Application-specific information-processing method, system, and apparatus

Abstract

An information-processing method for application-specific processing of messages. A message is received. Whether the message is in a selected application format is ascertained. If not, the message is routed to a next location. If so, the message is routed to a selected application processor, processed by the processor, and routed to the next location.

Description

BACKGROUND OF THE INVENTION

[0005] 1. Field of the Invention

[0006] The present invention relates generally to an information-processing system and in particular to an information-processing system for use in a network which processes information for use by one or more specific applications.

[0007] 2. Description of Related Art

[0008] Advances in the field of application-specific information-processing systems have solved many issues. But problems remain. The present invention solves some of the remaining problems.

[0009] An example of a traditional implementation of a network protocol stack includes a hardware-link layer—also called a transport layer—that verifies and delivers packets to a software protocol stack. Each packet is selected based on its frame format and stripped of any frame header that may be present. Each packet is then sent to the appropriate protocol stack and subsequently to the appropriate application.

[0010] If the implementation is a single processor implementation, the processor shepherds each packet through each software stack, processing one packet at a time, and queuing other packets for later processing. In a multiple processor implementation, each processor also acts on one packet at a time. Access to shared data structures is carefully serialized, so the total number of packets being processed is limited to the number of processors available.

[0011] One example is a TCP packet's handling on a TCP/IP Ethernet network. A hardware network interface card checks and delivers the incoming TCP packet to a software stack. The TCP packet's Ethernet frame header is removed, and the packet is sent to an IP stack. Then its IP header is verified and removed, and the packet is sent to a TCP stack. The packet's TCP header is verified and removed, and the packet's data is sent to an application. This process is used in many embedded devices.

BRIEF SUMMARY OF THE INVENTION

[0012] A preferred embodiment is an information-processing method for application-specific processing of messages. A message is received. Whether the message is in a selected application format is ascertained. If not, the message is routed to a next location. If so, the message is routed to a selected application processor, processed by the processor, and routed to the next location.

BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The following drawings form part of the present specification and are included to further demonstrate certain aspects of the present invention. The figures are not necessarily drawn to scale. The invention may be better understood by reference to one or more of these drawings in combination with the detailed description of specific embodiments presented herein.

[0014] FIG. 1 shows a high-level diagram of an information-processing system, in accordance with an embodiment of the present invention.

[0015] FIG. 2 shows a high-level diagram of a client to server proxy, in accordance with an embodiment of the present invention.

[0016] FIG. 3 shows a high-level diagram of a server to client proxy, in accordance with an embodiment of the present invention.

[0017] FIG. 4 shows a lower-level diagram of a section of FIG. 2, in accordance with an embodiment of the present invention.

[0018] FIG. 5 illustrates how application messages are parsed across TCP packets, in accordance with an embodiment of the present invention.

[0019] FIG. 6 shows a high-level diagram of a fabric and application service devices, in accordance with an embodiment of the present invention.

[0020] FIG. 7 shows the organization of a packet protected by IPSec in tunnel mode, in accordance with an embodiment of the PRIOR ART.

[0021] FIG. 8 shows a block diagram of a possible VPN device, in accordance with an embodiment of the present invention.

[0022] FIG. 9 shows a possible VPN scenario, in accordance with an embodiment of the present invention.

[0023] FIG. 10 shows a possible VPN device implementation, in accordance with an embodiment of the present invention.

[0024] FIG. 11 shows an outgoing packet handler, in accordance with an embodiment of the present invention.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS

[0025] In a preferred embodiment, an information-processing system encodes and decodes data to and from a network protocol. Processing steps select and prepare the data for use by one or more specific applications.

[0026] In a preferred embodiment, an information-processing system encodes and decodes multiple network protocols. The information-processing system uses data being carried over the network for an application. The information-processing system includes hardware state machines or simple programmable processors as modules. Each such module specializes in a specific task. The modules are interconnected to process data in a generally pipelined fashion. Thereby, module specialization contributes to the information-processing system's capability of processing networking traffic at very high speeds.

