Accelerated network packet processing

Abstract

Devices and techniques for accelerated packet processing are described herein. The device can match an action to a portion of a network data packet and accelerate the packet-processing pipeline for the network data packet through the machine by processing the action.

Description

TECHNICAL FIELD

Embodiments described generally herein relate to processing of data packets sent or received through a network. Some embodiments relate to hardware acceleration of data packet processing.

BACKGROUND

Hardware switches when combined with specialized hardware functions provide network capabilities including packet switching, security, deep packet inspection, and other capabilities. Recently, there has been a trend in providing virtual switches and virtual functions executing on high-volume computer architectures. The industry has ongoing efforts focused on improving the coordination between the virtual switches for purposes of optimally taking advantage of the throughput benefits, provided by hardware switches along with the flexibility and power of virtual switches.

BRIEF DESCRIPTION OF THE DRAWINGS

In the drawings, which are not necessarily drawn to scale, like numerals may describe similar components in different views. Like numerals having different letter suffixes may represent different instances of similar components. The drawings illustrate generally, by way of example, but not by way of limitation, various embodiments discussed in the present document.

FIG. 1 illustrates components of a single system that deploys multiple switching platforms for implementing accelerated network packet processing in accordance with some embodiments.

FIG. 2 illustrates components of a system that deploys a virtualization environment for implementing accelerated network packet processing in accordance with some embodiments.

FIG. 3 illustrates a control device to accelerate network processing in accordance with some embodiments.

FIG. 4 illustrates a method for accelerating network packet processing in accordance with some embodiments.

FIG. 5 illustrates a system for accelerating network packet processing in accordance with some embodiments.

DETAILED DESCRIPTION

Some network packet processing solutions have focused on hardware the use top-of-rack (ToR) switches and special function hardware to provide network functions including packet switching, security, deep packet inspection, and other functions. However, customers may experience reduced functionality caused by hardware limitations such as limited memory, limited Ternary Content-Addressable Memory (TCAM), reduced total number of supported data flows, etc. Furthermore, hardware switches may be overly rigid with respect to packet parsing, and hardware switches can exhibit a general lack of platform flexibility and configurability.

Therefore, the industry trend has been to provide software-defined networking (SDN) for decoupling network functions from underlying hardware, which can help increase agility and lower costs. Similarly, network function virtualization (NFV) can replace fixed-function hardware with implementations fully deployed in software that runs more cost-effectively on general purpose, standards-based servers, high-volume servers, and the like. However, such software-defined systems may not take advantage of some desirable characteristics of hardware switches.

These software-based solutions include a variety of software-based abstractions to the underlying physical architectures. For example, a virtual switch permits one or more Virtual Machines (VMs) to communicate with one another. A Virtual Network Function (VNF) may include one or more VMs (running different Operating Systems (OSs)) executing on one or more high volume hardware server architectures, hardware switches, hardware storage, and/or cloud infrastructures. VNFs process to provide specialized network processing in place of customized network appliances.

A variety of Application Programming Interfaces (APIs) and software platforms exist in the industry for permitting network automation utilizing virtual switches. A significant benefit of these approaches is the ability to define and customize packet processing rules and corresponding actions at the user-level of the Operating System (OS). One problem with these approaches is the underutilization of the underlying physical hardware switch, because a significant amount of rule matching and action identification for packet processing takes place within the kernel space of the OS and not on the underlying hardware switch, which has substantially better processing throughput.

One approach, associated with better utilization of the underlying switch architecture, is Single Root Input/Output (I/O) Virtualization (SR-IOV). With SR-IOV, an interface is provided that allows a device adapter to separate hardware resources among packet processing functions. However, this is a binary approach where the feature is: on or off. Moreover, SR-IOV activation alters the architecture's configuration and management. Consequently, SR-IOV is a packet processing “offload” solution, rather than a packet acceleration solution.

Embodiments provide a way to accelerate existing virtual switch solutions to better utilize the underlying physical switch hardware without changing or modifying how the existing virtual switch solutions interact with upper layers of device management and configuration.

Packet processing coordinates and manages, in a fine-grained fashion, multiple data plane components to utilize desirable features of both hardware switching and SDN/NFV usages. Control plane components include mechanisms to determine where traffic (e.g., data packets or flows) should be directed, while data plane components include mechanisms that forward traffic to those destinations. Embodiments provide control plane methods, systems, and devices to accelerate packet processing for multiple data plane components. Data plane components can include, by way of non-limiting example, Data Plane Development Kit (DPDK) components, field programmable gate array (FPGA) components, and Red Rock Canyon (RRC)/FM10K switch components available from Intel of Santa Clara, California, among other components. Methods in accordance with various embodiments can coordinate utilization of these and other components in a dynamic and flexible fashion based on user-defined and user-configured actions to reduce or minimize energy consumption or to enhance speed and performance. In embodiments, a control plane can offload a simple fast packet-processing pipeline from software-based switches or virtual switches to switch hardware, while providing for more-complicated processing on the CPU-based software data plane.

FIG. 1 illustrates components of a single system 100 that deploys multiple switching platforms for implementing methods in accordance with some embodiments. Illustration of the embodiments present just those components necessary for appreciating the depicted embodiments, such that other components are foreseeable without departing from the teachings herein.

The system 100 implements techniques having enhanced match-action acceleration in existing software-based packet processing pipelines (series of packet processing stages) for optimized increased bandwidth, lower latency and jitter, and lower Central Processing Unit (CPU) consumption. Match-action processing refers to a mechanism by which the system 100 enables packet processing acceleration. Match-action resources found in the system 100 execute this acceleration. Existing software-based control planes 101 and 102 (such as OpenStack, OpenDaylight, etc.) do not need modification in order to realize the acceleration presented herein. In addition, existing virtualization constructs (providing virtualized environment management), such as Virtio 113 and 133 do not need modified to use the architecture underlying the system 100, which include the optimizations for packet acceleration.

There are no additional interfaces or control planes, which a user is, required to learn in order to realize the optimizations presented herein with respect to accelerated packet processing/communication between devices (such as VMs managed by VNFs 110, 120, and 130).

Specifically, a Network Interface Controller (NIC) provides one or more novel physical functions (PF 141) and one or more virtual functions (VFs 142 and 143). These functions 141-143 respond to one or more novel match-action (e.g., P4) tables and cooperate with existing control planes 101 and 102 and their existing API commands for executing enhanced network packet processing between VNFs (110, 120, and 130) by offloading software-based match-action network packet processing to the underlying switch architecture of the NIC 140. The processing occurs below the operating system (OS) kernel stack 115 on the NIC 140 within TCAM, the PFs 141, and/or the VFs 142 and 143.

