Explicit congestion notification in a virtual environment

Abstract

Some embodiments provide a method of reducing network congestion in a virtual network. The method, at a first CFE of the virtual network, receives multiple encapsulated data packets of a data stream. The encapsulated data packets having been encapsulated by a second CFE, operating on a server of the virtual network. The second CFE identifies a load percentage of the server, sets explicit congestion notification (ECN) bits on a percentage of the data packets based on the load percentage of the server, and encapsulates each data packet. The first CFE determines whether to forward a new connection to the second CFE based at least on the percentage of data packets from the first CFE with the ECN bits set.

Description

In networks that transmit data using internet protocol (IP) packet transmission, multiple routers between the source and destination points of data packets receive and forward the data packets. Each of these routers has a finite capacity for handling data packets and it is possible for a router to receive more data packets in a given period of time than that router is capable of handling. This is referred to as network congestion. In the IP standard, network congestion is handled in various ways. In some IP systems, congested routers drop excess packets, resulting in the destination of the packets failing to acknowledge receipt of those packets. In response to the dropped packets, the source of the data packets reduces the frequency at which it sends subsequent packets.

A more efficient alternative to responding to dropped packets is to use explicit congestion notification (ECN). In an ECN enabled system, all routers along a path between a source and a destination make use of a dedicated set of 2 ECN bits in the IP standard header. These ECN bits are set to binary values 10 or 01 in each packet to indicate that the endpoints support ECN. When a router along the path of a data stream is congested when a packet arrives, the router changes the ECN value to a binary value of 11 to indicate congestion along the route. This status is reported to the source as an ECN echo (ECE) bit in acknowledgement (ACK) packets sent back by the destination. The ACK packets with ECE bits indicating congestion are treated by the source in a similar manner to dropped packets. That is, the source reduces the frequency at which it sends subsequent packets. One of ordinary skill in the art will understand that there is an ECN system implemented as part of an existing extension of internet protocols (IP). Some embodiments work with the existing ECN system. However other embodiments will work with an updated, modified, or otherwise different ECN system than the existing ECN system.

When a packet is sent through a virtual tunnel (sometimes called an “overlay tunnel” or a “VPN tunnel”) of a VPN, the original IP packet headers are encapsulated (e.g., encrypted) at a cloud forwarding element (CFE) of the VPN along with the payload of the data packets. A new header is then prepended to the encapsulated packet. The new header is used by the physical network underlying the VPN and includes ECN bits which the routers of the underlying physical network may use to identify congestion. However, the routers of the physical network are unable to recognize the encapsulated ECN bits of the original packet. Accordingly the encapsulated ECN bits may be repurposed in a method of reducing network congestion in a virtual network by efficiently assigning new data packet streams to receiving CFEs of the VPN tunnels with more available resources.

BRIEF SUMMARY

Some embodiments provide a method of reducing network congestion in a virtual network. The method, at a first CFE of the virtual network, receives multiple encapsulated data packets of a data stream, the encapsulated data packets having been encapsulated by a second CFE, operating on a server of the virtual network. The second CFE identifies a load percentage of the server, sets explicit congestion notification (ECN) bits on a percentage of non-encapsulated data packets based on the load percentage of the server, and encapsulates each non-encapsulated data packet. The first CFE determines whether to forward a new connection to the second CFE based at least on the percentage of data packets from the first CFE with the ECN bits set. In some embodiments, the load percentage is the larger of (i) a percentage of CPU resources of the server in use and (ii) a percentage of memory resources of the server in use. In other embodiments, the second CFE uses the first ECN bit of each of multiple packets to indicate the memory usage percentage of the server and uses the second ECN bit of each of the packets to indicate the CPU usage percentage, rather than both bits indicating a single load percentage value. The encapsulated data packets, in some embodiments are encapsulated acknowledgement (ACK) packets each sent in response to a set of encapsulated packets (e.g., one ACK packet for every two original encapsulated packets) received by the second CFE from the first CFE.

The method of some embodiments sets the ECN bits on a percentage of the data packets by randomly determining, for each data packet, whether to set an ECN bit of the data packet to a positive status or a negative status with a probability to set the ECN bit to a positive status being based on the load percentage. In some embodiments, the probability to set the ECN bit to a positive status is equal to the load percentage.

The second CFE may be one of a set of CFEs able to process data packets of the new connection. Each CFE of the set of CFEs sets data packets to the first CFE with a percentage of the data packets having an ECN bit set. The method, in some embodiments, determines whether to forward the new connection to the second CFE by comparing the percentage of data packets from the second CFE with the ECN bits set to percentages of data packets from each of the CFEs in the set of CFEs with the ECN bits set. For example, the method may compare the percentage of data packets from the second CFE with the ECN bits set to percentages of data packets from each of the CFEs in the set of CFEs with the ECN bits set by calculating a relative free capacity of the second CFE. The relative free capacity may be determined by calculating a free capacity for each CFE and dividing the free capacity of the second CFE by the sum of the free capacities of the CFEs in the set of CFEs. Determining whether to forward the new connection to the second CFE may be performed by randomly assigning the new connection either to the second CFE or to another CFE of the set of CFEs, with the probability of assigning the new connection to the second CFE being based on the relative free capacity of the second CFE.

The ECN bits of an encapsulated data packet may be a first encapsulated set of ECN bits and the encapsulated data packet may further include a second, non-encapsulated set of ECN bits. A hardware routing system underlies the virtual network. The second, non-encapsulated set of ECN bits, in some embodiments, is identifiable by the hardware routing system, and the first, encapsulated set of ECN bits is not identifiable by the hardware routing system.

The preceding Summary is intended to serve as a brief introduction to some embodiments of the invention. It is not meant to be an introduction or overview of all inventive subject matter disclosed in this document. The Detailed Description that follows and the Drawings that are referred to in the Detailed Description will further describe the embodiments described in the Summary as well as other embodiments. Accordingly, to understand all the embodiments described by this document, a full review of the Summary, the Detailed Description, the Drawings, and the Claims is needed. Moreover, the claimed subject matters are not to be limited by the illustrative details in the Summary, the Detailed Description, and the Drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

The novel features of the invention are set forth in the appended claims. However, for purposes of explanation, several embodiments of the invention are set forth in the following figures.