[0027] Modules are described as physically distinct entities for convenience, but they are computational units. As such, they may have any physically adequate embodiment. Some examples include being embodied in software running on a general processing device, in hardwired circuits, in a combination of software and hardware, etc. Therefore, any number of modules may be encapsulated in a single IC. Similarly, the functionality of a single module may be enabled using several physical devices. Modules are described by the computations they perform and not by the techniques, mechanisms, and combinations and subcombinations thereof that achieve the specified computations. In a preferred embodiment, the modules are hardware state machines or simple programmable processors.

[0028] FIG. 1 shows a high-level diagram of an information-processing system, according to a preferred embodiment. In this embodiment, network data in the form of packets pass from the client side through a physical-link-level layer to the information-processing system—also called a pipelined engine—where the packets are processed using shared resources, such as shared memory and queues, and then passed to the server side.

[0029] In a preferred embodiment, multiple networking modules work in conjunction. A packet is received at an information-processing system from a physical-link-level layer and then passes through the modules, typically in sequential fashion. The networking modules decode various network protocol frame headers. The application modules process the data contained in the packets.

[0030] Some examples of network protocols are described in Appendix A of this description. It is contemplated that for any message-based network protocol, an embodiment could be implemented to process messages of that protocol without departing from the spirit of the invention.

[0031] Application streams can have state information that is collected and processed over the lifetime of the application stream. In some preferred embodiments, an application service device (ASD) performs a function on one or more application streams. Examples of ASDs include SSL/TLS processors, IPSEC processors, TCP offload engines, content accelerators, firewalls, load balancers, IP-based storage devices, etc.

[0032] State information is collected from each packet to encrypt or decrypt subsequent packets. Typically, increasing bandwidth is difficult, further increasing the importance of achieving higher speeds via other solutions. As illustrated in FIG. 6, Group B, an ASD may perform a service via a number of connected devices. For example, a service that requires multiple tasks could be implemented using a different device to perform each service. In a preferred embodiment, an SSL/TLS device handles SSL/TLS handshakes and encryptions over a TCP/IP network. The SSL/TLS device includes a handshake unit and a data encryption unit. The handshake unit performs the handshake, while the data encryption unit performs encryption.

[0033] In a preferred embodiment, a collection of fabric elements is included in a fabric. The fabric switches each application stream to an appropriate application service device. An application stream is a stream of application data between a set of end points. Examples of application streams include an SSL/TLS connection between a web browser and a web server, e-mail transfers, etc. In this preferred embodiment, the data from each application stream is kept separate from the data of other application streams. Application data can originate at any of the end points and flow to any other end point. Typically, an application stream can be processed by a subset of the ASDs. An example of this is a session in which all data packets of an SSL/TLS session require handling by the particular ASD that handled that session's SSL/TLS handshake.

[0034] FIG. 6 illustrates an embodiment system where packets are gathered from one or more different networks, processed by ASDs and passed out to one or more networks for further processing. Related packets may arrive over different networks. The fabric elements in Group A are responsible for gathering information from related packets and routing them to the appropriate ASDs in Group B. The elements in Group C are responsible for taking the processed output from the ASDs and passing them on to one or more networks.

[0035] Group A preprocessing elements are each responsible for receiving packets from a network. There may be one or more Group A elements. For example, there will typically be more than one Group A element in a large installation where multiple network connections are used to avoid a single point of failure. The Group A preprocessing elements are responsible for routing different type of packets to different ASDs in Group B.

[0036] Group B ASDs are each specialized protocol processors. One or more may be SSL/TLS processors, and one or more may be VPN processors. Group C elements are responsible for load balancing streams of packets from the ASDs to backend servers. There may be one or more Group C elements.

[0037] So Group A and Group C represent a fabric that includes fabric elements. As stated above, Group B represents ASDs. Referring now to a preferred embodiment represented by Group A, Group B, and Group C, each module processes a single packet or data block at a time. The result of the processing is then output from the module and can then be passed to another module for further processing. The modules have access to a shared memory area, which need not be the case in every preferred embodiment. The modules also notify each other with messages about global state change conditions, which need not be the case in every preferred embodiment.

[0038] Referring now to a preferred embodiment depicted in FIG. 1, an information-processing system may be implemented as a pipeline. In the pipeline, modules are related sequentially such that the output of each module—except the last module—is typically the input of its succeeding module. However, the modules are adapted to make decisions and accordingly the pipeline analogy does not apply absolutely. While the general flow of information will typically be from one module to the next in a linear sequence, the relatedness of the modules creates a more complex connectivity design than a simple pipeline.

