Accelerating data intensive computing applications by reducing overhead associated with communication protocols

• US 8,671,414 B1
• Filed: 11/02/2011
• Issued: 03/11/2014
• Est. Priority Date: 05/09/2011
• Status: Active Grant

• Current Assignee: Emc IP Holding Company LLC (Dell Technologies Inc.)
• Original Assignee: EMC Corporation (Dell Technologies Inc.)
• Inventors: Forecast, John, Loafman, Zachary Merlynn, Paterra, Frank Charles, Sunny Veliath, Deepak, Cardente, John Thomas
• Primary Examiner(s): Ho, Andy
• Assistant Examiner(s): Seye, Abdou
• Application Number:
  US13/287,866
• Time in Patent Office: 860 Days
• Field of Search: None
• US Class Current: 719/312
• CPC Class Codes: G06F 2009/45595 Network integration; Enabling network access in virtual machine instances
• CPC Class Codes: G06F 9/45545 Guest-host, i.e. hypervisor is an application program itself, e.g. VirtualBox
• CPC Class Codes: G06F 9/544 Buffers; Shared memory; Pipes
• CPC Class Codes: G06F 9/545 where tasks reside in different layers, e.g. user- and kernel-space

Abstract

Various embodiments are directed toward improving Virtual Machine performance by reducing the amount of communication protocol processing performed between co-resident Virtual Machine guest applications executing on the same physical machine. Overhead can be reduced by short-circuiting the full communication protocol stack when a plurality of Virtual Machine guest applications are hosted and executing on the same physical machine such as a data storage node or a compute accelerator node. Improvements may also be realized by enabling the sharing in virtual memory of physical memory pages between co-resident Virtual Machine guest applications, and employing a low latency high bandwidth network for enabling guest applications to communicate with each other that are located in different nodes in a data storage cluster.

Claims

1. A method for accelerating communication between a first guest application and a second guest application executing on at least one data storage node, comprising:

generating a first virtual machine for the first guest application and a second virtual machine for the second guest application, wherein each virtual machine includes at least a virtual memory and a multiple layer communication protocol stack based at least in part on a plurality of layers of the Open Systems Interconnect model;

if the first guest application and the second guest application are determined to be co-resident on a data storage node, bypassing each lower layer of the communication protocol stack to directly communicate a request for data from the first guest application to the second guest application at a highest layer of the corresponding communication protocol stacks of their virtual machines, wherein the highest layer is an application layer that enables communication independent of a TCP/IP layer socket number;

determining a pointer to data requested by the first guest application, wherein the data resides in at least one physical page of memory at the data storage node and is controlled by the second guest application, wherein the first guest application maps the at least physical page into its virtual memory and the second guest application releases the at least one physical page for use by the first guest application;

providing the pointer to the first guest application to access the data in the at least one physical page of memory at the data storage node; and

if the first guest application and the second guest application are determined to be co-resident in a cluster on separate data storage nodes, employing a first network coupled between the separate data storage nodes for communication between the first and second guest applications, wherein the first network is a high bandwidth network.

2. The method of claim 1, further comprising employing a second network for communication between the first and second guest applications if the first guest application or the second guest application is determined to located on another node that is outside the cluster of nodes, wherein the second network provides lower bandwidth communication than the first network.

3. The method of claim 1, wherein if the first and second guest applications are co-resident in the data storage node further comprises transferring control to the first guest application from the second guest application for requested data residing in at least one physical page of memory at the data storage node.

4. The method of claim 1, wherein if the first and second guest applications are co-resident in the same data storage node further comprising:

sharing control between the first guest application and the second guest application for requested data residing in at least one physical page of memory at the data storage node; and

copying requested data from the at least one physical page of memory into each virtual memory of each virtual machine for the corresponding first guest application and the second guest application if either guest application writes to the requested data having shared control.