FIG. 1 presents a virtual network that is defined for a corporation over several public cloud datacenters and of two public cloud providers A and B.

FIG. 2 conceptually illustrates a process of some embodiments for determining a destination managed forwarding node (MFN) to send data packets to.

FIG. 3 illustrates communications using acknowledgement packets among a set of datacenters with MFNs.

FIG. 4 illustrates examples of an encapsulated packet and a non-encapsulated packet received by and sent by a CFE of a VPN.

FIG. 5 illustrates an example of a managed forwarding node of some embodiments of the invention.

FIG. 6 illustrates MFN selection elements of some embodiments.

FIG. 7 conceptually illustrates a computer system with which some embodiments of the invention are implemented.

DETAILED DESCRIPTION

In the following detailed description of the invention, numerous details, examples, and embodiments of the invention are set forth and described. However, it will be clear and apparent to one skilled in the art that the invention is not limited to the embodiments set forth and that the invention may be practiced without some of the specific details and examples discussed.

Some embodiments provide a method of reducing network congestion in a virtual network. The method, at a first CFE of the virtual network, receives multiple encapsulated data packets of a data stream, the encapsulated data packets having been encapsulated by a second CFE, operating on a server of the virtual network. The second CFE identifies a load percentage of the server, sets explicit congestion notification (ECN) bits on a percentage of non-encapsulated data packets based on the load percentage of the server, and encapsulates each non-encapsulated data packet. The first CFE determines whether to forward a new connection to the second CFE based at least on the percentage of data packets from the first CFE with the ECN bits set. In some embodiments, the load percentage is the larger of (i) a percentage of CPU resources of the server in use and (ii) a percentage of memory resources of the server in use. In other embodiments, the second CFE uses the first ECN bit of each of multiple packets to indicate the memory usage percentage of the server and uses the second ECN bit of each of the packets to indicate the CPU usage percentage, rather than both bits indicating a single load percentage value. The encapsulated data packets, in some embodiments are encapsulated acknowledgement (ACK) packets each sent in response to a set of encapsulated packets (e.g., one ACK packet for every two original encapsulated packets) received by the second CFE from the first CFE.

The method of some embodiments sets the ECN bits on a percentage of the data packets by randomly determining, for each data packet, whether to set an ECN bit of the data packet to a positive status or a negative status with a probability to set the ECN bit to a positive status being based on the load percentage. In some embodiments, the probability to set the ECN bit to a positive status is equal to the load percentage.

The second CFE may be one of a set of CFEs able to process data packets of the new connection. Each CFE of the set of CFEs sets data packets to the first CFE with a percentage of the data packets having an ECN bit set. The method, in some embodiments, determines whether to forward the new connection to the second CFE by comparing the percentage of data packets from the second CFE with the ECN bits set to percentages of data packets from each of the CFEs in the set of CFEs with the ECN bits set. For example, the method may compare the percentage of data packets from the second CFE with the ECN bits set to percentages of data packets from each of the CFEs in the set of CFEs with the ECN bits set by calculating a relative free capacity of the second CFE. The relative free capacity may be determined by calculating a free capacity for each CFE and dividing the free capacity of the second CFE by the sum of the free capacities of the CFEs in the set of CFEs. Determining whether to forward the new connection to the second CFE may be performed by randomly assigning the new connection either to the second CFE or to another CFE of the set of CFEs, with the probability of assigning the new connection to the second CFE being based on the relative free capacity of the second CFE.

The ECN bits of an encapsulated data packet may be a first encapsulated set of ECN bits and the encapsulated data packet may further include a second, non-encapsulated set of ECN bits. A hardware routing system underlies the virtual network. The second, non-encapsulated set of ECN bits, in some embodiments, is identifiable by the hardware routing system, and the first, encapsulated set of ECN bits is not identifiable by the hardware routing system.

As used in this document, data messages refer to a collection of bits in a particular format sent across a network. One of ordinary skill in the art will recognize that the term data message may be used herein to refer to various formatted collections of bits that may be sent across a network, such as Ethernet frames, IP packets, TCP segments, UDP datagrams, etc. Also, as used in this document, references to L2, L3, L4, and L7 layers (or layer 2, layer 3, layer 4, layer 7) are references, respectively, to the second data link layer, the third network layer, the fourth transport layer, and the seventh application layer of the OSI (Open System Interconnection) layer model. Network traffic refers to a set of data packets sent through a network. For example, network traffic could be sent from an application operating on a machine (e.g., a virtual machine or physical computer) on a branch of an SD-WAN through a hub node of a hub cluster of the SD-WAN.

FIG. 1 presents a virtual network 100 that is defined for a corporation over several public cloud datacenters 105 and 110 of two public cloud providers A and B. As shown, the virtual network 100 is a secure overlay network that is established by deploying different managed forwarding nodes 150 in different public clouds and connecting the managed forwarding nodes (MFNs) to each other through overlay tunnels 152. In some embodiments, an MFN is a conceptual grouping of several different components in a public cloud datacenter that with other MFNs (with other groups of components) in other public cloud datacenters establish one or more overlay virtual networks for one or more entities.

One of ordinary skill in the art will understand that the overlay tunnels are not physical entities, but instead are conceptual tunnels that are used to represent the actions of a CFE of the VPN encrypting (sometimes called “encapsulating”) data packets at one end of the virtual tunnel so that only another CFE, conceptually represented as the other end of the tunnel, can de-encapsulate/decrypt the packets to restore the original data packets. While the packets may be transferred along many different physical routes through the underlying network(s), the contents are protected, from third party inspection, by the encapsulation.

As further described below, the group of components that form an MFN include in some embodiments (1) one or more gateways for establishing VPN connections with an entity's compute nodes (e.g., offices, private datacenters, remote users, etc.) that are external machine locations outside of the public cloud datacenters, (2) one or more cloud forwarding elements (CFEs) for encapsulating data messages and forwarding encapsulated data messages between each other in order to define an overlay virtual network over the shared public cloud network fabric, (3) one or more service machines for performing middlebox service operations as well as L4-L7 optimizations, and (4) one or more measurement agents for obtaining measurements regarding the network connection quality between the public cloud datacenters in order to identify desired paths through the public cloud datacenters.