The functions 141-143 provide: 1) virtual switch (vSwitch) acceleration 118, 2) vSwitch offloading 126, 3) VNF acceleration 124, and 4) vitrio (virtualization (VM)) acceleration 135.

It is to be noted that although the system 100 depicts functions 141-143 supporting the four types of acceleration (118, 124, 126, and 135); this does not have to be the case in every instance. That is, in other embodiments, the NIC can implement one and/or configured for supporting, all, or various combinations of the four types of acceleration (118, 124, 126, and 135).

The functions (141, 142, and 143) and the supported four types of network packet acceleration (118, 124, 126, and 135) through match-action processing is as follows (with reference to the system 100 of the FIG. 1).

It is noted that the VNF 110 is programmed for processing (on the underlying hardware of the system 100) a variety of software, such as and by way of example, a network application 111 (performing one or more specialized network packet processing operations), a Data Plan Development Kit 122 (DPDK 122) API, and a virtualization service (virtio 113). The VNF 120 includes a network application 121, a DPDK API option 122, and a NIC VF Driver 123. The VNF 130 includes a network application 131, a DPDK API option 132, and a virtualization service (virtio 133).

The VNFs 110, 120, and 130 situated above the vSwitch 114. The vSwitch 114 situated above the kernel stack 115. Access by the vSwitch for matching and accessing the PF 141 over link 117. Access for accessing VF 142 by the NIC VF driver 123 of VNF 120 over link 125 (for direct TLV lookups through VF 142). Access for the VF 143 by the virtio instance 133 of VNF 130 over links 134 and 137 through 136 for virtio TLV driven lookups 134 and match TLV lookups 137 directly through the VF 143.

vSwitch Acceleration

By employing metadata generated in response to a match action table (e.g., P4 file(s)), acceleration of the packet-processing pipeline for the vSwitch 114 occurs.

A P4 file describes capabilities of the match-action pipeline. In an embodiment, the structure of the match-action table is a P4 file (defined by p4.org open source programming language and formats) is: [Parser]-[Ingress Quality of Service (QoS)]-[Filtering]-[Tunnel/Network Address Translation (NAT)]-[Replication]-[Egress QoS]

In an embodiment, the parser is in P4 format as well. This also provides frame formats (packet payload formats) for matching.

Additionally, P4 provides the mechanism for defining the match filed. A non-limiting set of files can appear as follows:

LI	L2	L3	L4	>L5
Source	Destination	Source	Destination L4	Virtual
Port	Media Access	Internet		Network
	Control (DMAC)	Protocol (IP)		Index
		(4/6)		(VNI)
Source	Source Media	Destination	Source L4	Network
Virtual	Access Control	IP (4/6)		Service
Port	(SMAC)			Header
				(NSH) Path
	Ethernet	Protocol	Transmission	NSH
			Control	Service
			Protocol (TCP)
	Outermost Virtual			L4 TCAM
	Local Area
	Network (VLAN)
	Second
	Outermost VLAN
	Outermost
	Multiprotocol
	Label Switching
	(MPLS)
	Second MPLS

The parse tree also recognizes a format for inner header tunnel-encapsulated frames:

Inner Header	Inner L2	Inner L3	Inner L4
VNI	DMAC	Source IP (4/6)	Destination L4
	SMAC	Destination IP (4/6)	Source L4
	Ethernet	Protocol
	Outermost VLAN

P4 also provides a mechanism for specifying the supported actions, as well as which tables support which actions. A non-limiting example set of actions can be as follows:

		Tunnel/NAT
Basic Actions	Modify Actions	Actions	Replication
Count	Set VLAN/VPRI	Virtual Extensible	Mirror
		Local Area Network
		(VXLAN)
Permit/deny/drop	Push VLAN	Decapsulate	Multicast
		VXLAN
Forward to port	Pop VLAN	Encapsulate NSH	Sample
Forward to virtual	Set Differentiated	Decapsulate NSH
port	Service Code Point
	(DSCP)
Spread across	route
ports
	Set_soft_id

vSwitch Acceleration Using Metadata

The host Ethernet controller (e.g., NIC 140) processes received-received frames first on system 100, and then the host Ethernet controller passes the processed frames up into the various software layers of the system 100 for further additional processing. On the receive path, the Ethernet controller can pre-process the frames and associate with the frame additional metadata. The semantics of this pre-processing are match-action based. The following examples illustrate the processing by which the vSwitch 114 (such as Open vSwitch (OvS) 114) pipeline processing accelerates with embodiments presented herein.

TCAM Pre-Classify

The host controller pushes received packet frames into the TCAM in the pipeline. The vSwitch 114 programs a set of matching rules into the TCAM (using and existing API of the vSwitch 114, such as an API associated with OvS). This causes some amount of metadata being set that include the result of the TCAM lookup. The PF 141, upon matching, appends the result (extra metadata and either in an encapsulating VLAN or a soft_id_value) within the received frame. The software pipeline uses that result in the metadata embedded in the received frame and can avoid having to achieve this TCAM lookup within the software pipeline.

In an embodiment and in the case of OvS (114), an enhanced-patch to OvS 114 detects this added metadata in the packet. If the metadata does not exist in the packet header in the packet header, the OvS 114 pursues its normal processing path. However, the OvS 114 skips 117 its typical software-based TCAM lookup processing within its pipeline when the metadata exists in the packet header (an indication that the PF 141 previously appended the TCAM look results within the NIC 140 before the OvS 114 processes the packet).

Tunnel Decapsulation

The host controller 140 receives a frame, the vSwitch 114 programs rules that map outer VXLAN, VXLAN-Generic-Protocol-Extension (GPE)+NSH, Generic Network Virtualization Encapsulation (Geneve), or Network Virtualization using Generic Routing Extension (NVGRE) headers into the metadata. In an embodiment, The PF 141 processes these rules for: 1) matching the rules on outer header (L2, L3, VNI, and Service Header); 2) decapsulates the matched outer header, and 3) and adds some additional metadata for signaling the removed header. The pipeline processing detects the removed outer header and processes the inner header accordingly.

In the case of OvS 114 (vSwitch 114), the PF 141 uses the decapsulation processing in conjunction with TCAM pre-classify processing for providing metadata on the outer header as well as metadata on the rules matched in the TCAM. When used in conjunction, the TCAM pre-classify rules apply to the inner header.

vSwitch Tap

Tapping an interface can be used for monitoring packet processing by copying some or all of the frames (associated with the interface) and sending the frames to a different location. This unduly taxes the software and hardware resources of the system 100 and is intrusive to the vSwitch 114 processing. Therefore, the teachings presented herein are particularly beneficial in such a scenario. The system 100 achieves this copying using a match rule that maps in the match-action table to a mirror action, which the functions (141-143) process within the NIC 140.

