Remote memory ring buffers in a cluster of data processing nodes

Patent metadata

Field	Value
Publication number	US-9304896-B2
Application number	US-201313959428-A
Country	US
Kind code	B2
Filing date	Aug 5, 2013
Priority date	Aug 5, 2013
Publication date	Apr 5, 2016
Grant date	Apr 5, 2016

Abstract

A data processing node has an inter-node messaging module including a plurality of sets of registers each defining an instance of a GET/PUT context and a plurality of data processing cores each coupled to the inter-node messaging module. Each one of the data processing cores includes a mapping function for mapping each one of a plurality of user level processes to a different one of the sets of registers and thereby to a respective GET/PUT context instance. Mapping each one of the user level processes to the different one of the sets of registers enables a particular one of the user level processes to utilize the respective GET/PUT context instance thereof for performing a GET/PUT action to a ring buffer of a different data processing node coupled to the data processing node through a fabric without involvement of an operating system of any one of the data processing cores.

First claim

What is claimed is:

1. A data processing node, comprising:
an inter-node messaging module including a plurality of sets of registers, wherein each set of registers is configured to define a GET/PUT context instance; and
a plurality of data processing cores each coupled to the inter-node messaging module, wherein each of the plurality of data processing cores is configured to map a user level process thereof to one of the plurality of sets of registers and thereby to a respective GET/PUT context instance;
wherein the user level process is configured to utilize the respective GET/PUT context instance to perform a GET/PUT action to a ring buffer of a target data processing node, and wherein the target data processing node is coupled to the data processing node through a fabric; and
wherein each of the plurality of data processing cores is further configured to modify a memory management unit page table to include a virtual address page for the user level process that maps to a physical address page for the respective GET/PUT context instance.

2. The data processing node of claim 1 , wherein:
each of the GET/PUT context instances includes a field for specifying a length of a GET/PUT payload for a respective GET/PUT action;
a first element of the ring buffer is configured to store a first GET/PUT payload for a first GET/PUT action of a first user level process;
a second element of the ring buffer is configured to store a second GET/PUT payload for a second GET/PUT action of a second user level process; and
the first element of the ring buffer is of a different size than the second element of the ring buffer.

3. The data processing node of claim 1 , wherein the user level process is further configured to:
determine a status of the respective GET/PUT context instance; and
populate the respective GET/PUT context instance with information for the GET/PUT action in response to a determination that a previously instantiated GET/PUT action is completed, wherein the previously instantiated GET/PUT action used the same respective GET/PUT context instance.

4. The data processing node of claim 3 , further comprising:
a memory management unit coupled to the plurality of data processing cores;
wherein each of the plurality of data processing cores is further configured to modify a page table to include a virtual address page for the user level process that maps to a physical address page for the respective GET/PUT context instance;
wherein the memory management unit is configured to translate a virtual address specified at the virtual address page to a physical address specified at the physical address page in response to the user level process populating the respective GET/PUT context instance with information for the GET/PUT action; and
wherein the inter-node messaging module is configured to associate the physical address with a GET/PUT module thereof in response to being provided with the physical address and to cause the GET/PUT module to perform the GET/PUT action for the user level process.

5. The data processing node of claim 4 , wherein:
each of the GET/PUT context instances includes a field for specifying a length of a GET/PUT payload for a respective GET/PUT action;
a first element of the ring buffer is configured to store a first GET/PUT payload for a first GET/PUT action of a first user level process;
a second element of the ring buffer is configured to store a second GET/PUT payload for a second GET/PUT action of a second user level process; and
the first element of the ring buffer is of a different size than the second element of the ring buffer.

6. The data processing node of claim 1 , further comprising:
a memory management unit coupled to the plurality of data processing cores; and
a kernel-level driver coupled to the plurality of data processing cores and to the memory management unit;
wherein the kernel-level driver is configured to modify a page table to include a virtual address page for the user level process that maps to a physical address page for the respective GET/PUT context instance in response to being called upon by the user level process in association with initiating the GET/PUT action;
wherein the user level process is further configured to populate the respective GET/PUT context instance with information for the GET/PUT action in response to determining that a previously instantiated GET/PUT action is completed, and wherein the previously instantiated GET/PUT action used the same respective GET/PUT context instance;
wherein the memory management unit is configured to translate a virtual address specified at the virtual address page to a physical address specified at the physical address page in response to population, by the user level process, of the respective GET/PUT context instance with information for the GET/PUT action; and
wherein the inter-node messaging module is configured to associate the physical address with a GET/PUT module thereof in response to being provided with the physical address and to cause the GET/PUT module to perform the GET/PUT action for the user level process.

7. The data processing node of claim 1 , wherein:
the GET/PUT action is a PUT action; and
the respective GET/PUT context instance includes a field that enables a PUT request for the PUT action to include a timestamp value, wherein the time stamp value designates a fabric time at which the PUT action was initiated.

8. The data processing node of claim 7 , wherein:
the data processing node is one of a plurality of data processing nodes connected to each other through a fabric;
the fabric time is determined using a time synchronization service, wherein the time synchronization service is configured to run across the fabric in a distributed manner on the plurality of data processing nodes; and
the time synchronization service is hardware-implemented on each of the plurality of data processing nodes.

9. The data processing node of claim 1 , wherein:
the GET/PUT action is a PUT action;
the inter-node messaging module is configured to generate a PUT request based on information from the respective GET/PUT context instance; and
the PUT request includes fabric time information corresponding to a fabric time at which the PUT action was initiated.

10. The data processing node of claim 1 , wherein:
the data processing node is configured to receive a remote node security key from the target data processing node; and
a GET/PUT request for the GET/PUT action includes an instance of the remote node security key, wherein the target data processing node is configured to verify authenticity of the GET/PUT request based on the instance of the remote node security key.

11. The data processing node of claim 10 , wherein:
the respective GET/PUT context instance includes a field that is configured to hold the remote node security key; and
the inter-node messaging module is configured to generate the GET/PUT request based on information from the respective GET/PUT context instance.

12. A data processing system, comprising:
a target node including a ring buffer in local memory thereof; and
a plurality of initiator nodes connected to each other and to the target node through a fabric, wherein each of the initiator nodes includes:
an inter-node messaging module comprising a plurality of sets of registers, wherein each set of registers is configured to define a GET/PUT context instance; and
a plurality of data processing cores each coupled to the inter-node messaging module, wherein each of the plurality of data processing cores is configured to map a user level process thereof to one of the sets of registers and thereby to a respective GET/PUT con

Assignees

• Iii Holdings 2 Llc

Inventors

• Chandra Prashant R
• Volpe Thomas A
• Davis Mark Bradley
• Dalton Niall Joseph

Classifications

• G06F9/54 Primary

  Interprogram communication · CPC title

• G06F12/0223

  User address space allocation, e.g. contiguous or non contiguous base addressing · CPC title

• G06F15/17375

  One dimensional, e.g. linear array, ring · CPC title

• G06F12/00 Primary

  Accessing, addressing or allocating within memory systems or architectures (digital input from, or digital output to record carriers, e.g. to disk storage units, G06F3/06) · CPC title

• Y02D10/00

  Energy efficient computing, e.g. low power processors, power management or thermal management · CPC title