaling? 3. How to realize distributed implementation? The schemes react to failures very fast because there are no calculations on the fly. They are also cost efficient because no bandwidth reservation is required. Our schemes guarantee 100% failure recovery without any assumption on the primary paths. Simulations show that our schemes yield comparable performance to shortest path route recalculation. This work illuminates the possibility of using pure IP layer solutions to build highly survivable yet cost-efficient networks.

Failure recovery in IP networks is critical to high quality service provisioning. One of the challenges is how to achieve fast recovery without introducing high complexity and resource usage. Today's main approaches are route recalculation and lower layer protection, where recalculation could take a long time to complete; while protection usually requires considerable bandwidth redundancy. IP fast rerouting achieves ultra fast failure recovery by calculating alternate paths in advance. When a failure is detected, the affected packets are immediately forwarded through alternate paths to shorten the service disruption. We present an algorithm called Efficient SCan for Alternate Paths (ESCAP) to achieve fast rerouting. The algorithm guarantees 100% recovery of single-link and single-node failures. In particular, it supports generic multi- path routing (where a router maintains multiple paths to a single destination) and does not require the paths to have equal cost. The implementation of ESCAP has low complexity and does not introduce explicit signaling between routers. Simulations show that our scheme yields comparable performance to shortest path route recalculation. This work illuminates the possibility of using pure IP layer solutions to enhance the Internet survivability.

Wavelength conversion is effective in reducing the connection blocking probability and increasing the link utilization in wavelength-routed WDM networks. However, wavelength converters are expensive in the foreseeable future, which means only a limited number of converters can be deployed in a network. In such case, placement algorithms are used to determine the locations of a given number of converters such that the connection blocking probability is minimized. This paper developed a low-complexity analytical model to reflect the impact of the converter locations on the network blocking probability. Based on this model, an algorithm is developed for converter placement. Since an analytical approach is taken, the algorithm has the advantage of high efficiency, allocation of 500 converters in two exiting networks with 14 and 19 nodes takes no more than 1 second using a personal computer. Simulations show that the proposed approach outperforms the best existing algorithm in terms of blocking probability.

With the dramatic increase of bandwidth from WDM technology, packet switching has caused a bottleneck for multi-hop networks, where electronic switches cannot scale up to high capacity while optical packet switches are still immature due to lack of optical memory. This paper proposes virtual fully connected (VFC) architecture for WDM networks to provide high performance edge-to-edge all optical transportation capability. With moderate wavelength resources, VFC architecture emulates a fully connected network by transporting traffic between edge nodes without intermediate buffering, thus a single-hop network is realized and high speed packet switches are avoided, which facilitates performance guarantee and reduces network cost. A scheduling algorithm is developed for contention resolution and bandwidth allocation, which is proved to provide 100% throughput as well as bandwidth guarantee under arbitrary traffic. Simulations show that VFC network achieves good delay performance under both uniform and non-uniform load.

With the increase of capacity and functionality in WDM networks, the control planes have become more and more important in state monitoring, resource management, and connection maintenance. The capability of fast failure restoration in a control plane not only enables effective network management but also helps efficient recovery of the transport plane. This work presents a signaling network that is able to achieve fast restoration under multiple failures. The mechanism employs distributed path restoration and introduces path probe using failure notification packets, which brings the benefits of low computation as well as short restoration time. Parallel processing architecture is developed and the complexity is analyzed. It is shown that the computation time can be easily controlled below 1 ms for a network with 256 nodes, which means the total restoration time for a wide area network is dominated by the propagation delay. Simulations show that the restored network provides good delay performance for the signaling traffic.