Abstract: A multi-mainframe system problem determination method includes recording, in a first computing system, diagnostic data, receiving, in the first computing system, a data collection trigger, responsive to the data collection trigger, coordinating, in the first computing system, synchronized collection of recorded diagnostic data with a second computing system and delivering collected diagnostic data to a storage medium.

Abstract: In a multiple interface, low power and lossy network comprising a plurality of nodes, a low transmission power and medium transmission power topology are defined for the network and a channel-hopping schedule is defined for the devices operating in each topology. A sender determines that data is capable of being transmitted via a link on the low transmission power topology. The sender determines the transmission parameters for the transmission of the data over the link on the low transmission power topology and determines a low transmission power channel for transmission of the data. The sender transmits the determined channel and the transmission parameters to the receiver. The sender transmits the data via the determined channel in the low transmission power topology.

Abstract: A method for controlling transit of routing messages in a network comprising multiple autonomous systems (AS) is disclosed. The method includes receiving, at a first AS, a routing message of an inter-AS routing protocol and identifying that the routing message comprises transit domain control (TDC) information specifying one or more autonomous systems to which the routing message may be propagated and/or one or more autonomous systems to which the routing message may not be propagated. The method further includes propagating the routing message from the first AS to a second AS in accordance with the TDC information.

Abstract: According to an implementation, a method for signaling that a first network node exists in a network is following entry of the first network node into a reduced power mode includes receiving, in a second network node, an indication that the first network node will enter into the reduced-power mode. The method also includes generating information that signals the existence of the first network node and that the first network node is reachable in the network and communicating the information to the network such that the second network node operates as a proxy advertiser for the first network node following entry of the first network node into the reduced-power mode.

Abstract: The present disclosure describes system and methods for network planning. The systems and methods can incorporate network traffic demands, availability requirements, latency, physical infrastructure and networking device capability, and detailed cost structures to calculate a network design with minimum or reduced cost compared to conventional methods. In some implementations, the method include providing an initial, deterministic set of failures, and then successively performing a network optimization and a network availability simulation to determine which failures most impact the performance of the network model. The high impact failures can then be provided back into the system, which generates an improved network design while still maintaining minimum cost.

Abstract: A method for engineering traffic in a communications system includes determining a set of delay constraints associated with a traffic flow over the communications system, and excluding non-convex constraints from the set of delay constraints, thereby producing a set of convex constraints. The method also includes selecting a path solution for the traffic flow in accordance with the set of convex constraints, and sending information regarding the path solution to nodes in the communications system.

Abstract: A heterogeneous network system, network apparatus, and rendezvous path selection method thereof are provided. The heterogeneous network system includes partner network apparatuses and a network apparatus. The transceiving interfaces of the network apparatus and the partner network apparatuses are not all the same. The network apparatus measures at least one first transmission parameter and receives a plurality of second transmission parameters. Each of the at least one first transmission parameter is associated with transmission from the network apparatus to one of its neighboring network apparatuses. Each of the second transmission parameters is associated with transmission between two network apparatuses neighboring each other. The network apparatus enumerates a plurality of paths from itself to a target network apparatus and calculates a path cost for each of the paths according to at least a portion of a combination of the at least one first transmission parameter and the second transmission parameters.

Abstract: A system and method for multicast routing using peer groups includes a router. The router includes a control unit and a memory coupled to the control unit. The memory stores one or more first multicast routing entries. The control unit is configured to receive a multicast packet on an input interface, determine one or more properties of the multicast packet, determine one or more output interfaces based on the properties and the first multicast routing entries, and forward copies of the multicast packet on each of the output interfaces. The router is configured to be coupled to a peer router via an inter-chassis link (ICL). The first multicast routing entries include one or more second multicast routing entries associated with multicast routing trees associated with the router and one or more third multicast routing entries associated with multicast routing trees associated with the peer router.

Abstract: In one embodiment, a source transmits one or more data packets to a destination over a primary pseudowire (PW). When a device on the primary PW detects a downstream failure of the primary PW, and in response to receiving one or more data packets from a source from the failed primary PW, the device adds a loopback packet identifier to the one or more received data packets, and returns the one or more data packets with the loopback packet identifier to the source upstream on the primary PW. Accordingly, in response to receiving the data packet returned with a loopback packet identifier from the primary PW (in response to the downstream failure), the source retransmits the one or more data packets to the destination over a backup PW.