In an embodiment, the vSwitch 114 sends the frames and then processes for multicast replication, tunnel encapsulation, and monitoring in the accelerated pipeline occurs after the frames occurs. In this manner, performance of packet acceleration takes place during transmit of the packets.

vSwitch Offload

vSwitch offload 126 is an extension of vSwitch acceleration 118 when the following conditions are true of a given set of traffic (network packets being processed through the system 100):

1) the accelerated pipeline has a direct connection 125 to the VM (associated with the traffic managed within VNF 120); and

2) the accelerated pipeline has the capability to fully process the traffic with all of the same rules (which the vSwitch 114 data plane processes).

Although identified as an “offload,” the rules within the vSwitch 114 never leave the vSwitch 114; here, the processing pushes and copies the rules into the accelerated pipeline, and pushes statistics for these rules back to the vSwitch 114 so that the rules stay in the software tables. A common configuration where vSwitch offload 126 can occur is when using OvS 114 to implement rules. The rules are applied by a separate control plane software layer; for example, OpenStack and OpenDaylight (e.g., Third Party APIs 101 and 102). The control flow is as follows:

1) A tunnel manager enumerates the virtual switches (multiple instances of vSwitch 114 not shown in the FIG. 1) and creates tunnels between them in a full mesh topology (every vSwitch connects to every other vSwitch). It is noted that in some embodiments, when vSwitches need to converse with one another, creation of the tunnels occur (lazy processing fashion).

2) When two VMs/containers want to connect to one another (from the VNFs 110-130), population of the forwarding rules to the VMs/containers respective vSwitches occurs. Certain systems will populate these rules via virtual L2 learning; other systems will provision the rules directly from a centralized controller.

In addition to tunnel rules (as discussed above), a vSwitch 114 can implement Access Control Lists (ACLs), service function classification, service function forwarding, basic ingress QoS, and connection tracking.

VNF Acceleration

In both vSwitch acceleration 118 and vSwitch offload 126, the vSwitch 114 is executing in the host system 100 and has full control of the underlying device associated with the system 100 for acceleration processing. Here, the connecting VM/container does not have any control over any sort of acceleration “from the network.” The VM/container does not have any control over what happens inside the vSwitch 114. The VNF 120 uses the same semantics as the vSwitch 114 does for: discovering which tables are available, how many entries are available in each table, and what match-action rules for application. However, restriction of the VNF 120 occurs in the following manners:

1) The host system 100 trusts the VNF 120 for enabling the acceleration (otherwise, the host system 100 cannot honor requests from the VNF 120).

2) Application or the VNF's rules occur on traffic sent to or received from that VNF 120. Therefore, the focus in VNF acceleration 124 is on grooming traffic before sending that traffic to the VNF 120. Thus, the VNF 120 cannot add rules to affect traffic going to/from other VFs (such as VF 143), other PFs 141, or the physical uplink ports.

3) The VNF 120 does not have visibility beyond its own interface. Therefore, the VNF 120 cannot add rules affecting traffic going to/from other VNFs (110 and/or 130), PFs 141, VFs 141 and 143, or physical uplink ports.

4) The VNF 120 may not have the same sets of actions available to it. That is, the VNF functions are generally limited to functions related to the VNF 120, such as ACLs (drops, counts, and policing), steering packets to queues, prioritizing packets into queues, and marking packets with metadata for pre-processing.

In an embodiment, the VNF 120 has multiple queues associated with that VNF 120. The VNF 120 is programmed (using for example Data Plan Development Kit 122 (DPDK 122) API) in the following manner:

1) Configure a default Receive Site Scaling (RSS) rule to spread traffic across multiple queues. Here, the VNF 120 can associate a CPU core for each queue to scale the packet processing across multiple flows.

2) Configure a set of FlowDirector® (Intel® packet steering product) rules that work as exceptions to the RSS default-spreading rule, which places specific flow types and mega flows into specific flows or given specific priorities. When RSS cannot efficiently spread with RSS, the configuration of the FlowDirector® aids in handling that traffic. The configuration of the FlowDirector® also aids in handling high priority traffic.

3) Configure a set of filtering rules to drop or police traffic. This can protect the VNF 120 from unwanted traffic, or from receiving traffic at too high a rate.

4) In a manner similar to vSwitch acceleration 118, the VNF 120 can also associate a 16-bit software identifier (ID) on flows, based on match-action rules. This can accelerate certain forwarding paths within the VNF 120 (such as TCAM processing).

Virtio Acceleration

Virtio acceleration 135 is a variant of vSwitch offload 126 still requiring that the hardware of the system 100 be capable of fully forwarding and performing packet processing the frame, but there is no direct connection between the VNF 130/VM and the underlying hardware (134 and 137). Instead, the VNF 130 connects (via software) with a virtio driver 136. In this scenario, there is a software loop (this may be OvS with the DPDK netdev, or some other entity that is copying frames to/from the VF 143 and the virtio queues within the VF 130/VM. A presented scenario implementing OvS with the DPDK netdev is as follows.

Discussion of the virtio acceleration provided with reference to the FIG. 2.

Although embodiments of FIG. 2 are within the context of virtio that other embodiments are deployable utilizing any available virtualization management system.

FIG. 2 illustrates components of a system 200 that deploys a virtualization environment (such as virtio) for implementing accelerated inter-device communication in accordance with some embodiments. Illustration of the embodiments present just those components necessary for appreciating the depicted embodiments, such that other components are foreseeable without departing from the teachings herein.

Again, system 200 illustrates a number of VNFs (210, 220, and 230) each having a network application (211, 221, and 231), DPDK option APIs (212, 222, and 232), and an instance of virtio (virtualization management services—213, 223, and 233). Each virtio instance (214, 223, and 233) connected to a corresponding vHost (241, 242, and 243) having the ability for virtio TLV lookups via 214, 224, and 234.

The vHosts (241, 242, and 245) situated in the OS with a vSwitch 240 and service provided by DPDK 244. The kernel stack 250 situated in the OS under the vSwitch 240 and the DPDK 244. The kernel stack 250 having direct access to the PF 261 and the DPDK 244 and vSwitch 240 having direct access to the VFs 262-265. Match TLVs lookup configurations occur over 266-268. Link 251 providing operations offloaded to the PF 261. The PF 261 and the VFs 262-265 part of the NIC 260.

