Automatic data replica manager in distributed caching and data processing systems
A method of data storage includes determining a latency distance from a primary node to each of two or more replica nodes, choosing a preferred replica node of the two or more replica nodes based on the determined latency distances, and write-caching data into the preferred replica node.
1. A method of data storage, the method comprising:
determining a first latency distance from a first node to a second node;
determining a second latency distance from the first node to a third node that is greater than the first latency distance;
writing first data associated with the first node from a fourth node to the second node; and
writing second data from the first node to the second node.
2. The method of claim 1 , wherein the first latency distance and the second latency distance are less than a third latency distance from the first node to the fourth node.
3. The method of claim 1 , wherein the second node is assigned a first ranking indicating a degree to which the second node is suitable for receiving data associated with the first node, and
wherein the third node is assigned a second ranking indicating a degree to which the third node is suitable for receiving data associated with the first node, the first and second rankings respectively being based on network delays associated with paths between the first node and the second node and between the first node and the third node.
4. The method of claim 3 , wherein at least one of the first node, the second node, or the third node comprises a physical host for running virtual machines.
5. The method of claim 4 , wherein one or more virtual machines share a solid-state drive tier and a hard disk drive tier in the first node, the second node, the third node, or the fourth node.
6. The method of claim 3 , wherein the first ranking or the second ranking is further based on workload characteristics comprising access frequency of data associated with the second node or the third node and on service level agreements (SLAs) associated with the second node or the third node.
7. The method of claim 3 , further comprising periodically updating the first ranking or the second ranking to account for changes in network delays.
8. The method of claim 1 , wherein:
the first node comprises a primary node; and
the second node, the third node, and the fourth node comprise replica nodes.
9. The method of claim 1 , wherein the first data comprises replica data, and
wherein the second data is different than the first data.
10. The method of claim 1 , further comprising maintaining the first data on the fourth node.
11. The method of claim 1 , wherein at least one of the first node, the second node, or the third node comprises:
a solid-state drive tier as a cache tier comprising a cache partition for storing data of local virtual machines; and
a replica partition for storing replica data from other nodes.
12. The method of claim 11 , wherein the at least one of the first node, the second node, or the third node comprises a hard disk drive tier as a back-end storage tier.
13. A method of using replicated data, the method comprising:
determining that a difference between an access speed corresponding to a first node and an access speed corresponding to a second node exceeds a threshold; or
determining a utilization ratio of throughput corresponding to the second node is less than an operating threshold; and
reading a first part of a dataset from the first node, and reading a second part of the data set from the second node.
14. The method of claim 13 , wherein:
the first node comprises a primary node of a datacenter cluster, such that the first node is configured to store the first part of the dataset at a cache partition of the first node; and
the second node comprises a replica node of the datacenter cluster, such that the second node is configured to store the second part of the dataset at a replica partition of the second node.
15. The method of claim 13 , further comprising splitting the dataset into the first part and the second part.
16. A method of prefetching data, the method comprising:
splitting a dataset into a first part and a second part;
determining that a first ratio of a total size of the dataset to a first access speed associated with a first node comprising the dataset is greater than or equal to a greater of a second ratio of a first size of the first part to the first access speed and a third ratio of a second size of the second part to a second access speed associated with a second node; and
loading the first part and the second part respectively from the first node and the second node.
17. The method of claim 16 , wherein the first access speed accounts for a first network delay associated with the first node, and wherein the second access speed accounts for a second network delay associated with the second node.
18. The method of claim 16 , further comprising:
determining a ratio of the first access speed to a sum of the first access speed and the second access speed; and
achieving a network makespan equal to a ratio of the total size of the dataset to the sum of the first access speed and the second access speed.
19. The method of claim 16 , further comprising:
determining a difference between the second access speed and the second access speed exceeds a threshold, or determining a utilization ratio of throughput corresponding to the second node is less than an operating threshold; and
triggering parallel prefetching.
20. The method of claim 19 , wherein the first node comprises a primary node, and wherein the second node comprises a replica node.