[0039] In a preferred embodiment, each module processes a single packet at a time, and each packet is processed by a single module at a time. Each module is assigned a time budget, so that the process(es) performed by the module are completed in a known number of clock cycles. The pipeline gates the flow of packets through the modules so that as each module outputs a packet, the module receives a new packet for processing. Each module's time budget is set relatively small in the design stage so that module operations are rendered relatively simple. Therefore, design of each module is driven by its time budget, network speed requirement, packet size, and task complexity.

[0040] For example, in FIG. 2, the Session Manager module establishes a new connection in processing a SYN packet. SYN packets are exchanged in establishing a TCP connection. Simultaneously, the Evaluator module checks a TCP packet to determine if it is in-sequence or out-of-sequence and the Packet Filter module checks the header of a packet to determine to which of the other modules the packet should be forwarded.

[0041] There are many advantages to this design approach, some of which are described. The modules are simpler and relatively independent which makes them easier to specify, design, and test. Also, the information-processing system may be tailored to a specific application or set of applications so that network packets that are part of different application streams can be quickly passed-thru or dropped without affecting the performance of the rest of the device.

[0042] For example, in a preferred embodiment, SSL/TLS information is processed. UDP packets and packets for non-managed ports are filtered out by the packet filter module, so they are not processed by any other modules. This design characteristic makes it easier to quantify the performance characteristic of each module and the entire information-processing system.

[0043] In a preferred embodiment of the invention, represented by FIG. 2, an information-processing system has been designed to process SSL/TLS connections. In this SSL/TLS information-processing system, a certificate table is defined accordingly. Each entry in the certificate table can contain an external IP address, a port number, certificate information, a server IP address, and a server port number. The external IP address and port number constitute a managed port that is used by each client attempting to establish an SSL/TLS connection. The certificate information contains a public certificate and associated private keys necessary for the SSL/TLS handshake. When a connection is made with a client over the managed port for an entry, a matching connection is made to the server port that is given by the server IP address and port number. Any data that arrives from the client over the SSL/TLS connection is decrypted and then forwarded to the server over the matching connection.

[0044] In the context of this description, two tasks are said to be processed concurrently if both tasks are processed at generally the same time, preferably in the same clock cycle. For example, in FIG. 2, a Packet Filter module can be determining whether a packet is for a managed port concurrently with an Evaluator module determining whether another packet is in-sequence or out-of-sequence.

[0045] In a preferred embodiment, an Evaluator module determines whether a packet is in-sequence or out-of-sequence. A sequential stream of packet-based data includes packets, each having an explicit or implicit sequence number specifying its place in the sequential stream. This characteristic allows the packets to be delivered in any order and requires the receiver to place the packets in order. For example, each TCP packet contains an explicit sequence number in its TCP frame header. Out-of-sequence packets can occur in many ways. For example, if a packet is dropped due to congestion, hardware failure, or other cause, the transmitter will not receive an acknowledgement of receipt of the dropped packet. After a period of time without receipt of an acknowledgement, the transmitter will resend the packet.

[0046] In a preferred embodiment, an information-processing system SSL/TLS proxy has two elements: a client to server (CS) proxy and a server to client (SC) proxy. Modules for the CS proxy, represented by FIG. 2, may be implemented as follows.

[0047] A CS Packet Filter module filters IP packets arriving from the client. If an arriving packet is not intended for a managed port, it is routed immediately to the server. SYN, FIN and RST packets are sent to a CS Session Manager module for establishing or breaking a connection, as appropriate. Arriving Ack packets are directly forwarded to the CS Inbound Ack Handler.

[0048] The CS Session Manager module is responsible for accepting the client's TCP connections from the CS Packet Filter and will establish a matching connection with the server in accordance with the Certificate table. The CS Session Manager will terminate a connection upon receipt of a FIN or RST packet or upon time-out.

[0049] The CS Packet Filter module sends each IP packet fragment to a CS Fragmentation Handler module. Once all the fragments of a fragmented packet are received, the fragmented packet is reassembled and sent back to the CS Packet Filter module for processing.

[0050] SSL/TLS handshake messages are created in response to handshake messages from the client. They are sent in reply to the client via the CS Outbound Ack Handler module.