Underneath OvS, instantiation of N (number of) DPDK netdevs (Linux virtual unction netdevs-functions for obtaining state and capabilities) occurs. Each netdev corresponds to a VF (such as VFs 261 and 262) within the hardware. The arrangement of ports underneath OvS then appears as follows:

1) PFs accelerated ports attached as regular non-DPDK to the OvS kernel data path. Using these interfaces, when a flow requires processing by the kernel (requires TCP, IP tables, etc.).

2) Default Rule VF(s): between 0 and 4 VFs, hardware accessible for sending frames (which are incapable of forwarded a packet processed) to the DPDK user space pipeline for processing. These VFs (such as VFs 142 and 143) have a similar function as the PF 141 ports, except that the packet processing data plane is in the DPDK (with higher performance but without kernel stacks).

• • In an embodiment, if the underlying Red Rock Canyon (RRC) switch components available from Intel® of Santa Clara, California, can support 50 G of bandwidth with on VF (142 or 143) with additional assignment of bandwidth allocated by connecting multiple VFs (142 and 143) from different PFs 117 into one virtio queue.

3) Virtio Alias VFs: remaining VFs (142 or 143 or 261-265 of FIG. 2) used as virtio alias ports and are under the control of the DPDK user-space Poll Mode Driver (PMD).

When attachment of a new VNF 130/VM to the OvS occurs, allocation of this VF 130 occurs as a virtio alias VF. When this happens, the OvS with the DPDK netdev enables a set of “Hardware Offload Virtual (HOV)” paths between its virtual Host (vHost) implementation and the virtio alias VF. Installation of the software logic is as follows:

1) If all of the rules pertaining to the packets coming FROM this VNF 130/VM has been put into the hardware, then zero-copy directly from the vHost into the corresponding virtio alias VF.

2) Else if a partial set of rules pertaining to the packets coming FROM this VF 130/VM has been put into the hardware, process this frame through the OvS user-space pipeline, and when finished push the frame into the corresponding virtio alias VF.

3) Else pass the packet through the OvS packet-processing pipeline. The packet can be sent into the network either via the PF netdevs (the frames are directly forwarded out of a port), via the DPDK VF 130 (in which case the transmitting uplink chosen by the hardware that uses the embedded switch beneath it, or pushed into the kernel for processing via Kernel NIC Interface (KNIC).

4) In the other direction, all received packets on the virtio alias VFs are sent directly from the hardware into the software virtio queues (via vHost) on the HOV path (in an embodiment, a third-party application may provide a zero-copy implementation for doing this).

In the hardware, plumbing of the reverse path occurs and the processing looks similar:

1) If application of all of the rules pertaining to the packets coming TO this VF 130/VM occurred, then forwarding the frame to the VF 130 corresponding to this VM. The PMD picks up this frame and pushes the frame to the VM over the HOV path.

2) If application of a partial set of rules pertaining to this packet occurred, or if this packet misses in the hardware tables, forward the frame to the default rule VF for handling. These frames picked up by the PMD and processed in the DPDK user-space pipeline.

3) Optionally, specific flows needing kernel processing trapped to the PF 141 for standard kernel processing.

When programming the hardware, rules that split traffic between virtio alias, default user space, and kernel data paths preserve conversational ordering. That is one rule does not forward a particular 5-tuple (Destination IP, Source IP, L4-Destination, L4-Source, Protocol) via one path and then forward another frame with the same 5-tuple through another path.

The inverse processing used on the software side: the vHost arbitrates between frames coming directly from the virtio alias, DPDK user-space and kernel data paths. As these are already in conversational order, just process a scheduling function. The virtio alias has highest processing priority, followed by the DPDK user-space and then kernel path.

In this scenario, a VNF 130 may request acceleration by sending Tag-Length-Values (TLVs) via virtio. If this VNF 130 has an aliased VF, these TLVs converted into VNF acceleration requests over a VF mailbox. For this reason and in an embodiment, a configuration provides support for multi-queue virtio, since this allows the VNF 130 to enable RSS and FlowDirector® to spread traffic to multiple queues within the VM. For VNFs using multiple cores, this allows the hardware to spread traffic to multiple cores appearing the connection between VNF 130 and the VF 130 is a direct connection.

In a scenario in which the DPDK or kernel data paths completely process the traffic, the processing through system 100 occurs in two manners:

1) In-line: The first data path to process the frame (hardware packets from the network, software for packets from the vHost) can minimally process the frame and send it to the primary data plane (user space, kernel, or hardware). This is a very efficient mechanism for choosing between data paths, as it does not “bounce-back” frames between data paths, but it does necessitate some amount of processing by the first data path, which may be redundant or overhead on top of the primary data path doing the processing.

2) Bounce-Back: In this scenario software “fastpaths” frames into the hardware, which then loopbacks frames back to the DPDK or the kernel if it cannot handle the frame itself. This has a lower amount of software overhead, but utilizes extra Peripheral Component Internet express (PCIe) bandwidth when frames bounced-back from the hardware back into the software.

In an embodiment, the above-mentioned techniques implemented as methods, an apparatus, and a system for computing-device architectures for achieving accelerated inter-device communication or operations (such as inter-VM communication on-chip for a single device architecture). This provides acceleration for one, all, or combinations of: 1) vSwitch acceleration using metadata, 2) offloaded vSwitch for Virtio connected VMs, 3) offloaded vSwitch for NSH service chaining, and/or 4) Multi-Layer Traffic Pre-Classify (VNF acceleration and NSH service chaining.

These and other embodiments of accelerated inter-device communication presented below with reference to the FIGS. 3-5.

FIG. 3 illustrates a control device 300 to accelerate network processing in accordance with some embodiments. The control device 300 includes a memory 301, processor 302 having instructions 303, a switch interface 304, one or more data plane interfaces 305. The control device 300 interacts with switch silicon 310 (through the switch interface 304) and one or more data plane processor(s) 320 (through the data plane interface(s) 305).

In an embodiment, the control device 300 is a NIC.

In an embodiment, the control device 300 is the NIC 140 of the FIG. 1.

In an embodiment, the control device 300 is the NIC 260 of the FIG. 2

In an embodiment, the control device 300 integrated within and/or interfaced to a multi-core processor. In an embodiment, the multicore processor is a server.

In an embodiment, the control device 300 operates within multiple heterogeneous and virtualized processing environments having a variety of VMs, operating systems, etc.

The control device 300 includes processor 302 to perform functionalities as described herein. It will be understood that any or all of the functions performed by processor 302 can be executed with hardware, software, firmware, or any combination thereof, on one or more processing cores, for example Intel® Architecture cores 114 or a core of the control device 300.

In an embodiment, the processor 302 performs the processing described above with respect to the PF 141 and the VFs 142 and 143 of the FIG. 1.