Abstract: Traffic recovery is supported at a switching node (20) of an OpenFlow network (5). The switching node (20) has a flow table (23) for storing flow entries (24) which determine forwarding of received packets between the ports. A switching node (20) installs (102) a flow entry for a back up path in the flow table (23). The switching node (20) renews (104, 105) the flow entry for the backup path based on at least one of: (i) an association between the flow entry for the backup path and a flow entry for a working path at the switching node, wherein the flow entry for the backup path is renewed when the flow entry for the working path is used to forward a received packet; (ii) receiving a flow entry renewal packet from another switching node on the backup path. A backup path can be configured for each of multiple points of failure in the working path.

Abstract: Expertise, for performing classification of a type of network traffic, can be encapsulated in a module. A set of modules, as currently available to a traffic controller, can be referred to as a collection. Programming language constructs are introduced that facilitate the writing of modules customized to identify network traffic that is peculiar to a particular user, or to a relatively small group of users. An example programming language, based on Tcl, is introduced. A key aspect is event-driven programming, and the “when” command construct is introduced. Three important event types, that can trigger a “when” command, are CLIENT_DATA, SERVER_DATA, and RULE_INIT. Another key aspect is an ability to keep state information between events. Constructs for intra-session, intra-module, and global state are introduced. A module can be blocked from executing more than once for a session. Successful execution of a module can be specified by a “match” statement.

Abstract: In one embodiment, a router located at an exit edge of an autonomous system (AS) receives a data packet in a data plane, and determines a destination of the data packet and an associated AS-path information to the destination. The router may then insert the AS-path information into the data packet, and forwards the data packet with the AS-path information toward the destination, such that a receiving device in a destination AS can validate whether the data packet was routed through a path that was secure from a control plane perspective based on a collection of one or more insertions of AS-path information.

Abstract: An apparatus generates routing data that defines, in a system in which a plurality of switches are connected to each other in a tree-shaped structure, a communication route between a plurality of end switches located at a lower level in the tree-shaped structure and a plurality of upper-level switches located at a level higher than the plurality of end switches in the tree-shaped structure, so that a first upper-level switch that switches communication data from a first node connected to a first end switch among the plurality of end switches to a second node connected to a second end switch among the plurality of end switches is identical with a second upper-level switch that switches communication data from the second node to the first node. The apparatus stores the generated routing data in a memory provided for the apparatus.

Abstract: This method (100) for encoding a list of identifiers in a network including a transmitter including a global list of identifiers and able to transmit a code corresponding to a coded list of identifiers, comprises the following steps: associating (105) with each identifier an ordinal number; ordering the list of identifiers in order to obtain a sorted list; defining (120) a variable equal to the number of identifiers in the list to be transmitted; if the variable is positive (125), coding (130, 135) the first identifier of the sorted list with a code corresponding to the number of sub-sets of the global list of cardinal equal to the number of identifiers of the sorted list and including at least one identifier, for which the ordinal number is in a strict order relationship with the ordinal number of the first identifier, removing (140) this identifier from the sorted list; coding (145) the list with the sum of the obtained codes.

Abstract: In one embodiment, an apparatus includes a memory, a hardware processor, and logic integrated with and/or executable by the processor. The logic is configured to receive one or more software defined network (SDN) routes dictating a path through a network comprising a plurality of devices. The logic is also configured to store the one or more SDN routes to the memory along with one or more traditional routes learned by the apparatus and/or configured by an administrator, and indicate the one or more SDN routes as being of a type different from the traditional routes. Moreover, the logic is configured to receive a priority ordering for a plurality of routes stored in the memory from the SDN controller, the plurality of routes including at least one SDN route, and construct a route information base (RIB) based on the plurality of routes and the priority ordering.

Abstract: In a CN access network with a virtual node for at least one third party, the virtual node being used for bearing and operating the function defined by the third party, the UE accesses from a fixed access network; and the access network provides the UE with the communication connection between the UE and the virtual node of the third network/service operator. According to the embodiment the access network accommodates the virtual node to operate the function related to the third party, and improves the function topology architecture of the network, to provide users with better user experience quality.

