JP2022505453A - Load balancing method based on NAT in DPDK environment - Google Patents

Abstract

The present invention provides a NAT-based load balancing method in a DPDK environment. This method is realized by a scheduler including a DPDK module and a NAT module, and includes the following steps. Initialize the DPDK module and NAT module to obtain the send / receive queue port, NAT rule table, and status recording table. The DPDK module receives and records the source / destination IP and address of the data packet, and calculates the hash value corresponding to the source / destination IP and address. The NAT rule table is searched based on the hash value, it is determined whether or not the server IP and port information corresponding to the obtained hash value exist in the NAT rule table, and if so, the source / destination IP is described. And replace the address to achieve datagram transmission, otherwise find the server with the highest residual load from the status recording table, send the data packet to the server with the highest residual load, and the server and client with the highest residual load. The communication information of the side is recorded in the NAT rule table and the NAT rule table is updated. According to the present invention, the applicability of the scheduler to a large-scale traffic is improved, the extensibility is good, and it is advantageous for equalizing the server load.
[Selection diagram] Fig. 1

Description

The present invention belongs to the field of network communication, and specifically relates to a load distribution method based on NAT (Network Address Translation) in a DPDK (Data Plane Development Kit) environment.

Today, with the explosive growth of computer users, network traffic is also increasing, and server load capacity is also a major issue. Today, even the best performing single servers can't withstand today's heavy traffic demands. Therefore, it is very important to use the idea of clustering to achieve load balancing. Traditional NAT-based load balancing is placed in the traditional TCP / IP protocol stack and is undoubtedly time consuming due to frequent system interruptions and continuous scheduling of context. Therefore, in the conventional use of NAT for load balancing, the scheduler itself is a bottleneck for load balancing. Due to scheduler performance issues, usually only about 10 servers can be scheduled. This is applicable in most cases, but with the explosive growth in network traffic, attacking hundreds of GB / S levels of traffic is no longer an issue, especially with the maturity of DDOS technology. .. Against this background, traditional load balancing based on NAT is clearly inapplicable, and there is an urgent need to find a load balancing technology that can withstand large volumes of traffic.

In the case of load balancing based on conventional NAT, after the scheduler intercepts an external request, a polling method is generally used to randomly find the address of the internal server from the address pool, replace the destination address, and send the source. Keep the address as it is. Furthermore, it is sent to the internal server for processing. Internal and external communication is the reverse process, where the scheduler replaces the source address of the data packet sent from the server with the all-stations address of the scheduler, keeps the destination address as it is, and sends it.

The advantage of traditional NAT-based load balancing is that the server can run any operating system that favors TCP / IP. Therefore, by arranging one IP in the scheduler, the server group can use its own IP address, which saves the IP address of the public network and is more flexible.

However, the scheduler needs to perform frequent search and replace operations on data packets. This workload is much less than the detailed processing of data packets by the server, but as the workload increases, the scheduler can no longer withstand such high frequency work. In general, as the number of server nodes increases to 20, the scheduler itself can become a new bottleneck in the system.

Further, in the case of load balancing based on the conventional NAT, a polling method is generally considered. That is, each server has an equal opportunity for the scheduler to take precedence, and to some extent the use of each server can be balanced. However, in this situation, the performance difference of the server itself is ignored. For example, when the load capacity of the server A is 10 times that of the server B, such a polling method reaches 1/10 of the load upper limit of the server A even when the load of the server B reaches the upper limit. This is unfair to Server B and undoubtedly degrades overall system performance.

In response to the above problems in the prior art in which the performance of the entire system is degraded due to unfair scheduling due to the inability to handle large-scale traffic, the present invention provides a NAT-based load balancing method in a DPDK environment. To. In this method, the DPDK module is used as a processing module of the scheduler, and the data packet is scheduled and transferred by combining the status recording table of the NAT module by the high-speed processing function of the DPDK for the data packet. This overcomes the performance limitations of the scheduler itself, raises the load balancing limit, and provides balanced transfers. The specific technical proposal is as follows.

