Network-based computational accelerator

• US 10,135,739 B2
• Filed: 05/04/2016
• Issued: 11/20/2018
• Est. Priority Date: 05/07/2015
• Status: Active Grant

• Current Assignee: Mellanox Technologies Limited (NVIDIA Corporation)
• Original Assignee: Mellanox Technologies Limited (NVIDIA Corporation)
• Inventors: Raindel, Shachar, Raikin, Shlomo, Liss, Liran
• Primary Examiner(s): Traore, Fatoumata
• Application Number:
  US15/145,983
• Publication Number:
  US 20160330112A1
• Time in Patent Office: 930 Days
• Field of Search: None
• CPC Class Codes: G06F 21/602 Providing cryptographic facilities or services
• CPC Class Codes: H04L 47/193 at the transport layer, e.g. TCP related
• CPC Class Codes: H04L 47/2441 relying on flow classification, e.g. using integrated services [IntServ]
• CPC Class Codes: H04L 63/04 for providing a confidential data exchange among entities communicating through data packet networks
• CPC Class Codes: H04L 63/0428 wherein the data content is protected, e.g. by encrypting or encapsulating the payload
• CPC Class Codes: H04L 63/0485 Networking architectures for enhanced packet encryption processing, e.g. offloading of IPsec packet processing or efficient security association look-up
• CPC Class Codes: H04L 67/63 Routing a service request depending on the request content or context
• CPC Class Codes: H04L 69/22 Parsing or analysis of headers

Abstract

A data processing device includes a first packet communication interface for communication with at least one host processor via a network interface controller (NIC) and a second packet communication interface for communication with a packet data network. A memory holds a flow state table containing context information with respect to multiple packet flows conveyed between the host processor and the network via the first and second interfaces packet communication interfaces. Acceleration logic, coupled between the first and second packet communication interfaces, performs computational operations on payloads of packets in the multiple packet flows using the context information in the flow state table.

Claims

1. A data processing device, comprising:

a first packet communication interface for communication with at least one host processor via a network interface controller (NIC);

a second packet communication interface for communication with a packet data network;

a memory, configured to hold a flow state table containing context information with respect to multiple packet flows conveyed between the host processor and the network via the first and second packet communication interfaces; and

acceleration logic, which is coupled between the first and second packet communication interfaces and is configured to perform computational operations on payloads of packets in the multiple packet flows using the context information in the flow state table,

wherein the acceleration logic is configured, upon receiving in a given flow, through the second packet communication interface, a packet that is out of order, to mark the given flow as an out-of-order flow in the flow state table and to convey the packet through the first packet communication interface along with a notification that the given flow is out of order, and

wherein the acceleration logic is configured, after conveying the notification that the given flow is out of order, to receive from the at least one host processor an update to the flow state table with respect to the given flow, and following the update, to resume performing the computational operations on the packets in the given flow.

2. The device according to claim 1, wherein the first and second packet communication interfaces comprise Ethernet interfaces.

3. The device according to claim 1, wherein the multiple packet flows comprise first packets conveyed by the NIC to the device through the first packet communication interface for transmission to the packet data network and second packets conveyed from the packet data network to the device through the second packet communication interface for delivery to the NIC, and

wherein the acceleration logic is configured to perform the computational operations on both the first and the second packets before passing the first and the second packets to the second and the first packet communication interfaces, respectively.

4. The device according to claim 1, wherein the computational operations performed by the acceleration logic comprise cryptographic computations, and wherein the context information comprises cryptographic parameters.

5. The device according to claim 4, wherein the cryptographic parameters comprise a respective encryption key and encryption state for each of the packet flows.

6. The device according to claim 1, wherein the multiple packet flows comprise transport-layer flows, and wherein the context information comprises multiple flow table entries that are respectively keyed to the transport-layer flows.

7. The device according to claim 6, wherein the transport-layer flows comprise sequences of the packets that are transmitted in accordance with the Transmission Control Protocol (TCP), and wherein the flow table entries contain TCP header information.

8. The device according to claim 6, wherein the acceleration logic is configured to store the entries in the flow state table in response to table update instructions conveyed by the host processor to the acceleration logic via the first packet communication interface.

9. The device according to claim 1, wherein the acceleration logic is configured to check the data packets against a predefined condition, and to pass the packets through between the packet communication interfaces without performing the computational operations on the payloads of the packets when the packets do not satisfy the predefined condition.

10. The device according to claim 9, wherein the predefined condition is that the flow state table contain an entry corresponding to a flow to which the packet belongs.

11. The device according to claim 9, wherein the predefined condition, when the packet is received by the acceleration logic through the first packet communication interface, is that the packet is conveyed from the NIC with a mark indicating that handling of the packet by the acceleration logic is needed.

12. The device according to claim 9, wherein the predefined condition is that the packet have a sequence number that matches an expected sequence number that is recorded in the flow state table for a flow to which the packet belongs.