5. A data storage node apparatus for accelerating communication between a first guest application and a second guest application, comprising:

a physical memory for storing data and instructions;

a data storage device for storing data and instructions; and

a processor for executing instructions that perform actions, including;

enabling generating a first virtual machine for the first guest application and a second virtual machine for the second guest application, wherein each virtual machine includes at least a virtual memory and a multiple layer communication protocol stack based at least in part on a plurality of layers of the Open Systems Interconnect model;

if the first guest application and the second guest application are determined to be co-resident on a data storage node, bypassing each lower layer of the communication protocol stack to directly communicate a request for data from the first guest application to the second guest application at a highest layer of the corresponding communication protocol stacks of their virtual machines, wherein the highest layer is an application layer that enables communication independent of a TCP/IP layer socket number;

determining a pointer to data requested by the first guest application, wherein the data resides in at least one physical page of memory at the data storage node and is controlled by the second guest application, wherein the first guest application maps the at least physical page into its virtual memory and the second guest application releases the at least one physical page for use by the first guest application;

providing the pointer to the first guest application to access the data in the at least one physical page of memory at the data storage node; and

if the first guest application and the second guest application are determined to be co-resident in a cluster on separate data storage nodes, employing a first network coupled between the separate data storage nodes for communication between the first and second applications, wherein the first network is a high bandwidth network.

6. The apparatus of claim 5, further comprising employing a second network for communication between the first and second guest applications if the first guest application or the second guest application is determined to located on another node that is outside the cluster of nodes, wherein the second network provides lower bandwidth communication than the first network.

7. The apparatus of claim 5, wherein if the first and second guest applications are co-resident in the data storage node further comprises transferring control to the first guest application from the second guest application for requested data residing in at least one physical page of memory at the data storage node.

8. The apparatus of claim 5, wherein if the first and second guest applications are co-resident in the same data storage node further comprising:

sharing control between the first guest application and the second guest application for requested data residing in at least one physical page of memory at the data storage node; and

copying requested data from the at least one physical page of memory into each virtual memory of each virtual machine for the corresponding first guest application and the second guest application if either guest application writes to the requested data having shared control.

9. A processor readable non-volatile hardware storage media that includes instructions for accelerating communication between a first guest application and a second guest application executing on at least one data storage node, wherein the execution of the instructions causes actions, comprising:

generating a first virtual machine for the first guest application and a second virtual machine for the second guest application, wherein each virtual machine includes at least a virtual memory and a multiple layer communication protocol stack based at least in part on a plurality of layers of the Open Systems Interconnect model;

if the first guest application and the second guest application are determined to be co-resident on a data storage node, bypassing each lower layer of the communication protocol stack to directly communicate a request for data from the first guest application to the second guest application at a highest layer of the corresponding communication protocol stacks of their virtual machines, wherein the highest layer is an application layer that enables communication independent of a TCP/IP layer socket number;

determining a pointer to data requested by the first guest application, wherein the data resides in at least one physical page of memory at the data storage node and is controlled by the second guest application, wherein the first guest application maps the at least physical page into its virtual memory and the second guest application releases the at least one physical page for use by the first guest application;

providing the pointer to the first guest application to access the data in the at least one physical page of memory at the data storage node; and

if the first guest application and the second guest application are determined to be co-resident in a cluster on separate data storage nodes, employing a first network coupled between the separate data storage nodes for communication between the first and second guest applications, wherein the first network is a high bandwidth network.

10. The media of claim 9, further comprising employing a second network for communication between the first and second guest applications if the first guest application or the second guest application is determined to located on another node that is outside the cluster of nodes, wherein the second network provides lower bandwidth communication than the first network.

11. The media of claim 9, wherein if the first and second guest applications are co-resident in the data storage node further comprises transferring control to the first guest application from the second guest application for requested data residing in at least one physical page of memory at the data storage node.

12. The media of claim 9, wherein if the first and second guest applications are co-resident in the same data storage node further comprising:

sharing control between the first guest application and the second guest application for requested data residing in at least one physical page of memory at the data storage node; and

copying requested data from the at least one physical page of memory into each virtual memory of each virtual machine for the corresponding first guest application and the second guest application if either guest application writes to the requested data having shared control.