In some embodiments, different MFNs can have different arrangements and different numbers of such components, and one MFN can have different numbers of such components for redundancy and scalability reasons. The CFEs in some embodiments are implemented by hardware routers, by software forwarding elements (e.g., software routers) executing on computers, by some other hardware, software, or combination of hardware or software that directs traffic through a physical or virtual network that implements a network cloud. Similarly, in some embodiments the actions described as being performed by CFEs may be performed by two or more hardware and/or software elements working together, such as a hardware router in conjunction with a software element working either on or outside the router to perform the encapsulation of packets.

Also, in some embodiments, each MFN's group of components execute on different computers in the MFN's public cloud datacenter. In some embodiments, several or all of an MFN's components can execute on one computer of a public cloud datacenter. The components of an MFN in some embodiments execute on host computers that also execute other machines of other tenants. These other machines can be other machines of other MFNs of other tenants, or they can be unrelated machines of other tenants (e.g., compute VMs or containers).

The virtual network 100 in some embodiments is deployed by a virtual network provider (VNP) that deploys different virtual networks over the same or different public cloud datacenters for different entities (e.g., different corporate customers/tenants of the virtual network provider). The virtual network provider in some embodiments is the entity that deploys the MFNs and provides the controller cluster for configuring and managing these MFNs.

The virtual network 100 connects the corporate compute endpoints (such as datacenters, branch offices and mobile users) to each other and to external services (e.g., public web services, or SaaS services such as Office365 or Salesforce) that reside in the public cloud or reside in private datacenter accessible through the Internet. This virtual network leverages the different locations of the different public clouds to connect different corporate compute endpoints (e.g., different private networks and/or different mobile users of the corporation) to the public clouds in their vicinity. Corporate compute endpoints are also referred to as corporate compute nodes in the discussion below.

In some embodiments, the virtual network 100 also leverages the high-speed networks that interconnect these public clouds to forward data messages through the public clouds to their destinations or to get close to their destinations while reducing their traversal through the Internet. When the corporate compute endpoints are outside of public cloud datacenters over which the virtual network spans, these endpoints are referred to as external machine locations. This is the case for corporate branch offices, private datacenters and devices of remote users.

In the example illustrated in FIG. 1, the virtual network 100 spans six datacenters 105a-105f of the public cloud provider A and four datacenters 110a-110d of the public cloud provider B. In spanning these public clouds, this virtual network connects several branch offices, corporate datacenters, SaaS providers and mobile users of the corporate tenant that are located in different geographic regions. Specifically, the virtual network 100 connects two branch offices 130a and 130b in two different cities (e.g., San Francisco, California, and Pune, India), a corporate datacenter 134 in another city (e.g., Seattle, Washington), two SaaS provider datacenters 136a and 136b in another two cities (Redmond, Washington, and Paris, France), and mobile users 140 at various locations in the world. As such, this virtual network can be viewed as a virtual corporate WAN.

In some embodiments, the branch offices 130a and 130b have their own private networks (e.g., local area networks) that connect computers at the branch locations and branch private datacenters that are outside of public clouds. Similarly, the corporate datacenter 134 in some embodiments has its own private network and resides outside of any public cloud datacenter. In other embodiments, however, the corporate datacenter 134 or the datacenter of the branch 130a and 130b can be within a public cloud, but the virtual network does not span this public cloud, as the corporate 134 or branch datacenter 130a connects to the edge of the virtual network 100. In some embodiments, a corporate 134 or branch datacenter 130a may connect to the edge of the virtual network 100 through an IP security (IPsec) tunnel.

As mentioned above, the virtual network 100 is established by connecting different deployed managed forwarding nodes 150 in different public clouds through overlay tunnels 152. Each managed forwarding node 150 includes several configurable components. As further described above and further described below, the MFN components include in some embodiments software-based measurement agents, software forwarding elements (e.g., software routers, switches, gateways, etc.), layer 4 proxies (e.g., TCP proxies) and middlebox service machines (e.g., VMs, containers, etc.). One or more of these components in some embodiments use standardized or commonly available solutions, such as Open vSwitch, OpenVPN, strongSwan, etc.

In some embodiments, each MFN (i.e., the group of components the conceptually forms an MFN) can be shared by different tenants of the virtual network provider that deploys and configures the MFNs in the public cloud datacenters. Conjunctively, or alternatively, the virtual network provider in some embodiments can deploy a unique set of MFNs in one or more public cloud datacenters for a particular tenant. For instance, a particular tenant might not wish to share MFN resources with another tenant for security reasons or quality of service reasons. For such a tenant, the virtual network provider can deploy its own set of MFNs across several public cloud datacenters.

In some embodiments, a logically centralized controller cluster 160 (e.g., a set of one or more controller servers) operate inside or outside of one or more of the public clouds 105 and 110, and configure the public-cloud components of the managed forwarding nodes 150 to implement the virtual network 100 over the public clouds 105 and 110. In some embodiments, the controllers in this cluster are at various different locations (e.g., are in different public cloud datacenters) in order to improve redundancy and high availability. The controller cluster in some embodiments scales up or down the number of public cloud components that are used to establish the virtual network 100, or the compute or network resources allocated to these components.

In some embodiments, the controller cluster 160, or another controller cluster of the virtual network provider, establishes a different virtual network for another corporate tenant over the same public clouds 105 and 110, and/or over different public clouds of different public cloud providers. In addition to the controller cluster(s), the virtual network provider in other embodiments deploys forwarding elements and service machines in the public clouds that allow different tenants to deploy different virtual networks over the same or different public clouds. The potential for additional tenants to operate on the same public clouds increases the security risk of unencrypted packets, providing a further incentive for a client to use VPN tunnels to protect data from third parties.

FIG. 2 conceptually illustrates a process 200 of some embodiments for determining a destination MFN (and thus the CFE of the MFN) to send data packets to. Some operations of FIG. 2 will be described in relation to FIG. 3. FIG. 3 illustrates communications using acknowledgement packets among a set of datacenters 110a-110d with MFNs 150 and 305-315. The MFN 150 is determining which of the two candidate MFNs 305 and 310 should be the next hop on the VPN for sending a new stat stream to one of the destination machines 300. In order to gather data used to make this determination, the MFN 150 receives acknowledgment packets (ACK packets) 320 and 325 in acknowledgment of earlier data packets (not shown) of earlier data streams. Although not used in determining the next hop from MFN 150 toward destination machines 300 in the illustrated case, the MFN 315 similarly sends ACK packets 335 and 340 to MFNs 305 and 310, respectively. In FIG. 2, the process 200 is performed by a first CFE, which is sending an initial stream of IP data packets in communication with a second CFE which is receiving the initial stream of data packets and sending acknowledgement (ACK) data packets back to the first CFE in reply.

