Patent Application
20190044869
Modern computing devices have become ubiquitous tools for personal, business, and social uses. As such, many modern computing devices are capable of connecting to various data networks, including the Internet, to transmit and receive data communications over the various data networks at varying rates of speed. To facilitate communications between computing devices, the data networks typically include one or more network computing devices (e.g., compute servers, storage servers, etc.) to route communications (e.g., via switches, routers, etc.) that enter/exit a network (e.g., north-south network traffic) and between network computing devices in the network (e.g., east-west network traffic). Such data networks typically have included complex, large-scale computing environments, such as high-performance computing (HPC) and cloud computing environments. Accordingly, it should be understood that network traffic has to be transmitted between and processed by computing devices with minimal latency. Depending on the specific purpose of each received network packet, the receiving computing device processes the data in a certain way. To identify how to process the data, the network packets are categorized or otherwise classified according to “flows,” which define operations and/or other rules for the processing of the network packets. As speed is a critical feature in computing systems, a variety of mechanisms have been employed to increase the speed at which such packet flow classifications occur.
Once such classification mechanism is the use of hash tables. Hash tables are often used to improve the speed of data lookup operations when it is necessary to sift through large amounts of data to identify a match. In particular, hash tables (e.g. associative arrays) utilize hash functions to deterministically map data of variable length to data of fixed size (i.e., the signature). Alternative classification mechanisms have been developed as well, including various algorithms and implementations for flow classification proposed by either industry or the research community to improve performance, such as tuple space search based algorithms, tree-based search algorithms, highly optimized decision tree based algorithms, combinations of tree and hash lookup tables, etc. However, present flow classification solutions are generally not optimal for all use cases. For example, some solutions improve lookup performance but penalize update speed, while some hash table based models, in general, provide efficient lookup and update, but are not typically considered to be memory “friendly”.
The concepts described herein are illustrated by way of example and not by way of limitation in the accompanying figures. For simplicity and clarity of illustration, elements illustrated in the figures are not necessarily drawn to scale. Where considered appropriate, reference labels have been repeated among the figures to indicate corresponding or analogous elements.
While the concepts of the present disclosure are susceptible to various modifications and alternative forms, specific embodiments thereof have been shown by way of example in the drawings and will be described herein in detail. It should be understood, however, that there is no intent to limit the concepts of the present disclosure to the particular forms disclosed, but on the contrary, the intention is to cover all modifications, equivalents, and alternatives consistent with the present disclosure and the appended claims.
References in the specification to “one embodiment,” “an embodiment,” “an illustrative embodiment,” etc., indicate that the embodiment described may include a particular feature, structure, or characteristic, but every embodiment may or may not necessarily include that particular feature, structure, or characteristic. Moreover, such phrases are not necessarily referring to the same embodiment. Further, when a particular feature, structure, or characteristic is described in connection with an embodiment, it is submitted that it is within the knowledge of one skilled in the art to effect such feature, structure, or characteristic in connection with other embodiments whether or not explicitly described. Additionally, it should be appreciated that items included in a list in the form of “at least one of A, B, and C” can mean (A); (B); (C): (A and B); (A and C); (B and C); or (A, B, and C). Similarly, items listed in the form of “at least one of A, B, or C” can mean (A); (B); (C): (A and B); (A and C); (B and C); or (A, B, and C).
The disclosed embodiments may be implemented, in some cases, in hardware, firmware, software, or any combination thereof. The disclosed embodiments may also be implemented as instructions carried by or stored on one or more transitory or non-transitory machine-readable (e.g., computer-readable) storage media, which may be read and executed by one or more processors. A machine-readable storage medium may be embodied as any storage device, mechanism, or other physical structure for storing or transmitting information in a form readable by a machine (e.g., a volatile or non-volatile memory, a media disc, or other media device).
In the drawings, some structural or method features may be shown in specific arrangements and/or orderings. However, it should be appreciated that such specific arrangements and/or orderings may not be required. Rather, in some embodiments, such features may be arranged in a different manner and/or order than shown in the illustrative figures. Additionally, the inclusion of a structural or method feature in a particular figure is not meant to imply that such feature is required in all embodiments and, in some embodiments, may not be included or may be combined with other features.
