Abstract: Disclosed are systems and methods for steering network traffic over different access networks based on a classification of the network traffic. A first access network provides a first network path between a client device and a destination network and a second access network provides a second network path between the client device and the destination network. The access networks have different characteristics, such as latency or available bandwidth. A traffic classifier receives network messages and assigns classifications to the received network messages. A policy enforcer determines routes for individual network messages. The determined route is based at least in part on the classification assigned by the traffic classifier.

Abstract: Described embodiments provide systems and methods for selecting between modes for delivering or providing access to a file. A server can be configured to deliver the file via a plurality of modes of delivery including. A first mode of delivery can include downloading the file and a second mode of delivery can include delivering the file via a virtualized stream. The server can receive a request and initiate downloading of the file for a length of time. The server can identify an amount of available bandwidth and size of the file and determine a first time for the first mode of delivery. The first time can be compared to a second time to deliver the file via the second mode of delivery. The server can select between the first mode of delivery and the second mode of delivery and provide the file via the selected mode of delivery.

Abstract: A method for configuring a BMC IP address of a bare-metal server, comprising: deploying an Ironic service and a TFTP service on a management control platform, and registering bare-metal servers based on serial numbers of the bare-metal servers and corresponding BMC IP addresses; in response to a boot signal of the bare-metal server, fetching, by a PXE client, a PXE configuration file from the TFTP service to enable an IPA to be booted, reading an initialization configuration identifier in the PXE configuration file, and confirming whether to perform an initialization configuration for the booted bare-metal server; and in response to performing the initialization configuration for the booted bare-metal server, obtaining, by the IPA, the serial number of the booted bare-metal server from the Ironic service, and bare-metal-node information corresponding to the serial number, parsing out the BMC IP address from the bare- metal-node information, and configuring the BMC IP address.

Abstract: A network interface device has an interface configured to interface with a network. The interface is configured to at least one of receive data from the network and put data onto the network. The network interface device has an application specific integrated device with a plurality of data processing pipelines to process at least one of data which has been received from the network and data which is to be put onto said network and an FPGA arranged in a path parallel to the data processing pipelines.

Abstract: A computer implemented method of decoding a signal. The method includes receiving a signal (which may be an electromagnetic signal), sampling the received signal to generate an input waveform having magnitude and phase components, applying a transform operation to the input waveform to generate a first decoded signal, and outputting the first decoded signal. The transform operation includes pre-processing the input waveform to generate a mirrored inverted waveform and applying a continuous wavelet transform to the mirrored inverted waveform to generate the first decoded signal. This allows inversion of the frequency and temporal resolution of the continuous wavelet transform, thereby enabling improved temporal and frequency decoding of a signal. The method is particularly suitable for signal filters and filtering units.

Abstract: This disclosure provides communications methods and apparatuses. One method comprises: receiving, by a first entity of an adaptation layer in a network node, a data packet sent by an entity of a lower protocol layer of the adaptation layer, wherein the data packet comprising an adaptation layer header and an adaptation layer payload; and processing the data packet in response to determining whether the network node is a destination node that the data packet is routed to at the adaptation layer.

Abstract: A system for processing data, comprising a compute node having a first processor that is configured to receive a digital data message containing a request for computing services and to allocate processing resources on a private network as a function of the request. A smart network interface controller (NIC) management system operating on a second processor and configured to cause the second processor to select a smart NIC associated with the private network to allocate the smart NIC to the computing services. The smart NIC includes a processor that is configured to interface with a public network and to send and receive data over the public network associated with the computing services.

Abstract: An information acquiring method is provided. The method is applied to a first network side device. The method includes acquiring second capability information for processing data sent by at least one second network side device. According to embodiments of the present disclosure, in addition to communicating with a terminal, the first network side device can acquire the second capability information for processing data from at least one second network side device. Based on the second capability information, the first network side device can determine the capability of the second network side device for processing data. Further, when there is to-be-processed data in the first network side device, the first network side device can determine, according to the second capability information, which second network side device can process the data, and then can send the data to the second network side device for processing.