The process 200 begins when the first CFE sends (at 205) data packets through a virtual tunnel to the second CFE. As mentioned above, sending data packets through a virtual tunnel encapsulates the original data packets and provides a new header for the encapsulated packets to send the packets to another CFE of the VPN. This encapsulation prevents the data in the original packet from being inspected by third parties. The encapsulation also prevents the original header of the original packet, including the ECN bits (e.g., the 15th and 16th bits in the header in the IPv4 standard) from being read by physical routers along the physical path the encapsulated packets travel to the second CFE. However, in some embodiments, the new header includes an additional set of ECN bits, which allows the underlying physical network (if ECN is enabled for all routers along the physical path between the CFEs) to use the ECN system to handle congestion on the physical network.

The process 200 then continues when the second CFE receives (at 210) the data packets and generates an ACK packet for each of the received data packets. The process 200 then identifies (at 215) a load percentage of the server on which the second CFE is operating. This load percentage may be referred to herein as the load percentage of the CFE, or the MFN (for brevity), or of the server on which the MFN is operating. The load percentage in some embodiments is the larger of the percentage of the memory resources in use and the percentage of the CPU resources in use. For example, if 40% of the memory resources were in use and 50% of the CPU resources were in use, then the load percentage would be 50%. In some embodiments, the load percentage is the percentage of the resources available to the CFE that are being used. The resources available may be the total memory and CPU resources of the server, or they may be an amount of those resources specifically allocated to the CFE.

The process 200 then sets (at 220) an ECN bit to positive in each of a percentage of the ACK packets based on the load percentage. As used herein, setting the ECN bit to “positive” indicates that the bit has been set to a value that indicates (to the first CFE) that more resources are in use than if the bit were set to “negative.” The positive setting may be a binary value of 1 in some embodiments, however in alternate embodiments, the binary value of 0 may be the positive value, with a binary value of 1 for the ECN bits indicating more resources are free.

In some embodiments, the ECN bit set (at 220) is the first ECN bit in the header, in other embodiments the ECN bit is the second ECN bit in the header. Still other embodiments may set both ECN bits in each packet (e.g., for more precise signaling of the load percentage). The percentage of the ECN bits set to positive is identical to or approximately the same as the load percentage in some embodiments. However, in other embodiments, the percentage of ECN bits set to positive may be a different value derived from the load percentage. In some embodiments, the process determines, for each packet, whether to set the ECN bit(s) to a positive value based on random selection, with the probability of a positive setting being based on the load percentage.

The process 220 then encapsulates (at 225) the ACK packets and sends them through a virtual tunnel to the first CFE. In FIG. 3, this is shown by ACK packets 320 being sent from MFN 305 to MFN 150 and ACK packets 325 being sent from MFN 310 to MFN 150. Here, the load percentage of the server that implements MFN 305 is 90% and therefore, in this embodiment, 90% of the ACK packets 320 are sent with the ECN bits set, as indicated by the value p=0.9 of the ACK packet. Similarly, the load percentage of the server that implements MFN 310 is 60% and therefore, in this embodiment, 60% of the ACK packets 325 are sent with the ECN bits set, as indicated by the value p=0.6 of the ACK packet. One of ordinary skill in the art will understand that the ACK packets 320 and 325 are sometimes sent in response to entirely different data streams and that CFEs of each MFN 305 and 310 are acting as the “second CFE” in the process 200 of FIG. 2.

One of ordinary skill in the art will understand that in some embodiments, the load percentage of an MFN and thus the percentage of ACK packets from that MFN with the ECN bits set are a property of the MFN. Therefore at any given time, any ACK packets from a particular server to any other server will have the same (or approximately the same, owing to weighted random selections) percentage of ECN bits set. This is illustrated in FIG. 3, as the load percentage of MFN 315 is 70% and the percentage of ACK packets 335 and 340 with ECN bits set is 70%.

At the first CFE, the process 200 receives (at 230) and de-encapsulates the packets. The process 200 then determines (at 235) based on the percentage of ACK packets with ECN bits set to positive whether to forward a new connection to the second CFE. In other embodiments, the first CFE may base the determination on ACK packets with ECN bits set to negative. That is, the determination in some embodiments is based directly on the relative amount of free resources rather than the relative amount of used resources.

In some embodiments, the first CFE similarly receives ACK packets from multiple CFEs (e.g., the CFEs of MFN 305 or 310 in FIG. 3) that include the second CFE. When the first CFE receives a new data stream (sometimes called a “new connection”), a set of packets with the same 5-tuple or 4-tuple, including e.g., source and destination IPs, source and destination ports, and possibly protocol of the packets, the first CFE may assign any of a set of candidate CFEs as the “next hop” within the VPN for packets in that data stream. Each of the CFEs in the set is a potential “next hop” within the VPN for packets from the first CFE to the destination of the packets. In some embodiments, in order to determine which CFE of the set to use as the next hop for a, the first CFE determines the free resources of the candidate CFEs. In some embodiments, the first CFE identifies the free resources of a particular candidate CFE based on the percentage of encapsulated ENC bits set in the ACK packets received from that candidate CFE. For example, if the load percentage of a particular CFE A is based on the percentage of resources in use, the first CFE may use the following example equation or a similar equation to calculate a relative free capacity (RFC) of the CFE A compared to the other available options.Relative Free Capacity A=(100%−load percent A %)/Sum(100%−load[A . . . Z]%)  (1)

That is, in such embodiments, the first CFE calculates a free resource capacity (e.g., the percent remaining given the load percentage) for each candidate CFE and for each CFE compares the free resource capacity of that CFE to the sum of the free resource capacities of all candidate CFEs (CFEs A to Z in equation (1)). In FIG. 3, the RFC for MFNs 305 and 310 are shown beside the ACK packets 320 and 325, respectively. These RFCs are shown to emphasize that they relate to the load percentages (p=0.9 and p=0.6) identified by the ACK packets 320 and 325. However, one of ordinary skill in the art will understand that unlike the load percentage, which is a property of the server of each MFN (at any given time), the RFC is based on the load percentages (and the consequent percent of ACK packets with ECN bits set) of the servers of multiple different MFNs. Therefore, the RFC value for a particular MFN can change even if the load percentage of that MFN changes.