Abstract: The method includes the following steps. First, a controller receives a data packet that is reported by a switch connected to a source host; then the controller searches a local media access control (MAC) address table of physical addresses of devices to determine whether a destination address of the data packet exists, and the controller acquires a destination port if no destination address of the data packet exists, where the destination port is a port of a switch connected to a destination host, and the destination host is another host different from the source host; and finally, the controller uses the destination port as a destination address to generate a routing table, and sends the routing table to the switch connected to the source host, so that the switch connected to the source host forwards the data packet according to the routing table.

Abstract: A method and apparatus of a device that exports Border Gateway Protocol (BGP) data in sFlow samples. The device generates separate tables for BGP attribute information, community list information, and autonomous systems path (AS_PATH) attribute information, where each entry in the BGP attribute information table is associated with an entry in the community list information table and an entry in the AS_PATH attribute information table. The device further populates a BGP route information table that defines a mapping from a network address prefix to an entry in the BGP attribute information table. The device generates a sFlow sample related to the network address prefix using the BGP attribute information table, the community list information table, the AS_PATH attribute information table, and the BGP route information table.

Abstract: A method for controlling large flows in a network includes asynchronously calculating a traffic rate associated with a network flow by a controller device based on a randomly sampled packet, comparing the calculated traffic rate with a threshold value, and generating a notification of a large flow when the rate exceeds the threshold value. The method further includes creating a filter in response to the notification, assigning an action to the filter, and installing the filter and the action in a network device. In addition, the method includes removes the installed filter and action from the network device after a time interval has elapsed.

Abstract: A method implemented by a network controller, comprising receiving an address filter from a service chain orchestrator, wherein the address filter comprises a plurality of service instance (SI) addresses associated with a plurality of SIs in a network, obtaining network topology information from a plurality of service function (SF) forwarders controlled by the network controller, wherein the network topology information indicates a plurality of network devices communicatively coupled to the plurality of controlled SF forwarders, and wherein at least some of the network devices are SI hosting devices, and generating a service chain topology map by building mappings between at least some of the plurality of controlled SF forwarders, the SI hosting devices, and the plurality of SIs according to the network topology information and the address filter.

Abstract: In one embodiment, a link utilization threshold is assigned to a first link. Link utilization of the first link and of one or more second links is monitored. The first link is considered to be out-of-policy (OOP) in the event that the link utilization of the first link surpasses the link utilization threshold. In response to the first link being considered OOP, action is taken. An excess utilization of the first link is determined. One or more prefixes currently routed over the first link whose collective per prefix utilization exceeds the excess utilization are determined. Provided that the collective per prefix utilization of the one or more prefixes currently routed over the first link, if added to the link utilization of the one or more second links, would not cause the one or more second links to be considered OOP, the one or more prefixes are redirected over them.

Abstract: Described herein is a Core Network (CN) node 200 and a Radio Access Network (RAN) node 100 and methods therein for handling quality of service, QoS, targets for traffic flows between the RAN, node and a radio terminal 300 served by the RAN node, disclosing an obtaining a number of traffic classes into which the traffic flows can be classified, and QoS targets are defined for each obtained traffic class for at least two congestion levels, and wherein the traffic classes with the associated QoS targets and congestion levels are provided to the RAN node so as to provide congestion level dependent QoS targets for the traffic classes to the RAN node.

Abstract: Various exemplary embodiments relate to a method for placing components of a plurality of instances of a cloud application on nodes in a cloud infrastructure, the method including: receiving scenario probabilities of a plurality of cloud application scenarios, wherein the plurality of cloud application scenarios define the modes of operation of the cloud application; receiving cloud infrastructure performance data; defining a performance metric of the cloud application based upon cloud infrastructure performance data; defining constraints on the placement of the application components; receiving a service level agreement performance requirement based upon a probability; optimizing the placement of the components of the plurality of instances of the cloud application on nodes in the cloud infrastructure based upon the scenario probabilities, cloud infrastructure performance data, the performance metric, and the constraints on the placement of the application components to meet the service level agreement perfo

Abstract: Embodiments presented herein describe techniques for isolating multicast and broadcast frames to a traffic class that is separate from a traffic class used for unicast frames. According to one embodiment, a network switch receives an incoming Ethernet virtual local area network (VLAN)-tagged frame. The switch evaluates priority bits of the VLAN tag of the frame. The switch also determines a type of frame (e.g., whether the frame is unicast, broadcast, multicast, or flood). Based on the priority field values and the type of the frame, the switch identifies a mapping of the frame to a particular traffic class. The network switch assigns the frame to the traffic class.