Abstract: Managing failover between information handling systems, including receiving, at an interface of a smart network interface card (smartNIC) of a primary information handling system, a packet, the primary information handling system communicatively couped to a secondary information handling system; determining whether the packet was transmitted by a network interface card (NIC) of the secondary information handling system; determining that the packet was transmitted by the NIC of the secondary information handling system, and in response, determining whether the packet is an address resolution protocol (ARP) request; determining that the packet is an ARP request, and in response, cloning a medium access control (MAC) address of the NIC of the secondary information handling system at the smartNIC of the primary information handling system; and forwarding the ARP request to a uplink connection.

Abstract: A computer program product, system, and computer implemented method for application-level redirect trapping and creation of NAT mapping to work with routing infrastructure for private connectivity in cloud and customer networks. The approach disclosed herein generally comprises a method of leveraging a reverse connection endpoint and IP address mapping controller to capture redirection messages from a private cloud or network (e.g., a service consumer network or a service consumer hybrid cloud). This allows at least the IP address mapping controller to manage a cloud networking infrastructure to provide for a service provider network (e.g., a public cloud) to support applications that overcome the isolation requirements of a private cloud or network to perform useful work. For example, without saddling the private cloud or network user with a heavy pre-configuration burden, the approach disclosed herein supports redirection to dynamically determined IP addresses at the private cloud or network.

Abstract: A method for detection of an anomaly in paths for routing data to an address prefix in a network composed of autonomous systems, where the paths are announced to a local autonomous system by neighboring autonomous systems. The method includes: a) receiving information representative of at least one part of the paths during a time interval; b) generating a matrix comprising elements representative of path segments respectively connecting pairs of autonomous systems located on the paths; c) repeating a) and b) to obtain a sequence of matrices according to b); d) analyzing a similarity between the matrices; and e) in a case where at least one of the matrices differs from other matrices, detecting a routing anomaly.

Abstract: Example computer-implemented methods, media, and systems for processing input/output (I/O) commands using block size aware polling are disclosed. One example method includes creating multiple polling queues and multiple interrupt queues in a transport drivers layer of a storage stack. A first I/O command is received from a core layer of the storage stack and by the transport drivers layer. A ratio of a total number of multiple small block size commands in the transport drivers layer to a total number of multiple outstanding I/O commands in the transport drivers layer is determined to be larger than a predetermined first threshold. In response to determining that the ratio is larger than the predetermined first threshold, the polling mode is applied to the first I/O command through the submission of the first I/O command to a first polling queue in the multiple polling queues.

Abstract: A method and system for communicating artificial intelligence (AI) tasks for a server chaining are presented. The method includes establishing a first connection between an AI client and a first AI server; encapsulating a request to process an AI task in at least one request data frame compliant with a communication protocol; and transporting the at least one request data frame over a network using a transport protocol over the first connection to the first AI server, wherein the first AI server spans the AI task over at least one second AI server, wherein the transport protocol provisions transport characteristics of the AI task and the transport protocol is different than the communication protocol, wherein AI task includes processing of a single compute graph thereby allow spanning the processing of the compute graph over one more AI servers.

Abstract: Virtual machines in a network may be isolated by encrypting transmissions between the virtual machines with keys possessed only by an intended recipient. Within a network, the virtual machines may be logically organized into a number of community-of-interest (COI) groups. Each COI may use an encryption key to secure communications within the COI, such that only other virtual machines in the COI may decrypt the message. Remote devices may gain access to virtual machines in a network through a virtual device relay. The virtual device relay receives data from the remote device, such as a tablet or cellular phone, and forwards the data to one of the virtual machines, when the virtual device relay shares a COI with the destination virtual machine.

Abstract: Disclosed herein are methods, systems, and processes for accepting and servicing interface agnostic validated unified platform queries. A request for shared data associated with web applications received from a user interface that is rendered by a client based on a specification is intercepted. The request is forwarded to a unified application programming interface (API) instead of a disparate unique API associated with the web applications. The request from the client is authenticated externally using the unified API and the request for the shared data is authorized to be displayed in the user interface based on the unified API instead of the specification.

Abstract: In a system with multiple host computers and one or more single-port non-volatile memory devices, a non-volatile memory switch receives memory transaction messages from different root complexes corresponding to the multiple host computers. Each of at least some of the memory transaction messages includes a host identifier that identifies a root complex from which the memory transaction was received. The non-volatile memory switch generates modified memory transaction messages at least by changing host identifiers within memory transaction messages to a common value indicative of a single root complex to present to the one or more single-port non-volatile memory devices the different root complexes as the single root complex. The non-volatile memory switch maintains associations of memory transaction messages with corresponding ones of the different root complexes, and sends the modified memory transaction messages to the one or more single-port non-volatile memory devices.