FIG. 3 illustrates a stage where MFN 150 is determining which MFN node 305 or 310 to send a new data stream to. No RFC values are shown for ACK packets 335 and 340 because the next hop is determined at each MFN, in some embodiments. Therefore, until data packets of the new connection reach either MFN 305 or MFN 310, those MFNs do not have a reason to calculate RFC values for a new connection. Additionally, in some embodiments, data streams with different destinations may have different sets of next hops. Since the RFC calculations are based on multiple candidate MFNs, the RFCs are calculated once the candidate MFNs for the next hop of a given data stream are known. Additionally, in some embodiments, when there is only a single candidate MFN for a next hop for a data stream, that data stream is simply assigned to that MFN without calculating an RFC for that candidate.

Once the first CFE determines the relative free capacity of each candidate CFE, the first CFE in some embodiments uses those values (and other factors, in some embodiments) to determine which candidate CFE to assign a new connection to. In some embodiments, the first CFE assigns the new connection to one of the candidate CFEs at random, with the probability of selecting a particular CFE being equal to, or influenced by, the relative free capacity of that CFE. That is, in some embodiments, CFEs with a large percentage of free resources are more likely to be selected than CFEs with a small percentage of free resources. In other embodiments the CFE to assign a new connection to is selected using the load percentage in another way, for example by assigning the new connection to the CFE with the lowest load percentage.

The IPv4 and IPv6 standards each use two bits for ECN. In some embodiments, each bit is used for identifying the usage of a separate resource. For example, in some embodiments, the first encapsulated ECN bit is used to indicate the memory usage percentage and the second encapsulated ECN bit is used to indicate the CPU usage percentage. In such embodiments, after the ACK packets are de-encapsulated by the first CFE, the percentage of de-encapsulated packets with the first ECN bit set to positive (e.g., a binary value of 1) indicates to the first CFE what the memory usage percentage of the second CFE is, and the percentage of de-encapsulated packets with the second ECN bit set to positive indicates to the first CFE what the CPU usage percentage of the second CFE is. The first CFE then uses those values as factors to determine whether to assign a new connection to the second CFE. CFEs that implement the process 200 are further described in context of managed forwarding nodes with respect to FIG. 5, below.

FIG. 4 illustrates examples of an encapsulated packet 400 and a non-encapsulated packet 405 received by and sent by a CFE of a VPN. The non-encapsulated packet 400 is received by a CFE, for example from an application on a VM served by the CFE. The non-encapsulated packet 400 includes IP headers 410, TCP headers 420, and a payload 430. In some embodiments, this payload may include a number of bytes limited based a maximum transmission unit of the route over which the packet will be sent or the payload may be omitted for some packets. The IP headers are comprised of bits arranged according to an IP standard (e.g., IPv4 or IPv6). The IP headers include two ECN bits 415. The headers 410 and 420 and payload 430 in the non-encapsulated packet 400 would be readable if the non-encapsulated packet were sent out before applying encapsulation. However, as described with respect to FIG. 2, in the methods of the present embodiment, a CFE encapsulates the non-encapsulated packet 400 to generate encapsulated packet 405.

Once encapsulated, the IP headers 410, TCP headers 420, and payload 430 are encrypted as the payload of the encapsulated packet 405. The encapsulated packet 405 is then prepended with a new IP header 440 and TCP header 450. The IP header 440 includes a set of ECN bits 445 that are accessible by the underlying network routers between the CFEs of the VPN. If the encapsulated packet 405 is sent over underlying network routers that support ECN, then the ECN bits 445 may be used conventionally to identify congestion along the physical route between the CFEs. However, ECN bits 415 are part of the encrypted non-encapsulated packet 400. Therefore, underlying network routers could not access the ECN bits 415 and those ECN bits are available for use in the manner described with respect to FIG. 2 in implementing the methods of the present invention.

In some embodiments, an intermediate CFE between a sending CFE and a receiving CFE may be the end point of one VPN tunnel and the start of another VPN tunnel. In some such embodiments, the intermediate CFE de-encapsulates packets coming in on one tunnel (e.g., with one encryption key) from a sending CFE. The intermediate CFE then reads the decrypted ECN bits of the packets to determine the load percentage (or individual resource usage) of the sending CFE. The intermediate CFE then sets the ECN bits according to the load percentage of the intermediate CFE, before re-encapsulating the packet and sending the re-encapsulated packet to the receiving CFE. The receiving CFE then de-encapsulates the packet and determines the load percentage or resource usage of the intermediate CFE. Some embodiments may repeat this process through a series of CFEs, with each CFE retrieving load percentage or resource usage data from the ECN bits 415 before re-setting those bits to values that reflect the present CFE's load percentage or resource usage.

FIG. 5 illustrates an example of a managed forwarding node 150 of some embodiments of the invention. In some embodiments, each managed forwarding node 150 is a machine (e.g., a VM or container) that executes on a host computer in a public cloud datacenter. In other embodiments, each managed forwarding node 150 is implemented by multiple machines (e.g., multiple VMs or containers) that execute on the same host computer in one public cloud datacenter. In still other embodiments, two or more components of one MFN can be implemented by two or more machines executing on two or more host computers in one or more public cloud datacenters. As shown, the managed forwarding node 150 includes a firewall 510, one or more optimization engines 520, edge gateways 525 and 530, and a CFE 535 (e.g., a cloud router). In some embodiments, each of these components 505-535 can be implemented as a cluster of two or more components. Further description of managed forwarding nodes of some embodiments is found in U.S. Pat. No. 11,005,684, which is incorporated herein by reference.