13. The device according to claim 1, wherein the first packet communication interface is configured to be coupled to communicate with multiple host processors, including at least first and second host processors, wherein the flow state table contains context information with respect to at least first and second packet flows conveyed via the device to and from the first and second host processors, respectively.

14. A data processing device, comprising:

a first packet communication interface for communication with at least one host processor via a network interface controller (NIC);

a second packet communication interface for communication with a packet data network;

a memory, configured to hold a flow state table containing context information with respect to multiple packet flows conveyed between the host processor and the network via the first and second packet communication interfaces; and

acceleration logic, which is coupled between the first and second packet communication interfaces and is configured to perform computational operations on payloads of packets in the multiple packet flows using the context information in the flow state table,

wherein the acceleration logic is configured to check the data packets against a predefined condition, and to pass a packet through between the packet communication interfaces without performing the computational operations on a payload of the packet when the packet does not satisfy the predefined condition,

wherein the acceleration logic is configured, after passing a given packet received from the second packet communication interface through to the first packet communication interface without performing the computational operations, to receive, via the first packet communication interface, a replay instruction with respect to the given packet, and in response to the replay instruction, to perform the computational operations on a payload of the given packet and to pass the payload, following performance of the computational operations, to the NIC via the first packet communication interface.

15. A method for data processing, comprising:

receiving in a computational accelerator, which is coupled via a first packet communication interface to communicate with at least one host processor via a network interface controller (NIC) and via a second packet communication interface to communication with a packet data network, a flow of data packets through one of the first and second interfaces for conveyance to the other of the first and second packet communication interfaces;

looking up context information with respect to the flow in a flow state table, which contains respective context entries with respect to multiple packet flows conveyed between the host processor and the network;

performing, in the computational accelerator, a predefined computational operation on a payload of a data packet in the flow using the context information;

after performing the predefined computational operation, transmitting the data packet through the other of the first and second packet communication interfaces;

upon receiving from the network in a given flow the packet that is out of order, marking the given flow as an out-of-order flow in the flow state table and conveying the packet to the NIC along with a notification that the given flow is out of order,

wherein performing the predefined computational operation comprises, after conveying the notification that the given flow is out of order, receiving from the at least one host processor an update to the flow state table with respect to the given flow, and following the update, resuming performance of the computational operations on the packets in the given flow.

16. The method according to claim 15, wherein receiving the flow of data packets comprises receiving and transmitting the data packets through Ethernet interfaces of the computational accelerator.

17. The method according to claim 15, wherein receiving the multiple packet flows comprises receiving first packets conveyed by the NIC to the computational accelerator for transmission to the packet data network and second packets conveyed from the packet data network to the computational accelerator for delivery to the NIC, and

wherein performing the predefined computational operation comprises applying computational operations to both the first and the second packets before transmitting the first and the second packets to the network and the NIC, respectively.

18. The method according to claim 15, wherein performing the computational operations comprises performing cryptographic computations, and wherein the context information comprises cryptographic parameters.

19. The method according to claim 18, wherein the cryptographic parameters comprise a respective encryption key and encryption state for each of the packet flows.

20. The method according to claim 15, wherein the multiple packet flows comprise transport-layer flows, and the context information comprises multiple flow table entries that are respectively keyed to the transport-layer flows, and

wherein looking up the context information comprises extracting a key from a header of the data packet, and reading the context information using the key.

21. The method according to claim 20, wherein the transport-layer flows comprise sequences of the packets that are transmitted in accordance with the Transmission Control Protocol (TCP), and wherein the flow table entries contain TCP header information.

22. The method according to claim 20, and comprising storing the entries in the flow state table in response to table update instructions conveyed by the host processor to the accelerator through one of the first and second packet communication interfaces.

23. The method according to claim 15, wherein performing the predefined computational operation comprises checking the data packets against a predefined condition, and passing the data packets through between the packet communication interfaces without performing the computational operations on payloads of the data packet when the data packets do not satisfy the predefined condition.

24. The method according to claim 23, wherein the predefined condition is that the flow state table contain an entry corresponding to a flow to which the packet belongs.

25. The method according to claim 23, wherein the predefined condition, when the packet is received from the NIC, is that the packet is conveyed from the NIC with a mark indicating that handling of the packet by the accelerator is needed.

26. The method according to claim 23, wherein the predefined condition is that the packet have a sequence number that matches an expected sequence number that is recorded in the flow state table for the flow to which the packet belongs.

27. The method according to claim 23, wherein performing the predefined computational operation comprises, after passing a given data packet through between the packet communication interfaces without performing the computational operations, receiving from the NIC a replay instruction with respect to the given packet, and in response to the replay instruction, performing the computational operations on a payload of the given packet and passing the payload, following performance of the computational operations, to the NIC.

28. The method according to claim 15, and comprising coupling the computational accelerator to communicate with multiple host processors, including at least first and second host processors, wherein the flow state table contains context information with respect to at least first and second packet flows conveyed via the accelerator to and from the first and second host processors, respectively.