Abstract: Examples of techniques for service discovery of wired host devices are described. In an example, a controller may receive a service advertisement message comprising information of a service hosted by a wired host device in a network. Based on a header information of the service advertisement message the controller determines that the service advertisement message is from a switch. The controller maps the wired host device to a set of Wireless Access Points (WAPs) connected to the switch. Further, the controller may send identification information of the wired host device to a client device connected to one of the set of WAPs, in response to receiving a query for the service from the client device.

Abstract: In one embodiment, an agent executed by a device receives a profile for an application executed by the endpoint device that specifies one or more flow treatment policies for the application. The agent obtains performance metrics regarding a plurality of network interfaces of the endpoint device. The agent selects a particular network interface of the plurality of network interfaces by comparing the profile for the application to the performance metrics. The agent sends a packet for an application flow of the application over the particular network interface.

Abstract: In various embodiments, a computer-implemented method comprises acquiring status data for settings that control operations of the edge sensor device, transmitting the status data to a backend system, receiving a command to modify settings that control the operations of the edge sensor device, the settings specifying (i) one or more subscribers, and (ii) a data collection limit, in response to receiving the command, modifying the setting, acquiring sensor data associated with a physical device operating within a physical environment, where a sensor service collects the the sensor data up to the data collection limit, and the edge sensor device is located proximal to the physical device, inputting the sensor data into an onboard message broker for publishing, and transmitting, from the onboard message broker, the sensor data, where the sensor data is addressed to a first set of one or more subscribers of the onboard message broker.

Abstract: Examples include registering a device driver with an operating system, including registering available hardware offloads. The operating system receives a call to a hardware offload, inserts a binary filter representing the hardware offload into a hardware component and causes the execution of the binary filter by the hardware component when the hardware offload is available, and executes the binary filter in software when the hardware offload is not available.

Abstract: A base station is partitioned into a part executed on virtualization platform and a plurality of Radio Transceiver Points (R-TPs). It is established at least one CCord (Central Controller or Central Coordinator) of the base station, connected with the virtualized part of the base station and to the plurality of R-TPs by at least one control logical interface. Communications with the User Equipment (UE) implement a protocol split: on each R-TP, at least MAC (Medium Access Control) functions and a modulation function as part of PHY (Physical Layer) functions and on the virtualization platform at least PDCP (Packed Data Convergence Protocol) functions. CCord collects reports from each R-TP, with respect to usage status of belonging, allocated or assigned resources to each R-TP for downlink or uplink communications, and based on the collected reports selects at least one new serving R-TP of the UE.

Abstract: The described embodiments transfer an activity from a source electronic device to a companion electronic device. The source electronic device receives activity information describing an activity performed in a first application at the source electronic device, determines an activity identifier for the activity information, and broadcasts an activity advertisement comprising the activity identifier. Upon receiving the activity advertisement, the companion electronic device determines whether a second application that is associated with the first application is available at the companion electronic device. If the second application is available, the companion electronic device requests extended activity data from the source electronic device. The source electronic device responds by sending extended activity data from the first application to the companion electronic device.

Abstract: A traffic anomaly detector of a Peripheral Component Interconnect express (PCIe) system, including filters configured to filter headers of PCIe transaction layer packets (TLPs) based on respective filter criterion; a classifier configured to trigger an event based on one of the filter criterion or a logical combination of a plurality of the filter criteria; an event counter configured to count a number of the events; and a processor configured to detect, based on a value of the event counter, an anomaly in the PCIe TLP traffic.

Abstract: Embodiments of network processing resource management in computing devices are disclosed therein. In one embodiment, a method includes receiving a request from a network interface controller to perform network processing operations at a first core of a main processor for packets assigned by the network interface controller to a queue of a virtual port of the network interface controller. The method also includes determining whether the first core has a utilization level higher than a threshold when performing the network processing operations to effect processing and transmission of the packets. If the first core has a utilization level higher than the threshold, the method includes issuing a command to the network interface to modify affinitization of the queue from the first core to a second core having a utilization level lower than the threshold.