Referring now to
The network appliance 106 is configured to receive network packets (e.g., Ethernet frames, messages, etc.), such as may be received from the source compute devices 102 via the network 104, perform some level of processing (e.g., one or more processing operations) on at least a portion of the data associated with the received network packets, and either drop or transmit each received network packet to a destination (e.g., to another network appliance in the same or alternative network, based to the source compute device 102, etc.). To perform the processing operations, the network appliance 106 may be configured to leverage virtualization technologies to provide one or more virtualized network functions (VNFs) (e.g., executing on one or more virtual machines (VMs), in one or more containers, etc.) to execute network services on commodity hardware. Such network services may include any type of network service, including firewall services, network address translation (NAT) services, domain name system (DNS) services, load-balancing services, deep packet inspection (DPI) services, transmission control protocol (TCP) optimization services, cache management services, Internet Protocol (IP) address management services, etc.
In network function virtualization (NFV) architecture, a VNF is configured to handle specific network functions that run in one or more VMs on top of hardware networking infrastructure traditionally carried out by proprietary, dedicated hardware, such as routers, switches, servers, cloud computing systems, etc. In other words, each VNF may be embodied as one or more VMs configured to execute corresponding software or instructions to perform a virtualized task. It should be understood that a VM is a software program or operating system that not only exhibits the behavior of a separate computer, but is also capable of performing tasks such as running applications and programs like a separate computer. A VM, commonly referred to as a “guest,” is typically configured to run a dedicated operating system on shared physical hardware resources of the device on which the VM has been deployed, commonly referred to as a “host.” It should be appreciated that multiple VMs can exist within a single host at a given time and that multiple VNFs may be executing on the network appliance 106 at a time (e.g., in parallel).
In use, as will be described in further detail below, the network appliance 106 is configured to identify a flow associated with each received network packet. To identify the flow for each received network packet, the network appliance 106, or more particularly a virtual router (see, e.g., the virtual router 212 of
Furthermore, present virtual router technologies are typically designed with a fixed algorithm that is configured to perform the “best” on average (e.g., for a targeted usage at the time of virtual router design). However, unlike those fixed algorithm virtual router technologies, the virtual router as described herein is configured to adapt to both traffic patterns and rule patterns, as well as other requirements, during run-time. In other words, the virtual router as disclosed herein is configured to function as an adaptive virtual router that can dynamically choose the optimal flow classification algorithm design internally to optimize the throughput for run time usages (e.g., traffic patterns, rule patterns, etc.). To do so, the virtual router includes a suite of flow classification algorithm designs that provide optimal performance for different usages and is configured to apply an approach to adaptively choose the appropriate classification algorithm design to optimize performance. Accordingly, the virtual router is configured to select the appropriate flow classification algorithm design internally, without requiring manual interference.
The network appliance 106 may be embodied as any type of computation or computing device capable of performing the functions described herein, including, without limitation, a server (e.g., stand-alone, rack-mounted, blade, etc.), a switch (e.g., a disaggregated switch, a rack-mounted switch, a standalone switch, a fully managed switch, a partially managed switch, a full-duplex switch, and/or a half-duplex communication mode enabled switch), a sled (e.g., a compute sled, a storage sled, an accelerator sled, a memory sled, etc.) a router, a web appliance, a processor-based system, and/or a multiprocessor system. In some embodiments, the network appliance 106 may be embodied as a distributed computing system, a composable computing system (e.g., composable by an orchestrator), or some other type of computing system in which resources can be pooled. In such embodiments, the network appliance 106 may be embodied as more than one computing device in which each computing device is configured to pool resources and perform at least a portion of the functions described herein.
As shown in
The compute engine 108 may be embodied as any type of device or collection of devices capable of performing the various compute functions as described herein. In some embodiments, the compute engine 108 may be embodied as a single device such as an integrated circuit, an embedded system, a field-programmable-array (FPGA), a system-on-a-chip (SOC), an application specific integrated circuit (ASIC), reconfigurable hardware or hardware circuitry, or other specialized hardware to facilitate performance of the functions described herein. Additionally, in some embodiments, the compute engine 108 may include, or may otherwise be embodied as, one or more processors 110 (i.e., one or more central processing units (CPUs)) and memory 112.
The processor(s) 110 may be embodied as any type of processor(s) capable of performing the functions described herein. For example, the processor(s) 110 may be embodied as one or more single-core processors, multi-core processors, digital signal processors (DSPs), microcontrollers, or other processor(s) or processing/controlling circuit(s). In some embodiments, the processor(s) 110 may be embodied as, include, or otherwise be coupled to an FPGA (e.g., reconfigurable circuitry), an ASIC, reconfigurable hardware or hardware circuitry, or other specialized hardware to facilitate performance of the functions described herein.
