Methods and systems for maintaining cache coherency between caches of nodes in a clustered environment
Disclosed herein are methods, systems, and processes to provide coherency across disjoint caches in clustered environments. It is determined whether a data object is owned by an owner node, where the owner node is one of multiple nodes of a cluster. If the owner node for the data object is identified by the determining, a request is sent to the owner node for the data object. However, if the owner node for the data object is not identified by the determining, selects a node in the cluster is selected as the owner node, and the request for the data object is sent to the owner node.
1. A computer-implemented method, comprising:
identifying an owner node of a data object, wherein
the owner node is one of a plurality of nodes of a cluster,
the owner node is communicatively coupled to a local storage device,
the local storage device has a first latency,
the plurality of nodes are communicatively coupled to a shared storage device,
the shared storage device has a second latency,
the second latency is greater than the first latency, and
the local storage device and the shared storage device each comprise a copy of the data object;
subsequent to identifying the owner node for the data object, sending a first request from a first node to the owner node for a copy of the data object;
receiving the copy of the data object from the owner node, wherein
subsequent to the receiving, the data object is updated by a second node of the cluster, and
updating the data object results in an updated copy of the data object on the shared storage device and further results in updated copy of the data object on the local storage device coupled to the owner node;
subsequent to the updating, sending a second request from the first node to the owner node for a copy of the data object, wherein
at least one of the first request or the second request is a Remote Direct Memory Access (RDMA) request; and
subsequent to the second request, receiving the updated copy of the data object from the owner node.
2. The computer-implemented method of claim 1 , further comprising:
instructing the owner node to cache the data object prior to sending the first request to the owner node for the data object, if the owner node is not identified by the identifying.
3. The computer-implemented method of claim 1 , further comprising:
receiving the first request for the copy of the data object from an application;
determining whether a master node maintains information identifying the owner node for the data object; and
selecting a node in the cluster as the owner node if the master node does not maintain information identifying the owner node for the data object.
4. The computer-implemented method of claim 3 , wherein
the owner node takes a lock comprising lock information on the data object, and
the lock information is maintained by the master node.
5. The computer-implemented method of claim 1 , further comprising:
selecting a node as the owner node, wherein
selecting the node as the owner node is based on at least one of: a random selection method, a circular selection method, a cache characteristic selection method, and node priority.
6. The computer-implemented method of claim 2 , further comprising:
selecting a node as the owner node, wherein
selecting the node as the owner node is performed by a reader node; and
a subsequent request for a copy of the data object from another reader node is serviced by the owner node by virtue of
the another reader node querying a master node for lock information maintained by a distributed lock manager, and
the another reader node determining that the owner node holds a lock on the data object.
7. The computer-implemented method of claim 6 , wherein
the owner node, the master node, the reader node, and the another reader node are part of the plurality of nodes,
the local storage device is a Solid State Drive (SSD), and
the shared storage device is a Hard Disk Drive (HDD).
8. A non-transitory computer readable storage medium comprising program instructions executable to perform a method comprising:
identifying an owner node of a data object, wherein
the owner node is one of a plurality of nodes of a cluster,
the owner node is communicatively coupled to a local storage device,
the local storage device has a first latency,
the plurality of nodes are communicatively coupled to a shared storage device,
the shared storage device has a second latency,
the second latency is greater than the first latency, and
the local storage device and the shared storage device each comprise a copy of
the data object;
subsequent to identifying the owner node for the data object, sending a first request from a first node to the owner node for a copy of the data object;
receiving the copy of the data object from the owner node, wherein
subsequent to the receiving, the data object is updated by a second node of the cluster, and
updating the data object results in an updated copy of the data object on the shared storage device and further results in updated copy of the data object on the local storage device coupled to the owner node;
subsequent to the updating, sending a second request from the first node to the owner node for a copy of the data object, wherein
at least one of the first request or the second request is a Remote Direct Memory Access (RDMA) request; and
subsequent to the second request, receiving the updated copy of the data object from the owner node.
