US 12,425,293 B2
Self-healing network of infrastructure processing units
Reshma Lal, Portland, OR (US); Pallavi Dhumal, Folsom, CA (US); Shubha Bommalingaiahnapallya, East Brunswick, NJ (US); and Asmae Mhassni, Portland, OR (US)
Assigned to Intel Corporation, Santa Clara, CA (US)
Filed by Intel Corporation, Santa Clara, CA (US)
Filed on Dec. 7, 2021, as Appl. No. 17/544,595.
Prior Publication US 2022/0094590 A1, Mar. 24, 2022
Int. Cl. H04L 41/0668 (2022.01); G06F 11/20 (2006.01); H04L 41/0604 (2022.01); H04L 41/0893 (2022.01); H04L 41/0897 (2022.01); H04L 41/344 (2022.01); G06F 11/14 (2006.01)
CPC H04L 41/0668 (2013.01) [G06F 11/203 (2013.01); H04L 41/0613 (2013.01); H04L 41/0893 (2013.01); H04L 41/0897 (2022.05); H04L 41/344 (2022.05); G06F 11/1441 (2013.01)] 20 Claims
OG exemplary drawing
 
8. A method performed in a networked environment including:
a plurality of compute platforms having infrastructure processing units (IPUs), at least a portion of the plurality of compute platforms including one or more accelerators comprising other processing units (XPUs);
managing, via a first IPU on a first compute platform, one or more XPUs in a first XPU cluster on the first compute platform;
detecting the first IPU has failed or is unavailable;
identifying a second IPU on a second compute platform that is comparable to the first IPU; and
migrating management of the one or more XPUs in the first XPU cluster on the first platform to the second IPU.