In some embodiments, the branch gateway 525 and remote device gateway 530 establish secure IPsec connections (e.g., IPsec tunnels) respectively with one or more branch offices 130 and remote devices (e.g., mobile devices 140) that connect to the MFN 150. The branch gateways 525 and 530 receive data packets through the IPsec tunnel from the branch office 130 and mobile devices 140, respectively. The gateways 525 and 530 send the data packets through firewall 510 and optimization engine 520 to the CFE 535. The CFE 535 then performs the encapsulation process described above and sends the packets to the CFE of another MFN. In some embodiments, the CFE 535 sends the data packets to an MFN determined by the load balancing MFN selector 540 (e.g., an MFN determined when a new connection is received). The MFN selector 540 receives data used to determine which MFN to select for a new connection from the ECN tracker 550. The MFN selector and ECN tracker 550 are further described with respect to FIG. 6, below. In alternate embodiments, some encapsulation and/or MFN determination functions may be performed by the gateways 525 and 530.

FIG. 6 illustrates MFN selection elements of some embodiments. In FIG. 6, first, the CFE 535 receives and de-encapsulates ACK packets from earlier data streams. One of ordinary skill in the art will understand that these earlier data streams do not have to have the same source or destination as a new data stream. Second, the CFE 535 sends the ECN values of the ACK packets to an ECN value tracker 610. In some embodiments, the CFE 535 sends, and the ECN value tracker 610 stores, the ECN values in association with an identifier of the MFN from which the packets were received. Third, the ECN values (and associated MFN identifiers) are then provided to or retrieved by a load percentage calculator 620. Fourth, the load percentages are stored in a load percentage value storage 630. In some embodiments, these first four operations go on in the background in order to prepare the data for later MFN selection.

Fifth, data packets of a new connection are received at the CFE 535. These packets may be received from a gateway (e.g., gateways 525 or 530 of FIG. 5) or through cloud routing fabric from another MFN (not shown). Sixth, the CFE 535, of FIG. 6, sends an MFN selection request to the MFN selector 540. Seventh, the load percentage values are retrieved from the load percentage value storage 630 by the MFN selector 540. Eighth, the MFN selector 540 sends the MFN selection to the CFE 535 (e.g., after calculating the RFC for each candidate MFN and selecting an MFN based at least partly on those calculations). Ninth, the CFE 535 encapsulates the data packets of the new data stream and sends them to the selected MFN. One of ordinary skill in the art will understand that FIG. 6 presents an example of an embodiment of a system for tracking load percentages and selecting MFNs for a new data stream, but that other systems which perform the same or similar functions, either separately or in combination, are used in other embodiments.

Although the above described embodiments use ACK packets as carriers of data concerning load percentages of servers on which MFNs with CFEs operate, one of ordinary skill in the art will understand that load percentage data can be carried in initial data packets instead of or in addition to being carried in ACK packets in some embodiments. Although the above description may refer to MFN selection, one of ordinary skill in the art will understand that in some embodiments, the selection is of a particular CFE or of a particular data center. Although the above described embodiments provided an example of calculations used to select a particular MFN, in some embodiments, one of ordinary skill in the art will understand that in other embodiments, other calculations may be used. Furthermore, in some embodiments, other considerations in addition to load percentage and/or resource use may influence the selection of an MFN. Although the above described embodiments determine a next hop MFN at each MFN along a path to a destination, in some embodiments, an entire path of hops is selected at a particular MFN (e.g., the first MFN to receive a new data stream). In the above described embodiments, each candidate MFN had previously sent ACK packets to an MFN determining a next hop. However, in some embodiments, one or more candidate MFN may not have previously sent packets to the first MFN. In such embodiments, the MFN selector may use other factors (e.g., a default load percentage, etc.) to determine whether to assign a new connection to the previously unknown candidate MFN.

This specification refers throughout to computational and network environments that include virtual machines (VMs). However, virtual machines are merely one example of data compute nodes (DCNs) or data compute end nodes, also referred to as addressable nodes. DCNs may include non-virtualized physical hosts, virtual machines, containers that run on top of a host operating system without the need for a hypervisor or separate operating system, and hypervisor kernel network interface modules.

VMs, in some embodiments, operate with their own guest operating systems on a host using resources of the host virtualized by virtualization software (e.g., a hypervisor, virtual machine monitor, etc.). The tenant (i.e., the owner of the VM) can choose which applications to operate on top of the guest operating system. Some containers, on the other hand, are constructs that run on top of a host operating system without the need for a hypervisor or separate guest operating system. In some embodiments, the host operating system uses name spaces to isolate the containers from each other and therefore provides operating-system level segregation of the different groups of applications that operate within different containers. This segregation is akin to the VM segregation that is offered in hypervisor-virtualized environments that virtualize system hardware, and thus can be viewed as a form of virtualization that isolates different groups of applications that operate in different containers. Such containers are more lightweight than VMs.

Hypervisor kernel network interface modules, in some embodiments, are non-VM DCNs that include a network stack with a hypervisor kernel network interface and receive/transmit threads. One example of a hypervisor kernel network interface module is the vmknic module that is part of the ESXi™ hypervisor of VMware, Inc.

It should be understood that while the specification refers to VMs, the examples given could be any type of DCNs, including physical hosts, VMs, non-VM containers, and hypervisor kernel network interface modules. In fact, the example networks could include combinations of different types of DCNs in some embodiments.

Many of the above-described features and applications are implemented as software processes that are specified as a set of instructions recorded on a computer-readable storage medium (also referred to as computer-readable medium). When these instructions are executed by one or more processing unit(s) (e.g., one or more processors, cores of processors, or other processing units), they cause the processing unit(s) to perform the actions indicated in the instructions. Examples of computer-readable media include, but are not limited to, CD-ROMs, flash drives, RAM chips, hard drives, EPROMs, etc. The computer-readable media does not include carrier waves and electronic signals passing wirelessly or over wired connections.

In this specification, the term “software” is meant to include firmware residing in read-only memory or applications stored in magnetic storage, which can be read into memory for processing by a processor. Also, in some embodiments, multiple software inventions can be implemented as sub-parts of a larger program while remaining distinct software inventions. In some embodiments, multiple software inventions can also be implemented as separate programs. Finally, any combination of separate programs that together implement a software invention described here is within the scope of the invention. In some embodiments, the software programs, when installed to operate on one or more electronic systems, define one or more specific machine implementations that execute and perform the operations of the software programs.