In an embodiment, the processor 302 performs the processing described above with respect to the PF 261 and the VFs 262-265.

In embodiments, the processor 302 can match an action referenced from a table to a portion of data in a network packet. The processor 302.

In an embodiment, the processor 302 is at least partial programmed by an Application Programming Interface (API). In an embodiment, the API provided in formats supported by one or more of: DPDK, OvS, OpenDaylight, and OpenStack.

In an embodiment, the processor 302 is programmed to be responsive to one or more tables or files that identify actions that the processor 302 is responsive to for purposes of performing predefined processing based on a particular identified action.

In an embodiment, a user interface permits a user to access the API for populating the table or file with actions.

In an embodiment, the file or table is the match-action files discussed above with reference to the FIGS. 1-2.

The processor 302 can match an action reference from a file/table to a portion of a packet header for a received network packet. The processor 302 processes the action identified by the action reference in cooperation with the memory 301 to accelerate a packet-processing pipeline for the network packet.

In an embodiment, the memory 301 is Random Access Memory (RAM) on the control device 300.

Processing the action does not have to be a single operation; rather, the action can trigger the processor 302 to perform a series of predefined operations.

In an embodiment, the processor 302 when processing the action performs a Ternary CAM lookup on some of the packet data. This was discussed above in the FIGS. 1-2 and specific with relevance to acceleration using packet pre-classify processing performed at the NIC.

In an embodiment, the processor 302 when processing the actions decapsulates an outer tunnel encapsulated header for the network packet, removes that header, and adds metadata into the header indicating that the header was removed. This processing discussed above with reference to the FIG. 1.

In an embodiment, the processor 302 when processing the actions copies a packet payload (frame) from the network packet and then sends that copied payload to a network location, which is independent of network locations, defined within or identify resources for processing the packet-processing pipeline. This scenario describes network traffic mirror and monitor operations having substantial improvements in processing throughput since resolution, processing, and sending for mirror packets occur on the control device 300 and not at upper layers within the network architecture, such as the OS layer for a specific VM, and the like.

In an embodiment, the processor 302 when processing the action copies the network packet forwarding rules from a first virtual switch to a second virtual switch. This scenario discussed above with reference to vSwitch offload acceleration and the discussion of the FIG. 1.

The processor 302 configured to communicate with the underlying switch silicon 310 through a switch interface 304.

FIG. 4 illustrates a method 400 for accelerating network packet processing in accordance with some embodiments. The method 400 implemented as executable instructions represented by one or more software modules (method 400 herein after referred to as “packet accelerator”) and executed by one or more hardware processor(s) from a non-transitory computer-readable storage medium.

In an embodiment, the executable instructions representing the packet accelerator is stored in a non-transitory computer-readable storage medium that when executed by one or more computing devices performs the method 400 processing.

In an embodiment, the packet accelerator configured for execution as firmware on a NIC. In an embodiment, the packet accelerator is the PF 141 of the FIG. 1. In an embodiment, the packet accelerator is the VF(s) 142 and/or 143 of the FIG. 1. In an embodiment, the packet accelerator is the PF 261 of the FIG. 2. In an embodiment, the packet accelerator is the VF(s) 262-265 of the FIG. 2.

In an embodiment, the packet accelerator configured for execution within one or more virtual switches, one or more physical switches, one or more device interfaces, one or more virtual device interfaces, and/or one or more operating systems.

In an embodiment, the packet accelerator configured for execution within one or more independent and virtualized environments (execution on system 100 and/or system 200 of the FIGS. 1 and 2, respectively).

In an embodiment, the packet accelerator configured for execution as the control device 300 of the FIG. 3.

At 410, the packet accelerator matches a portion of a network data packet to an action in a match-action table. For example, the processing and file structures of match-action files and matching to network packets discussed in the FIG. 1 above.

At 420, the packet accelerator accelerates processing of the network packet through the machine by executing the action as part of a packet-processing pipeline for the network data packet. That is, action processing accelerates packet processing in a number of manners that can include processing for: 1) offloading processing of software-based resources to hardware-based resources; 2) reorganizing processing of the software-based resources; and/or 3) reorganizing how the software-based resources access the hardware-based resources.

According to an embodiment, at 421, the packet accelerator inserts metadata into the network data packet as an indication that a portion of the packet processing has been handled/processed when the action processed. For example, the pre-classify processing and the TCAM processing presented above in the FIGS. 1-2 provides discussion relevant this processing.

In an embodiment, at 422, the packet accelerator copies a packet frame for the network data packet and sends the copied packet frame to a location that is independent of locations associated with the packet-processing pipeline. This is a packet mirroring processing situation discussed above with reference to the FIG. 1.

In an embodiment, at 423, the packet accelerator assigns the network data packet to a queue associated with a specific processing core. This may entail other configuration dependencies and processing presented above with the discussion of the virtio acceleration processing of the FIG. 2.

In an embodiment of 423 and at 424, the packet accelerator filters the network packet responsive to filtering rules.

In an embodiment of 424 and at 425, the packet accelerator sets a resource identifier on the network data packet that identifies a resource for processing against the network data packet when processing the network data packet through the packet-processing pipeline. So, tables, metadata, and/or the packet-processing pipeline control structures providing a mechanism by which different heterogeneous software and/or hardware resources that assist in accelerating the network data packet through the packet-processing pipeline (network/device(s)), and these mechanisms identified as a resource identifier for the network data packet. The packet accelerator configured for identifying the resource identifier and causing the corresponding resource to process for accelerating the pipeline processing of the packet through the network/device(s).

In an embodiment, at 426, the packet accelerator performs a variety of alternative processing depending on whether traffic rules are fully offloaded to hardware or other conditions. Such conditions discussed with reference to the virtio acceleration of the FIG. 2. Thus, the packet accelerator can: 1) zero-copy rules for processing the network data packet into an aliased-virtualized function; 2) process a frame for the network data packet through a user-space of an OS and when finished push the frame into the aliased virtualized function; or 3) pass the network data packet through the user-space pipeline for the OS.

Examples, as described herein, may include, or may operate on, logic or a number of components, modules, or mechanisms. Modules are tangible entities (e.g., hardware) capable of performing specified operations and may be configured or arranged in a certain manner. In an example, circuits may be arranged (e.g., internally or with respect to external entities such as other circuits) in a specified manner as a module. In an example, at least a part of one or more computer systems (e.g., a standalone, client or server computer system) or one or more processors of the control device 300 may be configured by firmware or software (e.g., instructions 303 (FIG. 3), an application portion, or an application) as a module that operates to perform specified operations. In an example, the software may reside on at least one machine-readable medium. In an example, the software, when executed by the underlying hardware of the module (e.g., the control device 300), can include instructions 303 (FIG. 3) to cause the hardware to perform the specified operations.