The memory 112 may be embodied as any type of volatile or non-volatile memory or data storage capable of performing the functions described herein. It should be appreciated that the memory 112 may include main memory (i.e., a primary memory) and/or cache memory (i.e., memory that can be accessed more quickly than the main memory). Volatile memory may be a storage medium that requires power to maintain the state of data stored by the medium. Non-limiting examples of volatile memory may include various types of random access memory (RAM), such as dynamic random access memory (DRAM) or static random access memory (SRAM).
The compute engine 108 is communicatively coupled to other components of the network appliance 106 via the I/O subsystem 114, which may be embodied as circuitry and/or components to facilitate input/output operations with the processor 110, the memory 112, and other components of the network appliance 106. For example, the I/O subsystem 114 may be embodied as, or otherwise include, memory controller hubs, input/output control hubs, integrated sensor hubs, firmware devices, communication links (e.g., point-to-point links, bus links, wires, cables, light guides, printed circuit board traces, etc.), and/or other components and subsystems to facilitate the input/output operations. In some embodiments, the I/O subsystem 114 may form a portion of a SoC and be incorporated, along with one or more of the processor 110, the memory 112, and other components of the network appliance 106, on a single integrated circuit chip.
The one or more data storage devices 116 may be embodied as any type of storage device(s) configured for short-term or long-term storage of data, such as, for example, memory devices and circuits, memory cards, hard disk drives, solid-state drives, or other data storage devices. Each data storage device 116 may include a system partition that stores data and firmware code for the data storage device 116. Each data storage device 116 may also include an operating system partition that stores data files and executables for an operating system.
The communication circuitry 118 may be embodied as any communication circuit, device, or collection thereof, capable of enabling communications between the network appliance 106 and other computing devices, such as the source compute device 102, as well as any network communication enabling devices, such as an access point, network switch/router, etc., to allow communication over the network 104. Accordingly, the communication circuitry 118 may be configured to use any one or more communication technologies (e.g., wireless or wired communication technologies) and associated protocols (e.g., Ethernet, Bluetooth®, Wi-Fi®, WiMAX, LTE, 5G, etc.) to effect such communication.
It should be appreciated that, in some embodiments, the communication circuitry 118 may include specialized circuitry, hardware, or combination thereof to perform pipeline logic (e.g., hardware algorithms) for performing the functions described herein, including processing network packets (e.g., parse received network packets, determine destination computing devices for each received network packets, forward the network packets to a particular buffer queue of a respective host buffer of the network appliance 106, etc.), performing computational functions, etc.
In some embodiments, performance of one or more of the functions of communication circuitry 118 as described herein may be performed by specialized circuitry, hardware, or combination thereof of the communication circuitry 118, which may be embodied as a SoC or otherwise form a portion of a SoC of the network appliance 106 (e.g., incorporated on a single integrated circuit chip along with a processor 110, the memory 112, and/or other components of the network appliance 106). Alternatively, in some embodiments, the specialized circuitry, hardware, or combination thereof may be embodied as one or more discrete processing units of the network appliance 106, each of which may be capable of performing one or more of the functions described herein.
The illustrative communication circuitry 118 includes the NIC 120, which may also be referred to as a host fabric interface (HFI) in some embodiments (e.g., high performance computing (HPC) environments). The NIC 120 may be embodied as any type of firmware, hardware, software, or any combination thereof that facilities communications access between the network appliance 106 and a network (e.g., the network 104). For example, the NIC 120 may be embodied as one or more add-in-boards, daughtercards, network interface cards, controller chips, chipsets, or other devices that may be used by the network appliance 106 to connect with another compute device (e.g., the source compute device 102).
In some embodiments, the NIC 120 may be embodied as part of a SoC that includes one or more processors, or included on a multichip package that also contains one or more processors. Additionally or alternatively, in some embodiments, the NIC 120 may include one or more processing cores (not shown) local to the NIC 120. In such embodiments, the processing core(s) may be capable of performing one or more of the functions described herein. In some embodiments, the NIC 120 may additionally include a local memory (not shown). In such embodiments, the local memory of the NIC 120 may be integrated into one or more components of the network appliance 106 at the board level, socket level, chip level, and/or other levels.
The one or more peripheral devices 122 may include any type of device that is usable to input information into the network appliance 106 and/or receive information from the network appliance 106. The peripheral devices 122 may be embodied as any auxiliary device usable to input information into the network appliance 106, such as a keyboard, a mouse, a microphone, a barcode reader, an image scanner, etc., or output information from the network appliance 106, such as a display, a speaker, graphics circuitry, a printer, a projector, etc. It should be appreciated that, in some embodiments, one or more of the peripheral devices 122 may function as both an input device and an output device (e.g., a touchscreen display, a digitizer on top of a display screen, etc.). It should be further appreciated that the types of peripheral devices 122 connected to the network appliance 106 may depend on, for example, the type and/or intended use of the network appliance 106. Additionally or alternatively, in some embodiments, the peripheral devices 122 may include one or more ports, such as a USB port, for example, for connecting external peripheral devices to the network appliance 106.
