Peer arrangements among switch-coupled endpoint devices
Peer-to-peer arrangements between endpoint devices are provided herein. A method includes establishing synthetic devices representing endpoint devices in an address domain associated with a host processor, where the endpoint devices have a different address domain than the host processor. The method also includes forming a peer arrangement between the endpoint devices such that data transfers between the endpoint devices in the different address domain can be initiated by the host processor interfacing with the synthetic devices.
1. A method, comprising:
initiating a first isolation function in a first communication switch to form a peer-to-peer arrangement between first endpoint devices coupled to the first communication switch, wherein the first isolation function isolates a first address domain associated with the first endpoint devices from at least a second address domain associated with a host processor by at least establishing first synthetic devices representing the first endpoint devices in the second address domain; and
initiating a second isolation function in a second communication switch to form a peer-to-peer arrangement between second endpoint devices coupled to the second communication switch, wherein the second isolation function isolates a third address domain associated with the second endpoint devices from at least the second address domain associated with the host processor by at least establishing second synthetic devices representing the second endpoint devices in the third address domain.
2. The method of claim 1 , further comprising:
communicatively coupling the first communication switch and the second communication switch over a communication interface such that the host processor can communicate with any among the first communication switch and the second communication switch.
3. The method of claim 2 , wherein the communication interface comprises at least one among Peripheral Component Interconnect Express (PCIe), Ethernet, Infiniband, or FibreChannel.
4. The method of claim 2 , wherein the communication interface comprises a communication protocol different than the first endpoint devices use to communicate with the first communication switch.
5. The method of claim 1 , wherein the first endpoint devices and the second endpoint devices each comprise a graphics processing unit (GPU).
6. The method of claim 1 , wherein the endpoint devices comprise at least one among graphics processing units (GPUs), solid state drives (SSDs), hard disk drives (HDDs), memory devices, or network interface elements.
7. The method of claim 1 , wherein the first endpoint devices are included in a first chassis or enclosure and the second endpoint devices are included in a second chassis or enclosure.
8. The method of claim 1 , further comprising:
redirecting traffic transferred by the host processor for the first endpoint devices in the second address domain for delivery to corresponding ones of the first endpoint devices in the first address domain; and
redirecting traffic transferred by the host processor for the second endpoint devices in the second address domain for delivery to corresponding ones of the second endpoint devices in the third address domain.
9. The method of claim 1 , further comprising:
adding a selected endpoint device into the first peer-to-peer arrangement by at least instantiating the selected endpoint device in the second address domain and initiating a synthetic device representing the selected endpoint device in the first address domain; and
removing the selected endpoint device from the peer-to-peer arrangement by at least inactivating the synthetic device representing the selected endpoint device in the first address domain.
10. The method of claim 1 , further comprising:
establishing the first isolation function in the first communication switch by at least providing first address traps that monitor for traffic in the first communication switch directed to first addresses associated with the first endpoint devices in the second address domain and translate the first addresses into corresponding addresses for the first endpoint devices in the first address domain; and
establishing the second isolation function in the second communication switch by at least providing second address traps that monitor for traffic in the second communication switch directed to third addresses associated with the second endpoint devices in the second address domain and translate the third addresses into corresponding addresses for the second endpoint devices in the third address domain.
11. A system, comprising:
a control processor configured to initiate a first isolation function in a first communication switch to form a peer-to-peer arrangement between first endpoint devices coupled to the first communication switch, wherein the first isolation function isolates a first address domain associated with the first endpoint devices from at least a second address domain associated with a host processor by at least establishing first synthetic devices representing the first endpoint devices in the second address domain; and
the control processor configured to initiate a second isolation function in a second communication switch to form a peer-to-peer arrangement between second endpoint devices coupled to the second communication switch, wherein the second isolation function isolates a third address domain associated with the second endpoint devices from at least the second address domain associated with the host processor by at least establishing second synthetic devices representing the second endpoint devices in the third address domain.
12. The system of claim 11 , comprising:
a communication interface configured to communicatively couple the first communication switch and the second communication switch such that the host processor can communicate with any among the first communication switch and the second communication switch.
13. The system of claim 12 , wherein the communication interface comprises at least one among Peripheral Component Interconnect Express (PCIe), Ethernet, Infiniband, or FibreChannel.
14. The system of claim 12 , wherein the communication interface comprises a communication protocol different than the first endpoint devices use to communicate with the first communication switch.
15. The system of claim 11 , wherein the first endpoint devices and the second endpoint devices each comprise a graphics processing unit (GPU).
16. The system of claim 11 , wherein the endpoint devices comprise at least one among graphics processing units (GPUs), solid state drives (SSDs), hard disk drives (HDDs), memory devices, or network interface elements.
17. The system of claim 11 , wherein the first endpoint devices are included in a first chassis or enclosure and the second endpoint devices are included in a second chassis or enclosure.
18. The system of claim 11 , further comprising:
the first isolation function configured to redirect traffic transferred by the host processor for the first endpoint devices in the second address domain for delivery to corresponding ones of the first endpoint devices in the first address domain; and
the second isolation function configured to redirect traffic transferred by the host processor for the second endpoint devices in the second address domain for delivery to corresponding ones of the second endpoint devices in the third address domain.
19. The system of claim 11 , further comprising:
the control processor configured to add a selected endpoint device into the first peer-to-peer arrangement by at least instantiating the selected endpoint device in the second address domain and initiating a synthetic device representing the selected endpoint device in the first address domain; and
the control processor configured to remove the selected endpoint device from the peer-to-peer arrangement by at least inactivating the synthetic device representing the selected endpoint device in the first address domain.
20. An apparatus comprising:
one or more non-transitory computer readable storage media; and
program instructions stored on the one or more non-transitory computer readable storage media, that when executed by a processing system, direct the processing system to at least:
initiate a first isolation function in a first communication switch to form a peer-to-peer arrangement between first endpoint devices coupled to the first communication switch, wherein the first isolation function isolates a first address domain associated with the first endpoint devices from at least a second address domain associated with a host processor by at least establishing first synthetic devices representing the first endpoint devices in the second address domain; and
initiate a second isolation function in a second communication switch to form a peer-to-peer arrangement between second endpoint devices coupled to the second communication switch, wherein the second isolation function isolates a third address domain associated with the second endpoint devices from at least the second address domain associated with the host processor by at least establishing second synthetic devices representing the second endpoint devices in the third address domain.