9. The non-transitory computer readable storage medium of claim 8 , wherein the method further comprises:
instructing the owner node to cache the data object prior to sending the first request to the owner node for the data object, if the owner node is not identified by the identifying.
10. The non-transitory computer readable storage medium of claim 8 , wherein the method further comprises:
receiving the first request for the copy of the data object from an application;
determining whether a master node maintains information identifying the owner node for the data object; and
selecting a node in the cluster as the owner node if the master node does not maintain information identifying the owner node for the data object.
11. The non-transitory computer readable storage medium of claim 10 , wherein
the owner node takes a lock comprising lock information on the data object, and
the lock information is maintained by the master node.
12. The non-transitory computer readable storage medium of claim 8 , wherein the method further comprises:
selecting a node as the owner node, wherein
selecting the node as the owner node is based on at least one of: a random selection method, a circular selection method, a cache characteristic selection method, and node priority.
13. The non-transitory computer readable storage medium of claim 9 , wherein the method further comprises:
selecting a node as the owner node, wherein
selecting the node as the owner node is performed by a reader node; and
a subsequent request for a copy of the data object from another reader node is serviced by the owner node by virtue of
the another reader node querying a master node for lock information maintained by a distributed lock manager, and
the another reader node determining that the owner node holds a lock on the data object.
14. The non-transitory computer readable storage medium of claim 13 , wherein
the owner node, the master node, the reader node, and the another reader node are part of the plurality of nodes,
the local storage device is a Solid State Drive (SSD), and
the shared storage device is a Hard Disk Drive (HDD).
15. A system comprising:
one or more processors; and
a memory coupled to the one or more processors, wherein the memory stores program instructions executable by the one or more processors to perform a method comprising:
identifying an owner node of a data object, wherein
the owner node is one of a plurality of nodes of a cluster,
the owner node is communicatively coupled to a local storage device,
the local storage device has a first latency,
the plurality of nodes are communicatively coupled to a shared storage device,
the shared storage device has a second latency,
the second latency is greater than the first latency, and
the local storage device and the shared storage device each comprise a copy of the data object;
subsequent to identifying the owner node for the data object, sending a first request from a first node to the owner node for a copy of the data object;
receiving the copy of the data object from the owner node, wherein
subsequent to the receiving, the data object is updated by a second node of the cluster, and
updating the data object results in an updated copy of the data object on the shared storage device and further results in updated copy of the data object on the local storage device coupled to the owner node;
subsequent to the updating, sending a second request from the first node to the owner node for a copy of the data object, wherein
at least one of the first request or the second request is a Remote Direct Memory Access (RDMA) request; and
subsequent to the second request, receiving the updated copy of the data object from the owner node.
16. The system of claim 15 , wherein the method further comprises:
instructing the owner node to cache the data object prior to sending the first request to the owner node for the data object, if the owner node is not identified by the identifying.
17. The system of claim 15 , wherein the method further comprises:
receiving the first request for the copy of the data object from an application;
determining whether a master node maintains information identifying the owner node for the data object; and
selecting a node in the cluster as the owner node if the master node does not maintain information identifying the owner node for the data object.
18. The system of claim 17 , wherein
the owner node takes a lock comprising lock information on the data object,
the lock information is maintained by master node, and
selecting a node as the owner node, wherein
selecting the node as the owner node is based on at least one of: a random selection method, a circular selection method, a cache characteristic selection method, or node priority.
19. The system of claim 18 , wherein
selecting a node as the owner node, wherein
selecting the node as the owner node is performed by a reader node; and
a subsequent request for a copy of the data object from another reader node is serviced by the owner node by virtue of
the another reader node querying a master node for lock information maintained by a distributed lock manager, and
the another reader node determining that the owner node holds a lock on the data object.
20. The system of claim 19 , wherein
the owner node, the master node, the reader node, and the another reader node are part of the plurality of nodes,
the local storage device is a Solid State Drive (SSD), and
the shared storage device is a Hard Disk Drive (HDD).
21. The method of claim 1 , wherein
the first request is a direct access request to read the data object from the local storage device.