[0051] TCP packets are sent to a CS Evaluator module for determination of whether they are in-sequence or out-of-sequence. In-sequence packets are sent to an CS Application Message Parser module. Out-of-sequence packets are sent to a CS Packet Sequencer module. The CS Packet Sequencer module re-sequences packets and then passes them on in proper sequence to the CS Application Message Parser module for further processing.

[0052] TCP packets can form a stream of data that can contain SSL/TLS messages. As shown in FIG. 5, the boundaries of a TCP packet do not necessarily align with the boundaries of an SSL/TLS message. The CS Application Message Parser module is responsible for combining and splitting TCP packets to form SSL/TLS messages. In FIG. 2, the CS Application Message Parser module sends packets to the CS Application Block module, and the CS Application Block module sends packets to the Results Processor. The communication relationships of these three modules are represented at a lower level in FIG. 4. The CS Application Message Parser module sends SSL/TLS handshake packets to a CS SSL/TLS Connection Manager module, and SSL/TLS data packets to a CS SSL/TLS Crypto Block module.

[0053] The CS SSL/TLS Connection Manager module performs an SSL/TLS handshake using Certificate Information for this managed port from a Certificate Table. The CS SSL/TLS Connection Manager module receives all the handshake messages from the client and creates handshake messages in response. The output from the CS SSL/TLS Connection Manager module will be to establish an SSL/TLS connection or refuse it. This block contains the public key cryptographic elements necessary for SSL/TLS.

[0054] The CS SSL/TLS Crypto Block module first receives SSL/TLS data messages from the client over an established SSL/TLS connection and then encryption processes the SSL/TLS data messages. The encryption processed SSL/TLS data messages are subsequently output to the CS Results Processor module.

[0055] The CS Results Processor module receives the output of the CS SSL/TLS Connection Manager module and SSL/TLS Crypto Block module. The CS Results Processor module sends decrypted SSL/TLS data to the server via the CS Inbound Ack Handler module.

[0056] The CS Inbound Ack Handler module and the CS Outbound Ack Handler module are responsible for tracking the Acks for packets sent out over the network to the server or client, respectively. Ack packets are sent over the network to the client via the CS Outbound Ack Handler module. If an Ack is not received within a set period of time, the packet is resent. If an Ack is not received after repeated attempts, the session is terminated and an error condition is raised.

[0057] In the same preferred embodiment, modules for the server to client (SC) proxy, represented by FIG. 3, may be implemented as follows.

[0058] An SC Packet Filter module receives packets from the server. For each packet, an SC Packet Filter determines whether the packet is a TCP packet and whether the packet is intended for the proxy server. If the packet is not a TCP packet that is intended for the proxy server, the packet is forwarded immediately out to the client. If the packet is an FIN packet or an RST packet, it is sent to an SC Session Manager module.

[0059] In this preferred embodiment, the server never connects to the proxy. Therefore, the SC Session Manager does not handle SYN packets: connection attempts are rejected immediately. Rather, the SC Session Manager module only manages FIN packets, RST packets, and time-out packets that terminate a session. The CS Session Manager is notified about these Session termination events.

[0060] The SC Packet Filter module sends each IP packet fragment to an SC Fragmentation Handler module (not shown in FIG. 3). Once all the fragments of a fragmented packet are received, the fragmented packet is reassembled and sent back to the SC Packet Filter module for processing.

[0061] TCP packets are sent to an SC Evaluator module for determination of whether they are in-sequence or out-of-sequence. In-sequence packets are sent to an SC SSL/TLS Crypto Block module. Out-of-sequence packets are sent to an SC Packet Sequencer module. The SC Packet Sequencer module re-sequences packets and then passes them on in proper sequence to the SC SSL/TLS Crypto Block module for further processing.

[0062] The SC Crypto Block module encrypts the data from each incoming TCP packet and places the encrypted data in an SSL/TLS message. The SC Crypto Block module then forwards the SSL/TLS message to the SC Results Processor module.

[0063] The SC Results Processor module places the SSL/TLS message in a TCP packet frame and sends it to the SC Ack Handler module. The SC Ack Handler module sends the TCP packet to the client over the link transport layer and awaits an Ack for a time period. If an Ack is not received within the time period, the packet is resent. Repeated failures causes the SC Ack Handler module to declare an error event and close the connection.