The source compute device 102 may be embodied as any type of computation or computer device capable of performing the functions described herein, including, without limitation, a smartphone, a mobile computing device, a tablet computer, a laptop computer, a notebook computer, a computer, a server (e.g., stand-alone, rack-mounted, blade, etc.), a sled (e.g., a compute sled, an accelerator sled, a storage sled, a memory sled, etc.), a network appliance (e.g., physical or virtual), a web appliance, a distributed computing system, a processor-based system, and/or a multiprocessor system. While not illustratively shown, it should be appreciated that source compute device 102 includes similar and/or like components to those of the illustrative network appliance 106. As such, figures and descriptions of the like/similar components are not repeated herein for clarity of the description with the understanding that the description of the corresponding components provided above in regard to the network appliance 106 applies equally to the corresponding components of the source compute device 102. Of course, it should be appreciated that the computing devices may include additional and/or alternative components, depending on the embodiment.
The network 104 may be embodied as any type of wired or wireless communication network, including but not limited to a wireless local area network (WLAN), a wireless personal area network (WPAN), an edge network (e.g., a multi-access edge computing (MEC) network), a fog network, a cellular network (e.g., Global System for Mobile Communications (GSM), Long-Term Evolution (LTE), 5G, etc.), a telephony network, a digital subscriber line (DSL) network, a cable network, a local area network (LAN), a wide area network (WAN), a global network (e.g., the Internet), or any combination thereof. It should be appreciated that, in such embodiments, the network 104 may serve as a centralized network and, in some embodiments, may be communicatively coupled to another network (e.g., the Internet). Accordingly, the network 104 may include a variety of other virtual and/or physical network computing devices (e.g., routers, switches, network hubs, servers, storage devices, compute devices, etc.), as needed to facilitate communication between the network appliance 106 and the source compute device 102, which are not shown to preserve clarity of the description.
Referring now to
Additionally, in some embodiments, one or more of the illustrative components may form a portion of another component and/or one or more of the illustrative components may be independent of one another. Further, in some embodiments, one or more of the components of the environment 200 may be embodied as virtualized hardware components or emulated architecture, which may be established and maintained by the NIC 120, the compute engine 108, and/or other software/hardware components of the network appliance 106. It should be appreciated that the network appliance 106 may include other components, sub-components, modules, sub-modules, logic, sub-logic, and/or devices commonly found in a computing device (e.g., device drivers, interfaces, etc.), which are not illustrated in
In the illustrative environment 200, the network appliance 106 additionally includes network traffic data 202, algorithm design data 204, and algorithm performance data 206, each of which may be accessed by the various components and/or sub-components of the network appliance 106. Further, each of the network traffic data 202, the algorithm design data 204, and the algorithm performance data 206 may be accessed by the various components of the network appliance 106. Additionally, it should be appreciated that in some embodiments the data stored in, or otherwise represented by, each of the network traffic data 202, the algorithm design data 204, and the algorithm performance data 206 may not be mutually exclusive relative to each other. For example, in some implementations, data stored in the network traffic data 202 may also be stored as a portion of one or more of the algorithm design data 204 and/or the algorithm performance data 206, or in another alternative arrangement. As such, although the various data utilized by the network appliance 106 is described herein as particular discrete data, such data may be combined, aggregated, and/or otherwise form portions of a single or multiple data sets, including duplicative copies, in other embodiments.
The network traffic ingress/egress manager 208, which may be embodied as hardware, firmware, software, virtualized hardware, emulated architecture, and/or a combination thereof as discussed above, is configured to receive inbound and route/transmit outbound network traffic. To do so, the network traffic ingress/egress manager 208 is configured to facilitate inbound/outbound network communications (e.g., network traffic, network packets, network flows, etc.) to and from the network appliance 106. For example, the network traffic ingress/egress manager 208 is configured to manage (e.g., create, modify, delete, etc.) connections to physical and virtual network ports (i.e., virtual network interfaces) of the network appliance 106 (e.g., via the communication circuitry 118), as well as the ingress/egress buffers/queues associated therewith. In some embodiments, information associated with the received network traffic (e.g., an associated descriptor, a pointer to a location in memory in which at least a portion of the received network traffic has been stored, a characteristic of the received network traffic, etc.) and/or network-related information (e.g., a number of active flows, a number of rules, a number of wild card formats, a required update rate, etc.) may be stored in the network traffic data.