It is a load balancing method based on NAT in the DPDK environment.
The method is realized by a scheduler including a DPDK module and a NAT module.
Step S1 in which the DPDK module and the NAT module are initialized, the send / receive queue of the scheduler and the corresponding port are arranged, and the corresponding NAT rule table and the status recording table are set.
Step S2 of recording the source / destination IP and port of the data packet received by the scheduler by the DPDK module and performing hash calculation based on the source / destination IP and port to obtain the hash value of the data packet.
Based on the hash value, the NAT rule table is traversed, and whether or not the IP and port information of the internal server corresponding to the hash value exists in the NAT rule table is searched. If so, the NAT rule table is searched for. In step S3, which replaces the source / destination IP based on the rule table, sends the data packet to the corresponding server, and if it does not exist, executes step S4.
If the NAT does not have the IP and port information of the internal server corresponding to the hash value, the server with the maximum residual load is searched for in the status recording table, and the data packet is transmitted to the server with the maximum residual load. In step S4, the communication information between the server and the client side having the maximum residual load is recorded in the NAT rule table, and the NAT rule table is updated.
including.

Further, the step S2 further includes the following.
It is determined whether the data packet is transmitted from the client side to the server or from the server to the client side, and if the data packet is transmitted from the client side to the server, steps S3 to S4 are executed, otherwise. If so, follow the steps below
The NAT rule table is traversed based on the hash value, the source / destination IP and port information corresponding to the hash value in the NAT rule table are searched, the source / destination IP and the port are replaced, and the data is described. Send the packet to the corresponding client side.

Further, the step S1 further includes the following.
The server in the status recording table is graded according to a predetermined rule based on the remaining load.

Further, the DPDK module receives the data packet by the rte_eth_tx_burst () function.

Compared with the prior art, the NAT-based load balancing method in the DPDK environment of the present invention has the following beneficial effects.
(1) It has adaptability for large-scale traffic. By using DPDK instead of the traditional protocol stack to speed up the processing of data packets and overcome the limitations of traditional NAT-based load balancing schedulers, the scheduler can schedule more servers. , Achieve load balancing for large-scale traffic. (2) It has a good balance and can improve the performance of the entire server cluster. By using the values obtained by comparing the residual load and performance of the status recording table, it is possible to quickly find the server with the maximum true residual load, and compared to the conventional method of polling each server, the server. By taking into account the difference in performance of, load balancing is surely realized. (3) It has expandability. The software can improve the processing performance of the scheduler and can realize load balancing of large-scale traffic. There are many general purpose API interfaces that provide better support for subsequent development compared to hardware systems, as well as further software development according to actual needs.

It is a flowchart which realizes one-time communication between a client side and a server by the method of this invention. It is a structural schematic diagram of the NAT rule table described in the Example of this invention. It is a schematic diagram of the model structure which completes the said method of the Example of this invention.

In order for those skilled in the art to better understand the invention, embodiments of the invention will be described in a clear and complete manner with reference to the following drawings.

In the embodiment of the present invention, a NAT-based load balancing method in a DPDK environment is provided. As shown in FIG. 3, a DPDK module and a NAT module are provided in the scheduler of the present invention. The method of the present invention is realized by the scheduler. FIG. 1 is a flowchart that realizes one communication between the client side and the server by the method of the present invention. As can be seen from the figure, the method of the present invention includes the following steps.

S1: Initialize the DPDK module and NAT module, arrange the send / receive queue of the scheduler and the corresponding port, and set the corresponding NAT rule table and status recording table.

In the embodiment of the present invention, the DPDK module realizes the transmission / reception processing of data packets, and the NAT module maintains the NAT rule table and the status recording table. The NAT rule table is a hash table. As shown in FIG. 2, each key value corresponds to two values. One value is used to replace the destination IP and port at the time of reception and send to the internal server, and the other value is used to replace the source address with the IP address and port of the external scheduler at the time of transmission and then send to the client side. Used for The NAT rule table is created at initialization and updated when the scheduler receives a data packet. Hash based on the client-side IP and port information each time, place the internal server IP and port information in the first position, and the external IP and port information used by the scheduler in the second position. When the scheduler searches, it first determines whether this data packet is received by the external client side or sent by the internal server, based on the source end address of the data packet, and then on the client side. Determines whether to use the first value of the key instead of the destination IP, port, or the second value instead of the source IP port, by hashing based on the IP and port information of ..