[0064] A preferred embodiment also provides for superior bandwidth scaling. The local service devices can operate at a lower bandwidth. A simpler design may be used to accomplish complex application processing, with a simpler implementation. ASDs may be connected in parallel as in FIG. 6, Group B to achieve superior aggregate bandwidth. And they may be connected serially to achieve superior aggregate capacity. Where each ASD is capable of a certain quantity of bandwidth (x), and several devices (n) are connected to the fabric as in FIG. 6, Group A the combined system achieves n times x bandwidth.

[0065] This preferred embodiment furthermore provides better path resiliency. An application stream can arrive over multiple networks and channels, where the fabric operationally combines them. Packets of a single application stream will be sent to the same ASD. Where different streams arrive at the fabric having different protocols, the data is converted to a single protocol. For example, where an ASD is only capable of processing IPV4 and that ASD interfaces with a network carrying IPV6 traffic, fabric elements are used to perform the conversion between IPV6 and IPV4.

[0066] One embodiment is a tunnel-mode IPSec VPN. FIG. 7 shows an example of an IP packet protected by IPSec in tunnel mode. The outer IP header is used to actually transmit a packet over an unsecured network. One unsecured network is the Internet. The IPSec header contains the authentication information.

[0067] Security Associations (SA) are the contract between the two ends of an IPSec connection. SAs determine the specific IPSec protocols used to secure packets, keys, and the duration for which the keys are valid. Each connection has a pair of SAs, one for incoming packets and one for outgoing packets. An SA database (SADB) is used to hold the SA.

[0068] FIG. 8 is a block diagram of a possible VPN device. It connects a secured network (accessible only by trusted parties) to an unsecured network. The device (1) provides the protection necessary to authenticate clients and (2) protects the data on the unsecured network by encrypting packet contents.

[0069] The VPN device offers secure channels over the unsecured network. It is possible that unsecured channels are also allowed between clients on the unsecured network and clients on the secured networks. Those channels, if they travel through the VPN device, can be passed through without interference.

[0070] FIG. 9 shows an embodiment where a VPN device is used. Packets transmitted from Client A to Client B—and from Client B to Client A—over the Internet, an unsecured network, are protected by the IPSec authentication and encryption services offered by the VPN device on the Client B side and possibly by software on the Client A side.

[0071] When the VPN device receives a packet from Client A over the unsecured network, it authenticates the sender using the IPSec header, decrypts the packet contents, and passes the decrypted packet over the secured network to the receiving Client B using the inner IP header.

[0072] When the VPN device receives a packet from Client B over the secured network, it encapsulates the entire packet by encrypting it, adding an IPSec header, adding an outer IP header, and transmits it over the unsecured network.

[0073] FIG. 10 is a block diagram of a possible VPN device implementation for incoming packets. A packet filter is responsible for receiving IP packets, recognizing which ones are on secured channels, as well as which have been fragmented. Packets for unsecured channels are sent on a bypass path directly to the transmit block for transmission to the secured network. Fragmented IP packets are sent to a fragmentation handler that FIG. 10 does not show. Packets for secure channels are sent to the IPSec authentication manager. The IPSec authentication manager authenticates the client transmitting the packet, and passes the contents of the packet, including the inner IP header, TCP header, and data blocks, to the decryption block. The authentication manager also communicates with the SADB block to retrieve the incoming SA. The authentication manager may comprise multiple blocks each responsible for a specific task involved with managing SAs.

[0074] The decryption block decrypts the packet content using the decryption algorithm and other information, as specified by the authentication manager. The decrypted packet is then transmitted on the secured network by the transmit block.

[0075] Each block described must process a packet within a specific time period. If the processing within a block is too complicated to meet its time budget, it must be split and then component blocks should meet the time budget for individual packets. Since each block is capable of processing different packets at the same time, total system throughput in terms of packets per time period goals should be met.

[0076] FIG. 11 describes an outgoing packet handler. The packet filter recognizes whether a packet is for a secured channel. If not, the packet is immediately routed to the transmit block. Otherwise, the IPSec session manager communicates with the SADB manager to fetch the necessary outgoing SA for the corresponding secure channel. The session manager gathers and builds the necessary information for authenticating this packet.