The network traffic pattern analyzer 210, which may be embodied as hardware, firmware, software, virtualized hardware, emulated architecture, and/or a combination thereof as discussed above, is configured to analyze network traffic received by the network appliance 106 and identify any patterns associated with the received network traffic. For example, the network traffic pattern analyzer 210 may be configured to identify access patterns based on information collected that is associated with received network traffic, such as traffic patterns (e.g., patterns of distribution of the network traffic to each active flow), rule set patterns (e.g., access patterns of rules, rule format distribution, etc.), wild card patterns (e.g., access patterns to wild cards), access patterns of flows (e.g., bursty or non-burtsy, sequential or random, etc.), and/or the like.
The virtual router 212, which may be embodied as hardware, firmware, software, virtualized hardware, emulated architecture, and/or a combination thereof as discussed above, is configured to perform routing/switching functionality of the network appliance 106 using a flow classification algorithm. To do so, the illustrative virtual router 212 includes a candidate classification algorithm design picker 214, a classification algorithm deployment manager 216, a classification algorithm performance monitor 218, and a classification algorithm design selector 220, each of which may be embodied as hardware, firmware, software, virtualized hardware, emulated architecture, and/or a combination thereof. For example, in some embodiments, one or more of the candidate classification algorithm design picker 214, the classification algorithm deployment manager 216, the classification algorithm performance monitor 218, and the classification algorithm design selector 220 may be embodied as circuitry or collection of electrical devices (e.g., candidate classification algorithm design picker circuitry 214, classification algorithm deployment management circuitry 216, classification algorithm performance monitoring circuitry 218, classification algorithm design selector circuitry 220, etc.).
The candidate classification algorithm design picker 214 is configured to identify a set of candidate classification algorithms from a plurality of candidate classification algorithms. In other words, the candidate classification algorithm design picker 214 is configured to select a set of candidate classification algorithms that can be used (e.g., tested, ranked, and selected) for the classification flow operation (e.g., to be performed by the virtual router). It should be appreciated that multiple high-level classification algorithm designs may be selectable by the candidate classification algorithm design picker 214. For example, the classification algorithm designs may include tuple space search algorithms in which headers of the received network packets go through each table to do masking and searching to look for the rules, multi-bit trie based search algorithms for IP prefix matching, exact match hash tables, and/or a combination thereof (e.g., a combination of a hash table cache and a tuple space search).
It should be appreciated that each classification algorithm design may have several sets of configurations. In an illustrative example, a multi-bit trie can have different stride sizes. Additionally, the hash table can have different numbers of way associativity and hash functions which will affect the lookup speed and table load factor. Accordingly, it should be further appreciated that such options can generate multiple candidate classification algorithm designs. For example, a first classification algorithm design may have two different classification algorithm design configurations, a second classification algorithm design may have thiee different classification algorithm design configurations, and a third classification algorithm design may have two different classification algorithm design configurations, for a total of seven different classification algorithm design configurations. In some embodiments, information associated with the classification algorithm designs may be stored in the algorithm design data 204.
The classification algorithm deployment manager 216 is configured to deploy the classification algorithms to one or more processors (e.g., one or more of the processors 110 of
The classification algorithm performance monitor 218 is configured to collect performance data associated with the implemented classification algorithm design. The performance data may include any type of resource usage data that is usable to rank a performance of one classification algorithm against another classification algorithm. For example, the classification algorithm performance monitor 218 may be configured to collect throughput data, compute data, power usage data, etc. In some embodiments, the collected performance data may be stored in the algorithm performance data 206.
The classification algorithm design selector 220 is configured to select a preferred classification algorithm design for flow classification that is to be deployed (e.g., by the classification algorithm deployment manager 216) across one or more processors/cores. To do so, the classification algorithm design selector 220 is configured to rank each candidate classification algorithm based on the performance data (e.g., as may be collected by the classification algorithm performance monitor 218) collected for each tested candidate classification algorithm. To select the preferred classification algorithm design, the classification algorithm design selector 220 is configured to select the highest ranking algorithm that exceeds a predetermined performance threshold. It should be appreciated that the classification algorithm design selector 220 may also apply a level of confidence to each tested candidate classification algorithm, relative to the performance data collected, such that the classification algorithm design selector 220 does not select the preferred classification algorithm design until a sufficient level of confidence has been achieved for each tested candidate classification algorithm.