Abstract: A path selector device of a network receives a network packet. A packet flow category to which the packet belongs is identified. A candidate outbound link set corresponding to the packet flow category, comprising a subset of the available outbound links of the path selector device, is determined. The packet is transmitted on a particular outbound link of the candidate outbound link set. Subsequent packets of the packet flow category are distributed among the members of the candidate outbound link set.

Abstract: A data flow is received at a network processor that includes a plurality of frames. A first set of frames in the plurality of frames are passed from the network processor to a general processor for processing by the general processor. A flow acceleration request is received at the network processor from the general processor based at least in part on inspection of a first frame in the first set of frames. The flow acceleration request is received subsequent to passing at least two of the first set of frames to the general processor. A particular frame in the plurality of frames received subsequent to the first set of frames is processed by the network processor such that it is accelerated relative to processing of the first set of frames by the general processor and bypasses the general processor.

Abstract: The present solution is directed to systems and methods for providing, by a device intermediary to a plurality of clients and one or more servers, analytics on a stream of network packets traversing the device. The systems and methods include the device identifying, while the device manages network traffic between the plurality of clients and the one or more servers, a stream of network packets, from a plurality of streams of network packets of the network traffic traversing the device, corresponding to a flow identifier, e.g., a selected one of an internet protocol address, a uniform resource locator or an application identifier. The systems and methods may include a collector of the analytics engine collecting, while the device manages network traffic, metrics on the identified stream of network packets and generating one or more stream objects that comprise the collected metrics.

Abstract: A communications controller is provided. The communications controller includes a flow manager that classifies a packet flow serviced by more than one transmission points (TPs) as one of a plurality of slices in accordance with at least one of a nature of the packet flow, a load status of each of the plurality of slices, and feedback information provided by the more than one TPs, and alters a classification of the packet flow in accordance with the load status of each of the plurality of slices, and feedback information provided by the TPs served by the communications controller. The communications controller also includes a memory coupled to the flow manager, the memory stores a packet of the packet flow in one of a plurality of packet queues in accordance with the classification of the packet flow.

Abstract: A method and apparatus for rate adaption of HTTP stream media is disclosed in this invention. The method comprises: acquiring from a server a stream media with a selected video rate at a client adopting the TCP protocol at its transmission layer and the HTTP protocol at its application layer, wherein the selected video rate allows an amount of buffered data to be maintained in a predetermined range in a buffer for buffering the stream media under a current download rate speed. This invention may improve user experience.

Abstract: In some embodiments, a motor drive system includes a communication subassembly electrically coupled to a control subassembly for providing communication between the control subassembly and at least one external device. The control subassembly and the communication subassembly transmit and receive control data independently of messaging data, greatly enhancing performance and reducing the workload of processors on the control subassembly and communication subassembly. Additionally, the control subassembly and the communication subassembly transmit message data via sequence count based messaging. Communications between the communication subassembly and the control subassembly may include a series of timeout periods and retries, increasing reliability.

Abstract: The invention relates to a method and device in a communications network when a User Equipment, UE, needs to update a next expected Transmission Sequence Number, TSN, continuously in order to avoid discarding successive transmissions. The method comprises: resetting a next expected TSN value in said network and UE to predetermined values after a predetermined inactivity time.

Abstract: A method for transporting communications signals includes receiving an analog IF signal at a first unit. The analog IF signal includes a first carrier having a first frequency and a first bandwidth and a second carrier having a second frequency different from the first frequency and a second bandwidth. The analog IF signal is converted to a digitally sampled IF signal having the first carrier located in a first Nyquist zone, the second carrier located in a second Nyquist zone, an image of the first carrier located in a third Nyquist zone, and an image of the second carrier located in the third Nyquist zone. The image of the first carrier and the image of the second carrier is transmitted from the first unit to a second unit, where the image of the first carrier and the image of the second carrier is then converted to the analog IF signal.

