IP Library Granted Patent US 8,862,865
Granted Patent B2
US 8,862,865 · App. 13/221,093 · Granted Oct 14, 2014

Rebooting infiniband clusters

Inventor: Constantine Gavrilov (Tel Aviv, IL)
Assignee: International Business Machines Corporation
G06F9/4401
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 8,862,865
App. No.
13/221,093
Granted
Oct 14, 2014
Kind
B2
Abstract

Methods, systems, and physical computer-readable storage media are now provided that reduce latency and improve efficiency of a system after reboot of an Infiniband (IB) cluster. A method of rebooting an Infiniband cluster includes preserving routing table configurations of each switch of the Infiniband cluster, rebooting endpoints of the Infiniband cluster, activating physical links at each endpoint, setting local identifier (LID) locally at each endpoint of the IB cluster, and activating logical links locally at each endpoint of the IB cluster.

Claims (35)

1. An Infiniband cluster system comprising:

a plurality of nodes including endpoints; and

at least one switch connecting at least two nodes,

wherein:

each endpoint includes a processor, and

during an Infiniband cluster system reboot, each endpoint locally sets a corresponding local identifier and activates a corresponding logical link.

2. The IB cluster system of claim 1 , wherein the at least one switch includes a routing table.

3. The IB cluster system of claim 1 , wherein at least one node of the plurality of nodes comprises a subnet manager provided with a mapping table including LIDs assigned to corresponding GUIDs.

4. A method of rebooting an Infiniband cluster, the method comprising:

preserving routing table configurations of each switch of the Infiniband cluster;

rebooting endpoints of the Infiniband cluster;

activating physical links by each rebooted endpoint;

setting local identifier (LID) locally by each endpoint; and

activating logical links locally by each endpoint.

5. The method of claim 4 , further comprising storing a present state of fabric before the step of rebooting.

6. The method of claim 4 , further comprising initiating subnet manager services on the rebooted endpoints, after the steps of setting LID and activating logical links locally.

7. The method of claim 4 , further comprising loading drivers on the rebooted endpoints, after the step of rebooting.

8. The method of claim 4 , wherein the step of activating logical links locally comprises sending a local MAD packet to set the LID.

9. The method of claim 8 , wherein the step of activating logical links locally further comprises sending MAD packets to switch a port and corresponding endpoint connected to the port to ARMED logical state.

10. The method of claim 9 , wherein the step of activating logical links locally further comprises sending MAD packets to switch a port and corresponding endpoint connected to the port to ACTIVE logical state.

11. The method of claim 4 , further comprising initiating subnet manager services after the steps of setting LID and activating logical links.

12. The method of claim 11 , further comprising storing a present state of fabric before the step of rebooting, and providing the present state of fabric to the subnet manager after the steps of setting LID and activating logical links.

13. The method of claim 11 , further comprising updating the routing table of the switch after discovering one of a new endpoint and a previously-known endpoint.

14. The method of claim 4 , further comprising locally predicting a LID associated with a global unique identifier of a corresponding endpoint, before the steps of setting LID and activating logical links.

15. The method of claim 4 , further comprising providing a mapping table to the subnet manager, the mapping table including global unique identifiers of each endpoint to a corresponding LID.

16. A non-transitory physical computer-readable storage medium comprising a computer program product method for rebooting an Infiniband cluster, the physical computer-readable storage medium comprising:

computer code for preserving routing table configurations of each switch of the Infiniband cluster;

computer code for rebooting endpoints of the Infiniband cluster;

computer code for activating physical links by each endpoint; and

computer code for setting local identifier (LID) locally by each endpoint of the Infiniband cluster; and

computer code activating logical links locally by each endpoint of the Infiniband cluster.

17. The non-transitory physical computer-readable storage medium of claim 16 , further comprising computer code for storing a present state of fabric before the step of rebooting.

18. The non-transitory physical computer-readable storage medium of claim 16 , further comprising computer code for initiating subnet manager services on the rebooted endpoints.

19. The non-transitory physical computer-readable storage medium of claim 16 , further comprising computer code for loading drivers on the rebooted endpoints.

20. The non-transitory physical computer-readable storage medium of claim 16 , further comprising computer code for initiating subnet manager services.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Sep 16, 2011
From: GAVRILOV, CONSTANTINE
To: INTERNATIONAL BUSINESS MACHINES CORPORATION
Reel/Frame 026920/0687 →
Continuity (1)
Related Publication 20130054947A1 · Feb 28, 2013