Abstract: A jobsite computer network has a local server connectable to a remote server via the internet, a network access transceiver connected to the local server, and a wireless device in wireless communication with the network access transceiver. The network access transceiver receives a message from the wireless device and sends a message to the local server. The local server utilizes information from the message to determine the location of the wireless device. The network access transceiver and/or the wireless device generate data which is sent to the remote server.

Abstract: A network interface controller (NIC) capable of efficient event management is provided. The NIC can be equipped with a host interface, a first memory device, and an event management module. During operation, the host interface can couple the NIC to a host device. The event management module can identify an event associated with an event queue stored in a second memory device of the host device. The event management module can insert, into a buffer, an event notification associated with the event. The buffer can be associated with the event queue and stored in the first memory device. If the buffer has met a release criterion, the event management module can insert, via the host interface, the aggregated event notifications into the event queue.

Abstract: The invention generally relates to a method and system for distribution of computational capacity for a plurality of end-devices in different localities using a decentralized edge architecture. The method and system clusters a plurality of moving nodes capable of offering highly virtualized computational and storage resources utilizing an Artificial Intelligence (AI) model. The clustering is performed by utilizing two prediction models: a mobility prediction model and a theoretical framework. The mobility prediction model learns timing and direction of movements such as mobility patterns of each of the plurality of moving nodes, as to ascertain computational capacity for the given locale at a certain time. The theoretical framework performs sequential to parallel conversion in learning, optimization and caching algorithms of the AI model under contingent circumstances.

Abstract: Wireless communications systems and methods related to full-duplex cooperative communication by a first device tethered with a second device communicating with a third device. With the tether, one device between the first device and the second device is designated or selected as a transmitter, and the other device as a receiver. In full-duplex operation, the transmitter may transmit outgoing data concurrent to the receiver receiving incoming data. In an example, one of the first and second devices may have a modem that is used to control the full-duplex communication. The modem controls both transmission and reception. In another example, both devices may have a modem. Each modem may assume responsibility for part of the full-duplex communication by performing some processing before either shipping received data, or transmitting outgoing data. The amount of processing may be split at a link layer. Interference cancellation may mitigate interference from full-duplex operation.

Abstract: The present disclosure provides a method of processing data based on QUIC protocol stack, the method including: obtaining a connection identifier of a data packet; determining a server program used for processing the data packet, wherein the server program comprises a plurality of worker threads, wherein QUIC protocol stacks run in the plurality of worker threads; determining a corresponding worker thread from the multiple worker threads based on the connection identifier; and distributing the data packet to the corresponding worker thread to make the QUIC protocol stack on the corresponding worker thread process the data packet. The disclosure further provides a system, a computing device and a computer-readable storage medium of processing data based on QUIC protocol stack.

Abstract: An information processing apparatus includes a communication unit capable of communicating with a server, a battery, and a controller. The controller is capable of executing processing regarding a predetermined function for which personal authentication is required at a time of use, detects a remaining charge level of the battery, and transmits, when the detected remaining charge level is less than a predetermined threshold value, a transfer request signal to the server, the transfer request signal requesting transfer of use authority of the function to another information processing apparatus.

Abstract: Provided is an onboard apparatus configured to connect to an other onboard apparatus via a first transmission path and a second transmission path, which are transmission paths of different types, the onboard apparatus comprising a control unit; and a transmission unit configured to transmit communication information to the other onboard apparatus, wherein the control unit, depending on contents of the communication information, selects at least one from among the first transmission path and the second transmission path as the transmission path to use to transmit the communication information and causes the transmission unit to transmit the communication information via a selected transmission path.

Abstract: A computing device includes a computing device controller hub and a plurality of parallelized nodes coupled to the computing device controller hub. Each node of the plurality of parallelized nodes includes a central processing module, a main memory, and at least one disk memory. The plurality of computing devices is operable to collectively execute query requests against at least one database table stored by the plurality of computing devices based on each node of each computing device performing corresponding operations independently from other nodes of the plurality of parallelized nodes.

Abstract: A cloud computing capacity management system can include a fine-grained admission control layer, a policy engine, and an enforcement layer. The fine-grained admission control layer can be configured to ingest capacity signals and create a capacity mitigation policy, based at least in part on the capacity signals, to protect available capacity of a cloud computing system for prioritized users. The capacity mitigation policy can be directed to users of the cloud computing system. The policy engine can be configured to control how the capacity mitigation policy is applied to the cloud computing system. The enforcement layer can be configured to handle incoming resource requests and to enforce resource limits based on the capacity mitigation policy as applied by the policy engine.