The status recording table of the present invention is used to record the specific status of a server having a residual load. In the present invention, the inquiry cycle of the internal server of one NAT module is set. Specifically, it is set according to the actual situation. In contrast, the present invention is not limited or fixed. Within one query cycle, the NAT module queries the internal servers on a regular basis, looks up the current load on the current internal servers, and considers them comprehensively in combination with the performance of each internal server. The process of comprehensive consideration is as follows. for t ← 0 ton; state record [t] ← server_load [t] * server_kind [t] * r, for server_kind = [S1, S2. .. .. Sn]. The performance information of each server is recorded, and if the numerical value is large, it indicates that the performance of the server is strong, and the current load server_load and server_kind are calculated to obtain the true remaining load status of the server. Here, r is a calibration constant and limits the state_record between 1 and 100. The last grade record obtained is inserted into the status record table state_record. According to such a method, in the conventional method in which NAT uses polling, the difference in server performance is ignored, and the deterioration of the performance of the entire system is prevented. Preferably, in this embodiment, the load status of the internal server is graded to a total of 100 grades of 1-100. It should be noted that this is merely a preferred embodiment of the present invention and does not limit the present invention.

S2: The source / destination IP and port of the data packet received by the scheduler are recorded by the DPDK module, and hash calculation is performed based on the source / destination IP and port to obtain the hash value of the data packet.

In a specific embodiment of the present invention, the DPDK module receives a data packet by the rte_eth_tx_burst () function. The process realization process of the rte_eth_tx_burst () function is as follows.

Place a send / receive queue from rte_eth_dev_configle (), and copy the placement parameters to the equipment data area with the memorycpy (& dev-> data-> dev_conf, sizeof (dev-> data-> dev_conf)) function, (* dev-> dev-> dev-> dev_conf. Equipment information is acquired by dev_ops-> dev_infos_get) (dev, & dev_info). At the same time, in the present invention, the port is initialized to the receive queue by the rte_eth_rx_queue_setup () function, eight queues are arranged, and rxq = rte_zmalloc_socket ("ethdev RX queue", sizeof (struct) After allocating a queue structure, filling the structure, and completing the arrangement, a data packet is received by the rite_eth_tx_burst () function, and the source / destination IP and port information of the corresponding data packet are recorded. For the received data packet, the scheduler determines whether the data packet is transmitted from the client side to the server or is transmitted from the server to the client side. If the data packet is sent from the server to the client side, it will perform a hash calculation based on the IP and port to obtain the corresponding hash value, traverse the NAT rule table based on the hash value, and in the NAT rule table. The source / destination IP and port information corresponding to the hash value are searched, the source / destination IP and port are replaced, and the data packet is transmitted to the corresponding client side via the DPDK module by the rte_eth_tx_burst () function. If not, that is, if the data packet was sent from the client side to the server, the hash value of the data packet was obtained by performing hash calculation based on the source / destination IP and port as well. The following steps S3 to S4 are executed based on the hash value.

S3: Traverses the NAT rule table based on the hash value, searches whether the IP and port information of the internal server corresponding to the hash value exists in the NAT rule table, and if so, in the NAT rule table. Based on this, the source / destination IP is replaced, the data packet is transmitted to the corresponding server, and if it does not exist, step S4 is executed.

Specifically, after the rte_eth_tx_burst () function receives the data packet, it records the information of the data packet source / destination IP address port, and the scheduler identifies this as the received data packet, and the source port, Hash calculation is performed based on the IP information, and whether or not the corresponding internal server IP and port information exists in the NAT rule table is searched based on the hash value. If it exists, it replaces the source / destination IP based on the NAT rule table and sends the data packet to the corresponding server. If not, step S4 is executed.