Abstract: A method for implementing the application for speaking right of a Long Term Evolution (LTE)-based broadband trunking system, a Mobility Management Entity (MME), a network subsystem, a broadband wireless access subsystem and the LTE-based broadband trunking system are disclosed. The method includes: the broadband wireless access subsystem sending a trunking speaking right update request to the network subsystem, receiving the trunking speaking right update accept message returned by the network subsystem, and sending the trunking speaking right update accept message to a speaking right seizing terminal; and the broadband wireless access subsystem receiving the speaking right occupation prompt message sent by the network subsystem, and sending the trunking speaking right occupation prompt message and updated configuration information via a multicast control channel of an enhanced Multimedia Broadcast Multicast Service (eMBMS).

Abstract: A communication device includes: a plurality of output ports; a plurality of queues in which packets are stored so as to be sorted into groups of packets that are output from an identical output port in an identical time period, from among the plurality of output ports; a plurality of first selectors that respectively corresponds to the plurality of output ports, and each of which switches a queue from which packets that are output from the output port are read, between the plurality of queues each time the time period elapses; and a second selector that switches a first selector from which packets are output, between the plurality of first selectors, at time intervals in accordance with output rates of packets of the plurality of output ports.

Abstract: A communication device, includes: a plurality of queues each configured to accumulate a packet; a scheduler configured to provide a permissible readout amount to each of the plurality of queues in accordance with an order that is based on a priority of each queue; a read processor configured to read out the packet from the plurality of queues, the permissible readout amount being consumed according to amount of the packets read out; and an accumulation amount counter configured to count an accumulation amount of the packets accumulated in each of the plurality of queues, wherein the accumulation amount counter notifies the scheduler of a change in the accumulation amount, and wherein the scheduler adjusts the priority of, among the plurality of queues, the queue of which the accumulation amount has changed, in response to the notification from the accumulation amount counter.

Abstract: Seamless path switching is made possible in a multi-hop network based upon stream marker packets and additional path distinguishing operations. A device receiving out-of-order packets on the same ingress interface is capable of determining a proper order for the incoming packets having different upstream paths. Packets may be reordered at a relay device or a destination device based upon where a path update is initiated. Reordering packets from the various upstream paths may be dependent upon a type of service associated with the packet.

Abstract: A rate adaptation system includes a barrel shift slot register and a rate adaptation register. The barrel shift slot register includes a plurality of slots with one of a valid read request or a dummy read request. A rate adaptation register is configured to sequentially cycle through the slots of the barrel shift register in response to a clock providing valid read requests to a FIFO buffer and to skip provision of valid read requests for clock cycles of the first clock associated with slots that include dummy read requests. The rate adaption register may also receive data blocks from the FIFO buffer and provide those data blocks to another FIFO buffer.

Abstract: Embodiments are directed to preventing flapping when auto-scaling cloud resources. In one scenario, a computer system accesses information specifying a target operational metric that is to be maintained on a plurality of cloud resources. The computer system determines a current measured value for the target operational metric for at least some of the cloud resources. The computer system further calculates a scaling factor based on the target operational metric and the current measured value, where the scaling factor represents an amount of variance between the target operational metric and the current measured value. The computer system also calculates a delta value representing a modified quantity of cloud resources modified by the calculated scaling factor and determines whether a scaling action is to occur based on the calculated delta value.

Abstract: A computing environment provisioning system includes a processing system and a memory for storing a locking service that is executed by the processing system to receive multiple provisioning requests from one or more provisioning scripts for provisioning multiple resources to be provided by one or more physical hosts or one or more virtual objects of a computing environment in which at least one resource being dependent upon operation of one or more of the other resources. The locking service may then identify the other resources that are required to be provisioned prior to the dependent resource being provisioned, and inhibit provisioning of the dependent resource until the other resources have been successfully provisioned.

Abstract: A method and an apparatus for managing task in mobile device. The method includes estimating a first amount of resources that will be consumed by the mobile device in executing the task; determining network configuration that is allocable to the mobile device; estimating a second amount of resources that will be consumed by the mobile device when a computing node in a mobile computing environment executes the task under the network configuration; and determining that the task is to be executed by the mobile computing environment in response to the second amount being less than the first amount by a predefined threshold.

Abstract: An example method is provided for a computing device to provide quality of service (QoS) for a container in a virtualized computing environment. The method may comprise receiving a traffic flow of packets from a virtual machine and identifying a container from which the traffic flow originates based on content of the received traffic flow of packets. The container may be supported by the virtual machine. The method may further comprise retrieving a QoS policy configured for the identified container. For example, the QoS policy may specify a network bandwidth allocation for the container. The method may further comprise forwarding the received traffic flow of packets according to the QoS policy.