Abstract: A system receives a first request from a first instance of a network function associated with a first address. The system may determine the first address and, based at least in part on the first address, may identify a second address with which to respond to the first request. The system may then send, to the first instance of the network function, a response to the first request specifying the second address. The system may also receive a second request from a second instance of the network function associated with a third address. The system may determine a fourth address with which to respond to the second request, and may thereafter send a response to the second request to the second instance of the network function, with the response specifying the fourth address.

Abstract: A method, apparatus, and system are described. A method in an access point (AP) configured for medium access control (MAC) address designation (MAAD) is described. The AP is configured to wirelessly communicate with a station (STA). The method comprises determining a first MAC address of the STA, where the first MAC address is usable as a transmitter address (TA) of the STA for a subsequent association to the AP by the STA, and transmitting the first MAC address to the STA in one of a response action frame and a message of a multiple-message handshake process.

Abstract: A data processing system is provided. A host processing device supports a host transport engine operable to establish a first transport stream over a network with a remote peer. Device hardware comprises a device transport engine. The device transport engine is configured to monitor the first transport stream to determine a state of the first transport stream and in response to an indication from the host processing device perform transport processing of the first transport stream.

Abstract: A method is provided for generating process protocols from process data in an IT system with at least one extraction node, at least one storage node, and at least one transformation node. The process may be stored in a client system of the IT system. An application catalog is run on a server system of the IT system, and application templates are managed using the application catalog, each application template having an installation program and configuration data. The installation program is used to instantiate, configure, and run each application template on the server system as an application. The applications being run are monitored and managed by a runtime component on the server system, and the process protocols are ascertained according to a number of control instructions run on nodes, the configuration data of each application comprising information on which control instructions are run on which nodes.

Abstract: An illustrative application executing on a first device establishes a first network connection between the first device and a collaboration platform server managing a set of components for a collaboration platform providing voice communication services with respect to a virtual space. Using the first network connection, the application: 1) sets a first location within the virtual space at which a first voice component is positioned, and 2) determines a second location within the virtual space at which a second voice component is positioned. The application receives, by way of a second network connection between the first and second devices, audio data representing voice communication originating at the second device. Based on the first and second locations, the application generates a spatialized rendering of the audio data for presentation to a user. Corresponding methods and systems are also disclosed.

Abstract: Apparatus and methods are disclosed herein for remote, direct memory access (RDMA) technology that enables direct memory access from one host computer memory to another host computer memory over a physical or virtual computer network. In one example, RDMA functionality is provided by combining a host executing instructions for RDMA software applications with dedicated hardware accelerator, where the host has access to substantially all RDMA connection context (e.g., configuration, status, state machine, counters, etc.) and a hardware accelerator stores a portion of the context information for a set of currently active transactions. The hardware accelerator derives a transaction identifier from header information in received RDMA packets and performs a local RDMA operation using at least a portion of the received data, a destination address of the location RDMA operation being based at least in part on a memory address determined using the transaction identifier.

Abstract: Technologies for processing network packets by a host interface of a network interface controller (NIC) of a compute device. The host interface is configured to retrieve, by a symmetric multi-purpose (SMP) array of the host interface, a message from a message queue of the host interface and process, by a processor core of a plurality of processor cores of the SMP array, the message to identify a long-latency operation to be performed on at least a portion of a network packet associated with the message. The host interface is further configured to generate another message which includes an indication of the identified long-latency operation and a next step to be performed upon completion. Additionally, the host interface is configured to transmit the other message to a corresponding hardware unit scheduler as a function of the subsequent long-latency operation to be performed. Other embodiments are described herein.

Abstract: A method for optimizing internet traffic from a plurality of local area networks (LANs) to an internet based server connected to internet includes instructions for analyzing an internet data by a gateway computer located in a first LAN of the plurality of LANs to determine a data that is latency sensitive and a data that is not latency sensitive; instructions for sending network steering data, by the gateway computer, to a Central Processing Matrix (CPM); instructions for sending data, by a plurality of customer premises equipment (CPE) located in a second plurality of LANs in the plurality of LANs to the CPM through a listener/reporter located outside of the plurality of LANs; instructions for receiving the network steering data by the listener/reporter from the CPM; and instructions for routing the data that is latency sensitive outside of the second plurality of LANs to the internet based server.