S4: If the NAT does not have the IP and port information of the internal server corresponding to the hash value, the server with the maximum residual load is searched for in the status recording table, and the data packet is transmitted to the server with the maximum residual load. Then, the communication information between the server and the client side having the maximum remaining load is recorded in the NAT rule table, and the NAT rule table is updated.

Specifically, if the corresponding information is not found in the NAT rule table, it indicates that this is the first client-side request. In this case, the scheduler retrieves the maximum recorded value from the status record table, retrieves the corresponding internal server IP, and the scheduler replaces the destination address of this data packet, internally by the DPDK module's rt_eth_tx_burst () function. It sends to the server, records the communication information between the server and the client side with the maximum remaining load in the NAT rule table, and updates the NAT rule table for the next use.

The process of performing one-time communication between the server and the client side by the method of the present invention has been described above with reference to FIG. 1, but since the principle and mechanism of multiple times of communication are the same as those of one-time communication, details are given. Explanation is omitted.

From the above, compared with the conventional technique, the method of the present invention speeds up the processing of data packets by using DPDK instead of the conventional protocol stack, and limits the load balancing scheduler based on the conventional NAT. By overcoming it, the scheduler can schedule more servers, achieve load balancing on large-scale traffic, apply to larger-scale traffic, and have a wider range of applications. In the present invention, by using the value obtained by comparing the residual load and performance of the status recording table, the server with the maximum true residual load can be found at high speed, and the conventional method of polling each server can be used. In comparison, by taking into account the difference in server performance, load balancing can be reliably achieved, the load between servers can be well balanced, and the performance of the entire server cluster can be improved. INDUSTRIAL APPLICABILITY According to the present invention, the processing performance of the scheduler can be improved by software, and load balancing of large-scale traffic can be realized. The scheduler has good extensibility because there are many general purpose API interfaces that provide better support for subsequent development compared to hardware systems, as well as the ability to further develop software according to actual needs.

The above description is a preferred embodiment of the present invention and does not limit the scope of protection of the present invention. Although the present invention will be described in detail by the above examples, those skilled in the art can modify or equivalently replace the technical means of each of the above embodiments. Any use in equivalent structures and related technical fields obtained based on the contents of this specification and drawings is within the scope of the present invention.

Claims (4)

It is a load balancing method based on NAT in the DPDK environment.
The method is realized by a scheduler including a DPDK module and a NAT module.
Step S1 in which the DPDK module and the NAT module are initialized, the send / receive queue of the scheduler and the corresponding port are arranged, and the corresponding NAT rule table and the status recording table are set.
Step S2 of recording the source / destination IP and port of the data packet received by the scheduler by the DPDK module and performing hash calculation based on the source / destination IP and port to obtain the hash value of the data packet.
Based on the hash value, the NAT rule table is traversed, and whether or not the IP and port information of the internal server corresponding to the hash value exists in the NAT rule table is searched. If so, the NAT rule table is searched for. In step S3, which replaces the source / destination IP based on the rule table, sends the data packet to the corresponding server, and if it does not exist, executes step S4.
If the NAT does not have the IP and port information of the internal server corresponding to the hash value, the server with the maximum residual load is searched for in the status recording table, and the data packet is transmitted to the server with the maximum residual load. In step S4, the communication information between the server and the client side having the maximum residual load is recorded in the NAT rule table, and the NAT rule table is updated.
A method characterized by including. The step S2 further includes the following.
It is determined whether the data packet is transmitted from the client side to the server or from the server to the client side, and if the data packet is transmitted from the client side to the server, steps S3 to S4 are executed, otherwise. If so, follow the steps below
The NAT rule table is traversed based on the hash value, the source / destination IP and port information corresponding to the hash value in the NAT rule table are searched, the source / destination IP and the port are replaced, and the data is described. The method of claim 1, wherein the packet is transmitted to the corresponding client side. The step S1 further includes the following.
The method according to claim 1, wherein the server in the status recording table is graded according to a predetermined rule based on the remaining load. The method according to any one of claims 1 to 3, wherein the DPDK module receives the data packet by the rte_eth_tx_burst () function.

Images