Referring now to
In block 308, the virtual router 212 deploys the selected default classification algorithm on one or more processors/cores. In block 310, the virtual router 212 monitors a performance level of the classification algorithm. For example, in block 312, the virtual router 212 may monitor a throughput of the classification algorithm in operation. It should be appreciated that, in other embodiments, additional and/or alternative telemetry metrics may be monitored, such as latency, jitter, etc. In block 314, the virtual router 212 determines whether the monitored performance level is less than a performance threshold (e.g., a minimum performance level that may be assigned by a controller/administrator). If not, the method 300 returns to block 310 to continue monitoring the performance level of the classification algorithm; otherwise, the method 300 proceeds to block 316.
In block 316, the virtual router 212 identifies a set of candidate classification algorithms from the available classification algorithms (e.g., based on the classification algorithm designs and different configurations for each classification algorithm design). In block 318, the virtual router 212 deploys at least one of the identified set of candidate classification algorithms. For example, in block 320, the virtual router 212 may deploy a candidate classification algorithm on the same processor/core as the default classification algorithm was deployed. Additionally or alternatively, in block 322, the virtual router 212 may deploy one or more candidate classification algorithms on a corresponding one or more other processors/cores.
In block 324, the virtual router 212 monitors a performance level of each of the one or more deployed classification algorithms. For example, in block 326, the virtual router 212 may monitor a throughput of each deployed classification algorithm in operation across the classified flows. As noted previously, it should be appreciated that, in other embodiments, additional and/or alternative telemetry metrics may be monitored, such as latency, jitter, etc. In block 328, shown in
If all of the candidate algorithms have not been tested, or sufficiently tested to an acceptable degree of confidence, the method 300 returns to block 318, in which the virtual router 212 deploys one or more of the candidate classification algorithms. Otherwise, if all of the candidate algorithms have been tested, and to an acceptable degree of confidence, if applicable, the method 300 proceeds to block 330. In block 330, the virtual router 212 ranks each tested classification algorithm based on the associated performance level. In block 332, the virtual router 212 selects the candidate classification algorithm with the highest ranked performance level (e.g., the highest throughput level).
In block 334, the virtual router 212 determines whether the performance level of the selected candidate classification algorithm exceeds a performance threshold (e.g., a minimum throughput threshold). If so, method 300 branches to block 336, in which the virtual router 212 deploys the selected classification algorithm on each of the applicable virtual routing processor(s)/core(s) configured to perform the flow classification operations. In some embodiments, in block 338, the virtual router 212 may additionally adjust the performance threshold relative to the performance level of the selected candidate classification algorithm.
However, referring back to block 334, if the virtual router 212 determines that the performance level of the selected candidate classification algorithm does not exceed the performance threshold, the method 300 branches to block 340. In block 340, the virtual router 212 transmits a report to a controller/administrator that indicates the performance level of the selected candidate classification algorithm does not exceed the performance threshold. Accordingly, in block 342, the virtual router 212 may automatically lower the performance threshold relative to the performance level of the selected candidate classification algorithm (e.g., lower the performance threshold to a value less than the performance level of the selected candidate classification algorithm).
Referring now to
The method 400 begins in block 402, in which the network appliance 106 determines whether to perform the offline profiling. If so, the method 400 advances to block 404, in which the network appliance 106 collects input parameters to perform the offline profiling. For example, in block 406, the network appliance 106 can collect flow information, classification rules, update patterns, etc., related to the expected network traffic. In block 408, the network appliance 106 identifies a set of candidate classification algorithms. In block 410, the network appliance 106 performs offline profiling for each of the identified set of candidate classification algorithms using automated scripts based on the collected input parameters.
In block 412, the network appliance 106 measures lookup and update performance levels for each candidate classification algorithm. In block 414, the network appliance 106 applies one or more machine learning algorithms to train a performance model for all of the candidate classification algorithms under various conditions (e.g., given various network traffic scenarios/patterns). It should be appreciated that, with offline profiling, extensive experiments (e.g., provided by automatic scripts) can be run with different input parameters (e.g., flows, rules, update pattern, etc.), and measure the lookup and update performance for all underlying algorithm design candidates. As such, these data sets are used as the training set for the machine learning algorithm to learn the performance for given algorithm design with given usages (indicated by a set of input parameters). Furthermore, by employing a machine learning algorithm (e.g., Random Forest Regression (RFR)) the performance with different algorithm designs for given scenarios can be predicted.