Abstract: A computing device may conduct a communication session. The communication session may be locally controlled by a main processor of the computing device. The main processor may be in a first power mode and may maintain an active state representation of the communication session. Packets of the communication session may pass through a network interface apparatus of the computing device, and a secondary processor in the network interface apparatus may examine at least some packets of the communication session to maintain a non-active state representation of the communication session. The main processor may transfer local control of the communication session to the secondary processor. The secondary processor may activate the non-active state representation and continue conducting the communication session using the activated state representation. The main processor may also transition from the first power mode to a second power mode that uses less power than the first power mode.

Abstract: A method for processing a resource access request in a network based on a software defined network (SDN) protocol includes receiving a resource access request forwarded by a first SDN switch in the network, wherein there exists no forwarding rule matching the resource access request in the first SDN switch; obtaining, according to the resource access request, resource server information of a resource server where the requested resource is stored, wherein the resource server is located in the network; generating a forwarding rule corresponding to the resource access request according to the resource server information; and sending to the first SDN switch at least one portion of the forwarding rule corresponding to the resource access request, so that the first SDN switch forwards the resource access request according to the at least one portion of the forwarding rule.

Abstract: Systems, computer-implemented methods, and non-transitory computer-readable medium having computer program stored therein can provide adaptive priority scheduling of communications over a communication network and dynamic resources allocation among a plurality of devices positioned in the communication network.

Abstract: A method for enabling network elements (NEs) operating at a bit rate R1 to communicate with NEs operating at a bit rate R2 is described. A ratio of R2 to R1 is represented by a ratio M:N, M and N are positive integers, and M>N.

Abstract: A Broadband Network Gateway (BNG) pool based responding method is described, including that a BNG device in the BNG pool receives a user access request, and the BNG device responds according to the user access request and a delayed response strategy corresponding to the BNG device. A BNG device, a user equipment and a BNG pool based responding system are also described. Thus a BNG device in a BNG pool can respond to a user access request according to a delayed response strategy, then a user equipment can take a BNG device corresponding to a first response message received as a server, thereby being capable of ensuring system stability.

Abstract: A device with dynamically tunable heterogeneous latencies includes an input port configured to receive a packet via a network, and a processing module configured to determine multiple values corresponding to a number of qualifying parameters associated with the packet. The processing module may use the values to generate a selector value and may allocate a latency mode to the packet based on the selector value.

Abstract: A switch appliance includes a first network port for communication with a first node, where the first network port is configured to receive a packet, and a second network port for communication with a second node. The switch appliance further includes a first instrument port for communication with a first inline tool, a buffer, and a processing unit coupled to the first network port, the second network port, the first instrument port and the buffer. The processing unit is configured to determine whether a packet processing state has been set as an inline-tool processing state or a bypass state, and is configured to pass the packet to the second network port for transmission to the second node, and to store a copy of the packet in the buffer, if the packet processing state has not been set as the inline-tool processing state nor the bypass state.

Abstract: A first computer sends an electronic message transparently to a second computer of intended recipients of an electronic mail (e-mail), in response to the e-mail addresses of the intended recipients being entered, by a first end-user on the first computer, into a ‘To’ message header field of the e-mail. The second computer sends a Boolean value to the first computer, wherein one of the intended recipients is not available to respond to the e-mail. The first computer queries a repository to return to the first end-user contact information of backup contact entities to respond the e-mail for the intended recipients that are not available to respond to the e-mail. The first computer sends the e-mail to the backup contact entities that are available to respond to the e-mail. The first computer deletes the e-mail from each inbox of the backup contact entities that received but did not read the e-mail before the intended recipients read the e-mail.

Abstract: A system and method are disclosed which enable more effective email response authoring by contact center agents, for example, by automatically suggesting prototypical (entire) email responses to the human agent and interactive suggestion of next sentence candidates during the writing process. In one method, a customer inquiry is received and a latent topic prediction is generated, based on a word-based representation of the customer inquiry. A latent topic prediction is generated for an entire agent's reply to the customer inquiry as a function of the latent topic prediction generated for the customer inquiry. A further latent topic prediction is generated for a next sentence of the agent's reply as a function of a topic prediction for the next sentence which is generated with a prediction model that has been trained on annotated sentences of agent replies. Information is output to assist the agent, based on the topic predictions.