The term “module” is understood to encompass a tangible entity, be that an entity that is physically constructed, specifically configured (e.g., hardwired), or temporarily (e.g., transitorily) configured (e.g., programmed) to operate in a specified manner or to perform at least part of any operation described herein. Considering examples in which modules are temporarily configured, a module need not be instantiated at any one moment in time. For example, where the modules comprise a general-purpose hardware processor configured using software; the general-purpose hardware processor may be configured as respective different modules at different times. Software may accordingly configure a hardware processor, for example, to constitute a particular module at one instance of time and to constitute a different module at a different instance of time. The terms “application, process, or service,” or variants thereof, is used expansively herein to include routines, program modules, programs, components, and the like, and may be implemented on various system configurations, including single-processor or multiprocessor systems, microprocessor-based electronics, single-core or multi-core systems, combinations thereof, and the like. Thus, the terms “application, process, or service” may be used to refer to an embodiment of software or to hardware arranged to perform at least part of any operation described herein.

While a machine-readable medium may include a single medium, the term “machine-readable medium” may include a single medium or multiple media (e.g., a centralized or distributed database, and/or associated caches and servers).

The term “machine-readable medium” may include any medium that is capable of storing, encoding, or carrying instructions 303 for execution by a machine (e.g., the control device 300 or any other module) and that cause the machine to perform any one or more of the techniques of the present disclosure, or that is capable of storing, encoding or carrying data structures used by or associated with such instructions. In other words, the processor 302 (FIG. 3) can include instructions and can therefore be termed a machine-readable medium in the context of various embodiments. Other non-limiting machine-readable medium examples may include solid-state memories, and optical and magnetic media. Specific examples of machine-readable media may include: non-volatile memory, such as semiconductor memory devices (e.g., Electrically Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM)) and flash memory devices; magnetic disks, such as internal hard disks and removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks.

The instructions 303 may further be transmitted or received over a communications network using a transmission medium utilizing any one of a number of transfer protocols (e.g., frame relay, internet protocol (IP), TCP, user datagram protocol (UDP), hypertext transfer protocol (HTTP), etc.). Example communication networks may include a local area network (LAN), a wide area network (WAN), a packet data network (e.g., the Internet), mobile telephone networks ((e.g., channel access methods including Code Division Multiple Access (CDMA), Time-division multiple access (TDMA), Frequency-division multiple access (FDMA), and Orthogonal Frequency Division Multiple Access (OFDMA) and cellular networks such as Global System for Mobile Communications (GSM), Universal Mobile Telecommunications System (UMTS), CDMA 2000 1×* standards and Long Term Evolution (LTE)), Plain Old Telephone (POTS) networks, and wireless data networks (e.g., Institute of Electrical and Electronics Engineers (IEEE) 802 family of standards including IEEE 802.11 standards (WiFi), IEEE 802.16 standards (WiMax®) and others), peer-to-peer (P2P) networks, or other protocols now known or later developed.

The term “transmission medium” shall be taken to include any intangible medium that is capable of storing, encoding or carrying instructions for execution by hardware processing circuitry, and includes digital or analog communications signals or other intangible medium to facilitate communication of such software.

FIG. 5 illustrates a system 500 for accelerating network packet processing in accordance with some embodiments. The system 500 shown in greatly simplified format with just particular components illustrated that are necessary for comprehending the system 500.

In an embodiment, the system is a multi-core server 510.

In an embodiment, the multi-core server 510 configured for performing vSwitch acceleration in the manners discussed above with reference to the FIG. 1.

In an embodiment, the multi-core server 510 configured for performing vSwitch offload acceleration in the manners discussed above with reference to the FIG. 1.

In an embodiment, the multi-core server 510 configured for performing VNF acceleration in the manners discussed above with reference to the FIG. 1.

In an embodiment, the multi-core server 510 configured for performing virtio acceleration in the manners discussed above with reference to the FIGS. 1 and/or 2.

In an embodiment, the multi-core server 510 configured for performing selective network packet acceleration for: vSwitch acceleration, vSwitch offload acceleration, VNF acceleration, and/or virtio acceleration in the manners discussed above with reference to the FIGS. 1 and/or 2.

The system 500 includes a means for matching a portion of a network packet to an action and a means for processing the action to accelerate a packet-processing pipeline, and, optionally, a means for configuring the means for matching and the means for processing.

The means for matching a portion of a network packet to an action during network-packet processing.

In an embodiment, the means for matching includes a match-action table 521 or file, such as a P4 file as discussed above with reference to the FIG. 1.

In an embodiment, the means for matching includes a memory, such as memory 301 of the FIG. 3.

In an embodiment, the means for matching includes volatile or non-volatile memory on a NIC, such as NIC 140 or 260 of the FIGS. 1 and 2, respectively.

In an embodiment, the means for matching includes volatile or non-volatile memory accessible on one or more devices representing the system 500.

In an embodiment, the means for matching includes a combination of memory and storage accessible on one or more device representing the system 500.

The means for processing configured to process the action (obtained from the means for matching). The action processed to accelerate a packet-processing pipeline (associated with or assigned to) for the network packet.

In an embodiment, the means for processing is one or more of: one or more device driver(s) (interface(s)) 540, one or more virtual interface(s) 540, one or more virtual switch(es) 550, one or more OS kernel processes (560), and/or a NIC with a physical switch 570.

In an embodiment, the means for processing is various combinations of the component devices and modules illustrated in the FIGS. 1, 2, and/or 3.

In an embodiment, the system 500 includes a means for configuring the means for processing to custom-define (or user-define) the action.

In an embodiment, the means for configuring is the API 520. In an embodiment, the API 520 is a specific configured instance of components illustrated in the FIGS. 1 and 2 using an API provided with OvS, DPDK, OpenStack, OpenDaylight, and/or virtio.

In an embodiment, the means for configuring also provides a means for configuring the means for matching. In an embodiment, the API 520 is the means for configuring and permits tables of files to be established for action in the match-action table/file 521.

Additional Notes & Examples

Example 1 includes subject matter (such as a control device, interplane control device, control plane processor, computer device and or any other electrical apparatus, device or processor) including a memory and processing circuitry. The processing circuitry configured to match an action reference from a table to a portion of data in a network data packet. The processing circuitry further configured to process an action identified by the action reference in cooperation with the memory that accelerates a packet-processing pipeline for the network data packet.