FIG. 7 conceptually illustrates a computer system 700 with which some embodiments of the invention are implemented. The computer system 700 can be used to implement any of the above-described hosts, controllers, gateway and edge forwarding elements. As such, it can be used to execute any of the above-described processes. This computer system 700 includes various types of non-transitory machine-readable media and interfaces for various other types of machine-readable media. Computer system 700 includes a bus 705, processing unit(s) 710, a system memory 725, a read-only memory 730, a permanent storage device 735, input devices 740, and output devices 745.

The bus 705 collectively represents all system, peripheral, and chipset buses that communicatively connect the numerous internal devices of the computer system 700. For instance, the bus 705 communicatively connects the processing unit(s) 710 with the read-only memory 730, the system memory 725, and the permanent storage device 735.

From these various memory units, the processing unit(s) 710 retrieve instructions to execute and data to process in order to execute the processes of the invention. The processing unit(s) may be a single processor or a multi-core processor in different embodiments. The read-only-memory (ROM) 730 stores static data and instructions that are needed by the processing unit(s) 710 and other modules of the computer system. The permanent storage device 735, on the other hand, is a read-and-write memory device. This device is a non-volatile memory unit that stores instructions and data even when the computer system 700 is off. Some embodiments of the invention use a mass-storage device (such as a magnetic or optical disk and its corresponding disk drive) as the permanent storage device 735.

Other embodiments use a removable storage device (such as a floppy disk, flash drive, etc.) as the permanent storage device 735. Like the permanent storage device 735, the system memory 725 is a read-and-write memory device. However, unlike storage device 735, the system memory 725 is a volatile read-and-write memory, such as random access memory. The system memory 725 stores some of the instructions and data that the processor needs at runtime. In some embodiments, the invention's processes are stored in the system memory 725, the permanent storage device 735, and/or the read-only memory 730. From these various memory units, the processing unit(s) 710 retrieve instructions to execute and data to process in order to execute the processes of some embodiments.

The bus 705 also connects to the input and output devices 740 and 745. The input devices 740 enable the user to communicate information and select commands to the computer system 700. The input devices 740 include alphanumeric keyboards and pointing devices (also called “cursor control devices”). The output devices 745 display images generated by the computer system 700. The output devices 745 include printers and display devices, such as cathode ray tubes (CRT) or liquid crystal displays (LCD). Some embodiments include devices such as touchscreens that function as both input and output devices 740 and 745.

Finally, as shown in FIG. 7, bus 705 also couples computer system 700 to a network 765 through a network adapter (not shown). In this manner, the computer 700 can be a part of a network of computers (such as a local area network (“LAN”), a wide area network (“WAN”), or an Intranet), or a network of networks (such as the Internet). Any or all components of computer system 700 may be used in conjunction with the invention.

Some embodiments include electronic components, such as microprocessors, storage and memory that store computer program instructions in a machine-readable or computer-readable medium (alternatively referred to as computer-readable storage media, machine-readable media, or machine-readable storage media). Some examples of such computer-readable media include RAM, ROM, read-only compact discs (CD-ROM), recordable compact discs (CD-R), rewritable compact discs (CD-RW), read-only digital versatile discs (e.g., DVD-ROM, dual-layer DVD-ROM), a variety of recordable/rewritable DVDs (e.g., DVD-RAM, DVD-RW, DVD+RW, etc.), flash memory (e.g., SD cards, mini-SD cards, micro-SD cards, etc.), magnetic and/or solid state hard drives, read-only and recordable Blu-Ray® discs, ultra-density optical discs, any other optical or magnetic media, and floppy disks. The computer-readable media may store a computer program that is executable by at least one processing unit and includes sets of instructions for performing various operations. Examples of computer programs or computer code include machine code, such as is produced by a compiler, and files including higher-level code that are executed by a computer, an electronic component, or a microprocessor using an interpreter.

While the above discussion primarily refers to microprocessors or multi-core processors that execute software, some embodiments are performed by one or more integrated circuits, such as application-specific integrated circuits (ASICs) or field-programmable gate arrays (FPGAs). In some embodiments, such integrated circuits execute instructions that are stored on the circuit itself.

As used in this specification, the terms “computer”, “server”, “processor”, and “memory” all refer to electronic or other technological devices. These terms exclude people or groups of people. For the purposes of the specification, the terms “display” or “displaying” mean displaying on an electronic device. As used in this specification, the terms “computer-readable medium,” “computer-readable media,” and “machine-readable medium” are entirely restricted to tangible, physical objects that store information in a form that is readable by a computer. These terms exclude any wireless signals, wired download signals, and any other ephemeral or transitory signals.

While the invention has been described with reference to numerous specific details, one of ordinary skill in the art will recognize that the invention can be embodied in other specific forms without departing from the spirit of the invention. For instance, several of the above-described embodiments deploy CFEs in public cloud datacenters. However, in other embodiments, the CFEs are deployed in a third-party's private cloud datacenters (e.g., datacenters that the third-party uses to deploy cloud CFEs for different entities in order to deploy virtual networks for these entities). Thus, one of ordinary skill in the art would understand that the invention is not to be limited by the foregoing illustrative details, but rather is to be defined by the appended claims.

Claims

1. A method of reducing network congestion in a virtual network, the method comprising, at a first cloud forwarding element (CFE) of the virtual network: receiving a plurality of encapsulated data packets of a data stream, the encapsulated data packets having been encapsulated by a second CFE, operating on a server of the virtual network, wherein the second CFE: identifies a load percentage of the server;sets explicit congestion notification (ECN) bits on a percentage of non-encapsulated data packets based on the load percentage of the server; andencapsulates each non-encapsulated data packet; anddetermining whether to forward a new connection to the second CFE based at least on the percentage of data packets received from the second CFE with the ECN bits set.

2. The method of claim 1, wherein the load percentage comprises a maximum of (i) a percentage of CPU resources of the server in use and (ii) a percentage of memory resources of the server in use.

3. The method of claim 1, wherein setting the ECN bits on a percentage of the data packets based on the load percentage comprises, for each data packet, randomly determining whether to set an ECN bit of the data packet to a positive status or a negative status with a probability to set the ECN bit to a positive status being based on the load percentage.

4. The method of claim 3, wherein the probability to set the ECN bit to a positive status is equal to the load percentage.