[0077] The Encryption block takes the authentication information, builds the outer IP header and IPSec header, and encapsulates the received packet to organize the IPSec packet as shown in FIG. 7. Under certain conditions, the outgoing packet could be big enough to be fragmented into two encapsulated packets. The transmit block then transmits the IPSec encapsulated packet onto the unsecured network.

[0078] As will be recognized by those skilled in the art, the innovative concepts described in the present application can be modified and varied over a tremendous range of applications, and accordingly the scope of patented subject matter is not limited by any of the specific examples given. For example, an ASD, with its described functions, could be physically incorporated into a firewall device, or could be incorporated into a network interface device that is physically located in a server system. Multiple ASDs can be implemented as one physical apparatus. ASDs could perform other data processing tasks instead of or in addition to decryption tasks.

Claims

1. An information-processing method including: receiving a message; ascertaining whether the message is in a selected application format; if the message is not in the selected application format: routing the message to a next location; and if the message is in the selected application format: routing the message to a selected application processor; processing the message by the selected application processor; and routing the message to the next location. A2. The method of claim A1, wherein receiving the message includes receiving a packet. A3. The method of claim A2, wherein receiving the packet includes receiving the packet from a network. A4. The method of claim A3, wherein receiving the packet from a network includes receiving the packet from a packet switched network. A5. The method of claim A4, wherein the network is the Internet. A6. The method of claim A1, wherein ascertaining whether the message is in a selected application format includes ascertaining whether the message is encrypted; and processing the message by the selected application processor includes decrypting the message by the selected application processor. B1. An information-processing system comprising: a fabric configured for communication with a network; a plurality of application service devices; wherein the plurality of application service devices are configured to receive a plurality of unprocessed application-specific messages from the fabric; wherein each unprocessed application-specific message is configured to be processed by a particular application; wherein the fabric is adapted to route each of the plurality of unprocessed application-specific messages to an application service device adapted to process the message with the particular application; wherein the plurality of application service devices are further configured to process the unprocessed application-specific messages in parallel, wherein each unprocessed application-specific message is processed with the particular application for which it is configured, whereby a plurality of processed application-specific messages is produced; and wherein the plurality of application service devices are further configured to send the each processed application-specific message to the fabric. B2. The information-processing system of claim B1, wherein each message comprises a packet. B3. The information-processing system of claim B2, wherein each application service device comprises a hardware state machine. B4. The information-processing system of claim B3, wherein the plurality of application service devices are included in a single integrated circuit. B5. The information-processing system of claim B1, wherein each application service device comprises a simple programmable processor. B6. The information-processing system of claim B1, wherein at least one of the plurality of application service devices comprises a plurality of interoperably configured distinct physical devices. B7. The information-processing system of claim B1, wherein at least one of the plurality of application service devices comprises an SSL/TLS processor. B8. The information-processing system of claim B1, wherein the plurality of unprocessed application-specific messages comprises an unprocessed application stream, and wherein the plurality of processed application-specific messages comprises a processed application stream. B9. The information-processing system of claim B8, wherein the application streams comprise an SSL/TLS connection between a web browser and a web server. B10. The information-processing system of claim B8, wherein the application streams comprise an e-mail transfer. B11. The information-processing system of claim B8, wherein the application streams comprise a virtual private networking communication. B12. The information-processing system of claim B8, wherein the application streams comprise a TCP offload engine communication. C1. An information-processing method, including: receiving a message; after receiving the message: ascertaining whether the message is susceptible to be processed by a particular application; if the message is susceptible to be processed by the particular application: routing the message to an application service device that is adapted to use the particular application to process the message; after routing the message to the application service device: processing the message by the application service device using the particular application; after processing the message: routing the message to a next location; and if the message is not an application-specific message: routing the message to the next location. C2. The information-processing method of claim C1, wherein the particular application comprises a decryption application, and wherein a message susceptible to be processed by the particular application comprises an encrypted message. C3. The information-processing method of claim C2, wherein the message is a packet. C4. An information-processing method, including: a first iteration of the method of claim C1; a second iteration of the method of claim C1; wherein the receiving a message of the second iteration corresponds to the routing of the message to the next location of the first iteration, whereby the message is processed in a pipeline fashion.