In Example 2, the subject matter of Example 1 can optionally include wherein when the processing circuitry processes the action, the processing circuitry is further configured to perform a Ternary CAM lookup on the portion of data and insert results from the lookup into the network packet as a pre-classification of the network data packet.

In Example 3, the subject matter of any of Examples 1-2 can optionally include wherein when the processing circuitry processes the action, the processing circuitry is further configured to decapsulate an outer tunnel encapsulated header of the network data packet, remove the tunnel encapsulated header from the network data packet, and add metadata into a header for the network data packet that indicates the tunnel encapsulated header was removed from the network data packet.

In Example 4, the subject matter of any of Examples 1-3 can optionally include wherein when the processing circuitry processes the action, the processing circuitry is further configured to copy a packet payload from the network data packet and send the copied packet payload to a location independent of locations that process the packet processing pipeline.

In Example 5, the subject matter of Examples 1-4 can optionally include wherein when the processing circuitry processes the action, the processing circuitry is further configured to copy data packet forwarding rules from a first virtual switch to a second virtual switch.

In Example 6, the subject matter of any of Examples 1-5 can optionally include wherein the processing circuitry is a physical function integrated into the control device.

In Example 7, the subject matter of any of Examples 1-5 can optionally include wherein the processing circuitry is a virtualized function programmed into the control device.

In Example 8, the subject matter of any of Examples 1-7 can optionally include a data plane interface configured to forward the network data packet to one of: an OS kernel stack, a virtualized switch, and a device driver.

In Example 9, the subject matter of any of Examples 1-8 can optionally include wherein the control device is a network interface controller (NIC).

In Example 10, the subject matter of any of Examples 1-9 can optionally include wherein the control device is interfaced and integrated into a multi-core hardware server.

Example 11 includes subject matter such as a machine-readable medium including instructions that, when executed on a machine (such as a control device, interplane control device, control plane processor, computing device, NIC card, etc.) cause the machine to match a portion of a network data packet to an action in a match action table and accelerate processing of the network data packet through the machine by executing the action as part of a packet processing pipeline for the network data packet.

In Example 12, the subject matter of Example 11 may optionally include wherein the instructions to accelerate further include insert metadata into the network data packet indicating that a portion of the packet-processing pipeline was processed when the action was processed.

In Example 13, the subject matter of any of Examples 11-12 may optionally include wherein the instructions to accelerate further include instructions to copy a packet frame for the network data packet and send the copied packet frame to a location that is independent of locations associated with the packet-processing pipeline.

In Example 14, the subject matter of any of Examples 11-13 may optionally include wherein the instructions to accelerate further include instructions to assign the network data packet to a queue associated with a specific processing core of the machine.

In Example 15, the subject matter of Example 14 may optionally include wherein the instructions to accelerate further include instructions to filter the network data packet responsive to filtering rules.

In Example 16, the subject matter of Example 15 may optionally include wherein the instructions to accelerate further include instructions to set a resource identifier on the network data packet that identifies a resource for processing against the network data packet when processing the network data packet through the packet-processing pipeline.

Example 17 include a system (e.g., a server, computer, set of cooperating computers, etc.) a means for matching a portion of a network packet to an action and a means for processing the action to accelerate a packet-processing pipeline associated with the network packet based on processing the action.

Example 18 includes the subject matter of Example 17, and optionally wherein the network packet is sent from a first Virtual Machine (VM) to a second VM, and each of the VMs executing on a same multi-core server that executes the means for matching and the means for processing.

Example 19 includes subject matter of any of the Examples 16-17, and optionally further including a means for configuring the means for processing to custom-define the action.

In Example 20, the subject matter of any of the Examples 16-17, and optionally includes wherein the means for processing is one of: a virtualized network switch, a hardware switch, a kernel process, a device driver, and a virtualized interface.

The above detailed description includes references to the accompanying drawings, which form a part of the detailed description. The drawings show, by way of illustration, specific embodiments that may be practiced. These embodiments are also referred to herein as “examples.” Such examples may include elements in addition to those shown or described. However, also contemplated are examples that include the elements shown or described. Moreover, also contemplate are examples using any combination or permutation of those elements shown or described (or one or more aspects thereof), either with respect to a particular example (or one or more aspects thereof), or with respect to other examples (or one or more aspects thereof) shown or described herein.

Publications, patents, and patent documents referred to in this document are incorporated by reference herein in their entirety, as though individually incorporated by reference. In the event of inconsistent usages between this document and those documents so incorporated by reference, the usage in the incorporated reference(s) are supplementary to that of this document; for irreconcilable inconsistencies, the usage in this document controls.

In this document, the terms “a” or “an” are used, as is common in patent documents, to include one or more than one, independent of any other instances or usages of “at least one” or “one or more.” In this document, the term “or” is used to refer to a nonexclusive or, such that “A or B” includes “A but not B,” “B but not A,” and “A and B,” unless otherwise indicated. In the appended claims, the terms “including” and “in which” are used as the plain-English equivalents of the respective terms “comprising” and “wherein.” Also, in the following claims, the terms “including” and “comprising” are open-ended, that is, a system, device, article, or process that includes elements in addition to those listed after such a term in a claim are still deemed to fall within the scope of that claim. Moreover, in the following claims, the terms “first,” “second,” and “third,” etc. are used merely as labels, and are not intended to suggest a numerical order for their objects.

The above description is intended to be illustrative, and not restrictive. For example, the above-described examples (or one or more aspects thereof) may be used in combination with others. Other embodiments may be used, such as by one of ordinary skill in the art upon reviewing the above description. The Abstract is to allow the reader to quickly ascertain the nature of the technical disclosure and is submitted with the understanding that it will not be used to interpret or limit the scope or meaning of the claims. Also, in the above Detailed Description, various features may be grouped together to streamline the disclosure. However, the claims may not set forth features disclosed herein because embodiments may include a subset of said features. Further, embodiments may include fewer features than those disclosed in a particular example. Thus, the following claims are hereby incorporated into the Detailed Description, with a claim standing on its own as a separate embodiment. The scope of the embodiments disclosed herein is to be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled.

Claims

1. An apparatus usable in association with Ternary Content-Addressable Memory (TCAM) and at least one network, the apparatus being for use in network switch forwarding operations associated with the at least one network, the apparatus comprising: packet processing circuitry configurable to implement, when the apparatus is in operation, at least one programmable match-action operation, the at least one programmable match-action operation being to match, based upon match-action table lookup data, at least one action with received packet data;wherein: the packet processing circuitry is programmable, when the apparatus is in the operation, (1) based upon received configuration data and (2) via at least one application programming interface;the received configuration data is to be generated based upon programming language instructions that define the match-action table lookup data and/or header format data of the received packet data;the apparatus is programmable to access at least one portion of the match-action table lookup data stored in the TCAM for use in the at least one programmable match-action operation; andthe at least one action is configurable to comprise packet data classification based upon tunnel header field data classification rules.