5. The method of claim 1, wherein: the second CFE is one of a set of CFEs able to process data packets of the new connection;each CFE of the set of CFEs sends data packets to the first CFE with a percentage of the data packets having an ECN bit set; anddetermining whether to forward the new connection to the second CFE comprises comparing the percentage of data packets from the second CFE with the ECN bits set to percentages of data packets from each of the CFEs in the set of CFEs with the ECN bits set.

6. The method of claim 5, wherein comparing the percentage of data packets from the second CFE with the ECN bits set to percentages of data packets from each of the CFEs in the set of CFEs with the ECN bits set comprises calculating a relative free capacity of the second CFE by calculating a free capacity for each CFE in the set of CFEs and dividing the free capacity of the second CFE by the sum of the free capacities of the CFEs in the set of CFEs.

7. The method of claim 6, wherein determining whether to forward the new connection to the second CFE comprises randomly assigning the new connection to the second CFE or to another CFE of the set of CFEs with the probability of assigning the new connection to the second CFE being based on the relative free capacity of the second CFE.

8. The method of claim 1, wherein the ECN bits of an encapsulated data packet comprise a first encapsulated set of ECN bits and the encapsulated data packet further comprises a second, non-encapsulated set of ECN bits.

9. The method of claim 8, wherein a hardware routing system underlies the virtual network, the second, non-encapsulated set of ECN bits is identifiable by the hardware routing system, and the first, encapsulated set of ECN bits is not identifiable by the hardware routing system.

10. The method of claim 1, wherein the plurality of encapsulated data packets is a plurality of encapsulated acknowledgement (ACK) packets each sent in response to an encapsulated packet received by the second CFE from the first CFE.

11. A method of reducing network congestion in a virtual network, the method comprising, at a first cloud forwarding element (CFE) of the virtual network: receiving a plurality of encapsulated data packets of a data stream, the encapsulated data packets having been encapsulated by a second CFE, operating on a server of the virtual network, wherein the second CFE: identifies a memory usage percentage and a CPU usage percentage of the server;sets first explicit congestion notification (ECN) bits on a percentage of non-encapsulated data packets based on the memory usage percentage and sets second ECN bits of a percentage of the non-encapsulated data packets based on the CPU usage percentage; andencapsulates each non-encapsulated data packet to generate an encapsulated data packet; anddetermining whether to forward a new connection to the second CFE based on at least one of the percentage of data packets from the second CFE with the first ECN bits set and the percentage of data packets from the second CFE with the second ECN bits set.

12. The method of claim 11, wherein determining whether to forward the new connection to the second CFE is based on both the percentage of data packets from the second CFE with the first ECN bits set and the percentage of data packets from the second CFE with the second ECN bits set.

13. The method of claim 11, wherein the memory percentage of the server comprises a percentage of memory resources of the server that are allocated to the second CFE and in use handling data packets and the CPU usage percentage of the server comprises a percentage of CPU resources of the server that are allocated to the second CFE and in use handling data packets.

14. The method of claim 11, wherein the plurality of encapsulated data packets of the data stream is a first plurality of encapsulated packets of a first data stream and the server is a first server of the virtual network, the method further comprising: receiving a second plurality of encapsulated data packets of a second data stream, the encapsulated data packets of the second plurality of encapsulated data packets having been encapsulated by a third CFE, operating on a second server of the virtual network, wherein the third CFE: identifies a memory usage percentage and a CPU usage percentage of the second server;sets first explicit congestion notification (ECN) bits of a percentage of non-encapsulated data packets of the second server based on the memory usage percentage of the second server and sets second explicit congestion notification bits of a percentage of the non-encapsulated data packets of the second server based on the CPU usage percentage of the second server; andencapsulates each non-encapsulated data packet of the second server to generate an encapsulated data packet of the second plurality of encapsulated data packets; andwherein determining whether to forward the new connection to the second CFE comprises selecting between the second CFE and the third CFE based on (i) at least one of the percentage of data packets from the second CFE with the first ECN bits set and the percentage of data packets from the second CFE with the second ECN bits set, and (ii) at least one of the percentage of data packets from the third CFE with the first ECN bits set and the percentage of data packets from the third CFE with the second ECN bits set.

15. A machine readable medium storing a program which when executed by one or more processing units reduces network congestion in a virtual network, the program comprising sets of instructions for, at a first cloud forwarding element (CFE) of the virtual network: receiving a plurality of encapsulated data packets of a data stream, the encapsulated data packets having been encapsulated by a second CFE, operating on a server of the virtual network, wherein the second CFE: identifies a load percentage of the server;sets explicit congestion notification (ECN) bits on a percentage of non-encapsulated data packets based on the load percentage of the server; andencapsulates each non-encapsulated data packet; anddetermining whether to forward a new connection to the second CFE based at least on the percentage of data packets received from the second CFE with the ECN bits set.

16. The machine readable medium of claim 15, wherein the load percentage comprises a maximum of (i) a percentage of CPU resources of the server in use and (ii) a percentage of memory resources of the server in use.

17. The machine readable medium of claim 15, wherein setting the ECN bits on a percentage of the data packets based on the load percentage comprises, for each data packet, randomly determining whether to set an ECN bit of the data packet to a positive status or a negative status with a probability to set the ECN bit to a positive status being based on the load percentage.

18. The machine readable medium of claim 17, wherein the probability to set the ECN bit to a positive status is equal to the load percentage.

19. The machine readable medium of claim 15, wherein: the second CFE is one of a set of CFEs able to process data packets of the new connection;each CFE of the set of CFEs sends data packets to the first CFE with a percentage of the data packets having an ECN bit set; anddetermining whether to forward the new connection to the second CFE comprises comparing the percentage of data packets from the second CFE with the ECN bits set to percentages of data packets from each of the CFEs in the set of CFEs with the ECN bits set.

20. The machine readable medium of claim 19, wherein comparing the percentage of data packets from the second CFE with the ECN bits set to percentages of data packets from each of the CFEs in the set of CFEs with the ECN bits set comprises calculating a relative free capacity of the second CFE by calculating a free capacity for each CFE in the set of CFEs and dividing the free capacity of the second CFE by the sum of the free capacities of the CFEs in the set of CFEs.