Abstract: A method, system and computer-usable medium are disclosed for operating an endpoint agent at an endpoint device. Certain embodiments include a computer-implemented method for operating an endpoint agent at an endpoint device, including: operating the endpoint agent to selectively subscribe to events corresponding to activities occurring at an endpoint platform; processing events received from a message bus by the endpoint agent, where the events processed by the endpoint agent are events to which the endpoint agent has subscribed; and communicating, to a service, information corresponding to the events processed by the endpoint agent. Other embodiments of this aspect of the invention may include corresponding stand-alone and/or network computer systems, apparatus, and computer programs recorded on one or more computer storage devices, each configured to perform one or more of these actions.

Abstract: Systems and methods for a virtual layer-2 network are described herein. The method can include providing a virtual Layer 3 network in a virtualized cloud environment. The virtual Layer 3 network can be hosted by an underlying physical network. The method can include providing a virtual Layer 2 network in the virtualized cloud environment. The virtual Layer 2 network can be hosted by the underlying physical network.

Abstract: A plurality of session signatures and a plurality of inference flags are stored in association with each other, each session signature representing a combination of flag values, each flag value signifying a key event. A plurality of session events are logged, each session event issuing from one of the remote client devices and having a corresponding session ID and associated data. The session events are grouped by their corresponding session IDs to create a plurality of session records, each session record including flag values for a plurality of key events. A session signature is identified for each session record from the stored session signatures based on the key event flag values of the session record, and a value of an inference flag associated with the identified session signature is obtained from a storage device. A predetermined operation is performed based on the obtained value of the inference flag.

Abstract: A computing device receives a modified first image frame from a client device, wherein the client device stores the first image frame. The computing device generates a second image frame that corresponds to the modified first image frame with a watermark. The computing device transmits, to a third-party application executing at the computing device, the second image frame and generates an instruction for modifying the first image frame, the instruction based on an altered version of the second image frame that is identified by the watermark, the second image frame altered by the third-party application. The computing device transmits, to the client device, the instruction for modifying the first image frame.

Abstract: Examples described herein include systems and methods for controlling access to a server, such as an email server or a gateway, in situations where the identity of the requesting device is unknown or where the user device accesses the server using an unknown or unmanaged application. In one example, the system can utilize a user authentication credential included in the request to identify other devices belonging to the user that happen to be enrolled with the system. An out-of-band message can be sent to those enrolled devices, requesting confirmation from the user and, in conjunction with an authentication token, allowing the system to trust the previously unknown device. In the example of an unmanaged application attempting to access an email server, the system can confirm compliance of the requesting device and issue an authentication token that, along with an appropriate command sent to the email server, provides access.

Abstract: One variation of a system for implementing a key-value data store includes one or more processors, storage media and instructions stored in the storage media which, when executed by the system cause the system to: receive a request store a particular key-value item; request a first networked distributed data storage system to store the particular key-value item; based on a determination that a set of one or more offload criteria is satisfied: retrieve a first set of key-value items from the first networked distributed data storage system, and request a second networked distributed data storage system to store the first set of key-value items in a first set of one or more data objects. The first networked distributed data storage system can have a lower data write latency and a higher data storage cost than the second networked distributed data storage system.

Abstract: A power and data transfer method includes providing a cable assembly for both power and data transmission, transferring power through at least two power conductors of the cable assembly, and transferring data through at least two data conductors of the cable assembly using a serial communication protocol.

Abstract: Exemplary embodiments provide a refrigeration system having an interior space cooled by a plurality of cooling. Each cooling unit is capable of operating either synchronously when in communication with a control panel or under independent operation. Each cooling unit is modularly and replaceable without the use of tools by means of a quick connect system. The cooling units use a heat exchanger cooled by chilled water and make use of an electronic super heat control and electronic expansion valve to regulate the flow of refrigerant for improved efficiency.

Abstract: Disclosed are systems and methods for steering network traffic over different access networks based on a classification of the network traffic. A first access network provides a first network path between a client device and a destination network and a second access network provides a second network path between the client device and the destination network. The access networks have different characteristics, such as latency or available bandwidth. A traffic classifier receives network messages and assigns classifications to the received network messages. A policy enforcer determines routes for individual network messages. The determined route is based at least in part on the classification assigned by the traffic classifier.