In block 416, the network appliance 106 constructs a learned data structure (e.g., a lookup table, a state machine, etc.) based the performance model. To construct the learned data structure, in block 418, the network appliance 106 uses various features as inputs, including network traffic patterns, rule patterns, etc. Additionally, in block 420, the network appliance 106 constructs the learned data structure using a predicted performance for each of the different candidate classification algorithms as output. In an illustrative example, to mitigate the costs of real-time machine learning overhead, a lookup table can be constructed using the offline learning process by using impact factors (e.g., rules, flows, etc.) as inputs and a predicted performance (e.g., based on the performance model) for different algorithm designs as outputs. In some embodiments, during real-time virtual router operation, the virtual router may continuously sample traffic characteristics. Accordingly, in such embodiments, when the performance level drops below the performance threshold, the virtual router can perform a lookup operation on the lookup table and obtain the best algorithm design to achieve optimal performance, as opposed to performing the real-time performance testing for different algorithm designs as described above.
Illustrative examples of the technologies disclosed herein are provided below. An embodiment of the technologies may include any one or more, and any combination of, the examples described below.
Example 1 includes a network appliance for classifying network flows using adaptive virtual routing, the network appliance comprising one or more processors; and virtual switch operation mode controller circuitry to identify a set of candidate classification algorithms from a plurality of classification algorithm designs to perform a flow classification operation; deploy each of the candidate classification algorithms to a processor of the one or more processors; monitor a performance level of each of the deployed candidate classification algorithms; identify a candidate classification algorithm of the deployed candidate classification algorithms with a higher performance level than the performance level of each of the other deployed candidate classification algorithms; and deploy the identified candidate classification algorithm on each of the one or more processors configured to perform the flow classification operation.
Example 2 includes the subject matter of Example 1, and wherein a candidate classification algorithm comprises one of the plurality of classification algorithm designs having a unique configuration relative to the other candidate classification algorithms.
Example 3 includes the subject matter of any of Examples 1 and 2, and wherein the processor includes a plurality of processor cores, and wherein to deploy each of the candidate classification algorithms to the processor comprises to deploy each of the candidate classification algorithms to a respective one of the processor cores in parallel.
Example 4 includes the subject matter of any of Examples 1-3, and wherein to deploy each of the candidate classification algorithms to the processor comprises to deploy each of the candidate classification algorithms serially to the processor.
Example 5 includes the subject matter of any of Examples 1-4, and wherein to monitor the performance level comprises to monitor a throughput level.
Example 6 includes the subject matter of any of Examples 1-5, and wherein the virtual switch operation mode controller circuitry is further to compare the performance level of the identified candidate classification algorithm to a performance threshold, and wherein to deploy the identified candidate classification algorithm comprises to deploy the identified candidate classification algorithm subsequent to having determined the performance level of the identified candidate classification algorithm is greater than the performance threshold.
Example 7 includes the subject matter of any of Examples 1-6, and wherein the virtual switch operation mode controller circuitry is further to perform an offline profiling operation, and wherein to perform the offline profiling operation comprises to collect a plurality of input parameters; identify a plurality of candidate classification algorithms from the plurality of classification algorithm designs to perform a flow classification operation; and apply, for each of the plurality of candidate classification algorithms, one or more automated scripts using the plurality of collected input parameters.
Example 8 includes the subject matter of any of Examples 1-7, and wherein to collect the plurality of input parameters comprises to collect at least one of a classification rule, an update pattern, and flow information.
Example 9 includes the subject matter of any of Examples 1-8, and wherein to perform the offline profiling operation further comprises to measure the performance level for each of the plurality of candidate classification algorithms as a result of the applied automated scripts; apply, for each of the plurality of candidate classification algorithms, one or more machine learning algorithms to train a performance model for the plurality of candidate classification algorithms in a plurality of network traffic conditions; and construct a learned data structure as a function of the performance model.
Example 10 includes the subject matter of any of Examples 1-9, and wherein the virtual switch operation mode controller circuitry is further to monitor a present performance level of the identified candidate classification algorithm; compare the present performance level of the identified candidate classification algorithm against a performance threshold; and identify, in response to a determination that the present performance level of the identified candidate classification algorithm is less than the performance threshold, a replacement classification algorithm that is predicted to exceed the performance threshold based on the learned data structure.
Example 11 includes the subject matter of any of Examples 1-10, and wherein learned data structure comprises one of a lookup table or a state machine.
