Volatile memory data recovery based on independent processing unit data access
In a server system, a host computing platform can have a processing unit separate from the host processor to detect and respond to failure of the host processor. The host computing platform includes a memory to store data for the host processor. The processing unit has an interface to the host processor and the memory and an interface to a network external to the host processor and has access to the memory. In response to detection of failure of the host processor, the processing unit migrates data from the memory to another memory or storage.
1 . A server system, comprising:
a processing core of a first computing platform;
a volatile memory on the first computing platform to store data for the processing core; and
a processing unit having access to the volatile memory, the processing unit including an interface to the processing core and the volatile memory and including an interface to a network external to the processing core, the processing core to offload processing tasks to the processing unit while the processing core is available, the processing unit to detect a failure of the processing core that makes the processing core unavailable, and migrate the data from the volatile memory to a second computing platform in response to detection of the failure.
2 . The server system of claim 1 , wherein the processing core comprises a central processing unit (CPU) or a graphics processing unit (GPU).
3 . The server system of claim 1 , wherein the volatile memory comprises a random access memory (RAM) device or a cache device.
4 . The server system of claim 1 , wherein the volatile memory comprises a multitier memory hierarchy.
5 . The server system of claim 1 , wherein the processing unit comprises an intelligent processing unit coupled to the processing core via a peripheral component interconnect express (PCIe) link.
6 . The server system of claim 1 , wherein the second computing platform comprises a remote storage platform or a pooled memory platform.
7 . The server system of claim 1 , wherein the first computing platform comprises a first server node and the second computing platform comprises a second server node.
8 . The server system of claim 1 , wherein the processing unit is to migrate selected critical data from the volatile memory to the second computing platform.
9 . The server system of claim 1 , wherein the processing unit is to migrate a selected address range of the volatile memory to the second computing platform.
10 . The server system of claim 1 , wherein the data includes cryptographic keys and computational results.
11 . The server system of claim 1 , wherein the processing unit has a different computational domain from the processing core.
12 . The server system of claim 11 , wherein the processing unit is to continue to operate after the failure of the processing core.
13 . An apparatus comprising a computer readable storage medium having content stored thereon, which when executed causes a computing system to perform a method for mitigating server failure including:
selecting data of a volatile memory to protect in event of a failure of a host processor of a first computing platform by a host operating system executing on the host processor;
identifying the data to a processing unit having an interface to the host processor and the volatile memory and having an interface to a network external to the host processor; and
offloading processing tasks to the processing unit;
wherein the processing unit is to detect a failure of the host processor that makes the host processor unavailable and migrate the data from the volatile memory to a second computing platform in response to detection of the failure.
14 . The apparatus of claim 13 , wherein selecting the data of the volatile memory to protect comprises determining critical data to be migrated in case the host processor fails.
15 . The apparatus of claim 13 , wherein selecting the data of the volatile memory to protect comprises specifying address ranges to store cryptographic keys and computational results.
16 . The apparatus of claim 13 , wherein identifying the data to the processing unit comprises passing address ranges to the processing unit.
17 . A server system, comprising:
multiple computing platforms having a host processor with a host operating system, a volatile memory to store data for the host processor, and an infrastructure processing unit (IPU) separate from the host processor with a separate operating system, the IPU having access to the volatile memory; and
an orchestrator server node to manage the multiple computing platforms;
wherein the IPU of a first computing platform of the multiple computing platforms capable to detect a failure of the host processor of the first computing platform that makes the host processor unavailable, notify the orchestrator of the failure, and migrate the data out of the volatile memory in response to detection of the failure.
18 . The server system of claim 17 , wherein the IPU of the first computing platform is to migrate the data to a remote storage platform or to a pooled memory platform or to a second computing platform.
19 . The server system of claim 17 , wherein the IPU of the first computing platform is to migrate the data to an internal storage of the first computing platform.
20 . The server system of claim 17 , wherein the IPU is configured to pass to the orchestrator a list of services with process address space identifiers (PASIDs) for the services and memory ranges to migrate from the volatile memory.
21 . The server system of claim 17 , wherein the orchestrator is to instantiate a new instance of an application associated with the migrated data on a second computing platform different from the first computing platform prior to completion of migration of the data.
22 . The server system of claim 21 , wherein the IPU is to send requests from the new instance of the application to the second computing platform.