2. The apparatus of claim 1, wherein: the match-action table lookup data is configurable to define, at least in part, tunnel header field data classification rules.

3. The apparatus of claim 1, wherein: the at least one action is also configurable to comprise tunnel header data decapsulation associated with the received packet data; andthe tunnel header data decapsulation comprises removal of one or more of: virtual extensible local area network (VXLAN) header data;network virtualization using generic routing extension (NV-GRE) header data;generic network virtualization encapsulation (GENEVE) header data; and/ormultiprotocol label switch (MPLS) header data.

4. The apparatus of claim 1, wherein: the apparatus also comprises programmable parser circuitry to recognize header field data of the received packet data for use in the at least one programmable match-action operation; and/orthe apparatus is configurable to be comprised in a network switch that comprises physical ports.

5. The apparatus of claim 1, wherein: the match-action table lookup data also comprises access control list data.

6. The apparatus of claim 1, wherein: the apparatus is configurable to be comprised in a server system.

7. At least one non-transitory machine-readable medium storing instructions for being executed by an apparatus, the apparatus being usable in association with Ternary Content-Addressable Memory (TCAM) and at least one network, the apparatus being for use in network switch forwarding operations associated with the at least one network, the apparatus comprising packet processing circuitry, the instructions when executed by the apparatus resulting in the apparatus being configured for performance of operations comprising: implementing, by the packet processing circuitry, at least one programmable match-action operation, the at least one programmable match-action operation being to match, based upon match-action table lookup data, at least one action with received packet data;wherein: the packet processing circuitry is programmable, when the apparatus is in the operation, (1) based upon received configuration data and (2) via at least one application programming interface;the received configuration data is to be generated based upon programming language instructions that define the match-action table lookup data and/or header format data of the received packet data;the apparatus is programmable to access at least one portion of the match-action table lookup data stored in the TCAM for use in the at least one programmable match-action operation; andthe at least one action is configurable to comprise packet data classification based upon tunnel header field data classification rules.

8. The at least one non-transitory machine-readable medium of claim 7, wherein: the match-action table lookup data is configurable to define, at least in part, tunnel header field data classification rules.

9. The at least one non-transitory machine-readable medium of claim 7, wherein: the at least one action is also configurable to comprise tunnel header data decapsulation associated with the received packet data; andthe tunnel header data decapsulation comprises removal of one or more of: virtual extensible local area network (VXLAN) header data;network virtualization using generic routing extension (NV-GRE) header data;generic network virtualization encapsulation (GENEVE) header data; and/ormultiprotocol label switch (MPLS) header data.

10. The at least one non-transitory machine-readable medium of claim 7, wherein: the apparatus also comprises programmable parser circuitry to recognize header field data of the received packet data for use in the at least one programmable match-action operation; and/orthe apparatus is configurable to be comprised in a network switch that comprises physical ports.

11. The at least one non-transitory machine-readable medium of claim 7, wherein: the match-action table lookup data also comprises access control list data.

12. A method implemented using an apparatus, the apparatus being usable in association with Ternary Content-Addressable Memory (TCAM) and at least one network, the apparatus being for use in network switch forwarding operations associated with the at least one network, the apparatus comprising packet processing circuitry, the method comprising: implementing, by the packet processing circuitry, at least one programmable match-action operation, the at least one programmable match-action operation being to match, based upon match-action table lookup data, at least one action with received packet data;wherein: the packet processing circuitry is programmable, when the apparatus is in the operation, (1) based upon received configuration data and (2) via at least one application programming interface;the received configuration data is to be generated based upon programming language instructions that define the match-action table lookup data and/or header format data of the received packet data;the apparatus is programmable to access at least one portion of the match-action table lookup data stored in the TCAM for use in the at least one programmable match-action operation; andthe at least one action is configurable to comprise packet data classification based upon tunnel header field data classification rules.

13. The method of claim 12, wherein: the at least one action is also configurable to comprise tunnel header data decapsulation associated with the received packet data; andthe tunnel header data decapsulation comprises removal of one or more of: virtual extensible local area network (VXLAN) header data;network virtualization using generic routing extension (NV-GRE) header data;generic network virtualization encapsulation (GENEVE) header data; and/ormultiprotocol label switch (MPLS) header data.

14. The method of claim 12, wherein: the apparatus also comprises programmable parser circuitry to recognize header field data of the received packet data for use in the at least one programmable match-action operation; and/orthe apparatus is configurable to be comprised in a network switch that comprises physical ports.

15. The method of claim 12, wherein: the match-action table lookup data also comprises access control list data.

16. Network switch circuitry usable in association with Ternary Content-Addressable Memory (TCAM) and at least one network, the network switch circuitry being for use in network switch forwarding operations associated with the at least one network, the network switch circuitry comprising: physical ports for use in carrying out the network switch forwarding operations;packet processing circuitry configurable to implement, when the network switch circuitry is in operation, at least one programmable match-action operation, the at least one programmable match-action operation being to match, based upon match-action table lookup data, at least one action with received packet data;wherein: the packet processing circuitry is programmable, when the network switch circuitry is in the operation, (1) based upon received configuration data and (2) via at least one application programming interface;the received configuration data is to be generated based upon programming language instructions that define the match-action table lookup data and/or header format data of the received packet data;the network switch circuitry is programmable to access at least one portion of the match-action table lookup data stored in the TCAM for use in the at least one programmable match-action operation; andthe at least one action is configurable to comprise packet data classification based upon tunnel header field data classification rules.

17. The network switch circuitry of claim 16, wherein: the at least one action is also configurable to comprise tunnel header data decapsulation associated with the received packet data; andthe tunnel header data decapsulation comprises removal of one or more of: virtual extensible local area network (VXLAN) header data;network virtualization using generic routing extension (NV-GRE) header data;generic network virtualization encapsulation (GENEVE) header data; and/ormultiprotocol label switch (MPLS) header data.

18. The network switch circuitry of claim 16, wherein: the network switch circuitry also comprises programmable parser circuitry to recognize header field data of the received packet data for use in the at least one programmable match-action operation; andthe at least one action is for use in implementation of, at least in part, one or more of the network switch forwarding operations that are to be carried out using the physical ports.

19. The network switch circuitry of claim 16, wherein: the match-action table lookup data also comprises access control list data.

20. The network switch circuitry of claim 16, wherein: the network switch circuitry is configurable to be comprised in a server system.