Example 12 includes one or more machine-readable storage media comprising a plurality of instructions stored thereon that, in response to being executed, cause a network appliance to identify a set of candidate classification algorithms from a plurality of classification algorithm designs to perform a flow classification operation; deploy each of the candidate classification algorithms to a processor of one or more processors of the network appliance; monitor a performance level of each of the deployed candidate classification algorithms; identify a candidate classification algorithm of the deployed candidate classification algorithms with a higher performance level than the performance level of each of the other deployed candidate classification algorithms; and deploy the identified candidate classification algorithm on each of the one or more processors configured to perform the flow classification operation.
Example 13 includes the subject matter of Example 12, and wherein a candidate classification algorithm comprises one of the plurality of classification algorithm designs having a unique configuration relative to the other candidate classification algorithms.
Example 14 includes the subject matter of any of Examples 12 and 13, and wherein the processor includes a plurality of processor cores, and wherein to deploy each of the candidate classification algorithms to the processor comprises to deploy each of the candidate classification algorithms to a respective one of the processor cores in parallel.
Example 15 includes the subject matter of any of Examples 12-14, and wherein to deploy each of the candidate classification algorithms to the processor comprises to deploy each of the candidate classification algorithms serially to the processor.
Example 16 includes the subject matter of any of Examples 12-15, and wherein to monitor the performance level comprises to monitor a throughput level.
Example 17 includes the subject matter of any of Examples 12-16, and wherein the plurality of instructions further cause the network appliance to compare the performance level of the identified candidate classification algorithm to a performance threshold, and wherein to deploy the identified candidate classification algorithm comprises to deploy the identified candidate classification algorithm subsequent to having determined the performance level of the identified candidate classification algorithm is greater than the performance threshold.
Example 18 includes the subject matter of any of Examples 12-17, and wherein the plurality of instructions further cause the network appliance to perform an offline profiling operation, and wherein to perform the offline profiling operation comprises to collect a plurality of input parameters; identify a plurality of candidate classification algorithms from the plurality of classification algorithm designs to perform a flow classification operation; and apply, for each of the plurality of candidate classification algorithms, one or more automated scripts using the plurality of collected input parameters.
Example 19 includes the subject matter of any of Examples 12-18, and wherein to collect the plurality of input parameters comprises to collect at least one of a classification rule, an update pattern, and flow information.
Example 20 includes the subject matter of any of Examples 12-19, and wherein to perform the offline profiling operation further comprises to measure the performance level for each of the plurality of candidate classification algorithms as a result of the applied automated scripts; apply, for each of the plurality of candidate classification algorithms, one or more machine learning algorithms to train a performance model for the plurality of candidate classification algorithms in a plurality of network traffic conditions; and construct a learned data structure as a function of the performance model.
Example 21 includes the subject matter of any of Examples 12-20, and wherein the plurality of instructions further cause the network appliance to monitor a present performance level of the identified candidate classification algorithm; compare the present performance level of the identified candidate classification algorithm against a performance threshold; and identify, in response to a determination that the present performance level of the identified candidate classification algorithm is less than the performance threshold, a replacement classification algorithm that is predicted to exceed the performance threshold based on the learned data structure.
Example 22 includes the subject matter of any of Examples 12-21, and wherein learned data structure comprises one of a lookup table or a state machine.
Example 23 includes a network appliance for classifying network flows using adaptive virtual routing, the network appliance comprising means for identifying a set of candidate classification algorithms from a plurality of classification algorithm designs to perform a flow classification operation; means for deploying each of the candidate classification algorithms to a processor of one or more processors of the network appliance; means for monitoring a performance level of each of the deployed candidate classification algorithms; means for identifying a candidate classification algorithm of the deployed candidate classification algorithms with a higher performance level than the performance level of each of the other deployed candidate classification algorithms; and circuitry to deploy the identified candidate classification algorithm on each of the one or more processors configured to perform the flow classification operation.
Example 24 includes the subject matter of Example 23, and further including means for comparing the performance level of the identified candidate classification algorithm to a performance threshold, and wherein to deploy the identified candidate classification algorithm comprises to deploy the identified candidate classification algorithm subsequent to having determined the performance level of the identified candidate classification algorithm is greater than the performance threshold.
Example 25 includes the subject matter of any of Examples 23 and 24, and further including means for performing an offline profiling operation, and wherein the means for performing the offline profiling operation comprises means for collecting a plurality of input parameters; means for identifying a plurality of candidate classification algorithms from the plurality of classification algorithm designs to perform a flow classification operation; and means for applying, for each of the plurality of candidate classification algorithms, one or more automated scripts using the plurality of collected input parameters.
