Method and system for multi-layer differential load balancing in tightly coupled clusters
A method and system for load balancing. The method includes discovering each of a plurality of hosts in a cluster, wherein the plurality of hosts is configured for accessing a LUN of a storage system through a storage network fabric. Global input/output (I/O) load characteristics are collected for each of the plurality of hosts at the device and/or volume level. A selected host is determined for processing an I/O originating at the local host, wherein the host is selected based on a current set of the global I/O load characteristics.
1. A computer implemented method for load balancing, comprising:
at a local volume manager associated with a local host, discovering each of a plurality of remote volume managers, wherein each remote volume manager is associated with a remote host of a plurality of hosts in a cluster, wherein said plurality of hosts is configured for accessing a LUN of a storage system through a storage network fabric;
at said local volume manager, collecting global I/O load characteristics from each remote volume manager for each corresponding associated remote host of said plurality of hosts in said cluster; and
determining a selected host for processing an I/O originating at said local host based on a current set of said global I/O load characteristics and on a rate of received I/O from each remote volume manager for said each corresponding associated remote host of said plurality of hosts in said cluster.
2. The computer implemented method of claim 1 , wherein determining the selected host comprises:
determining to ship said I/O from said local host based on said current set of said global I/O load characteristics to said selected host comprising a remote host; and
selecting said remote host from one or more remote hosts based on favorable differential I/O load balancing statistics determined in comparison to said local host; and
redirecting said I/O from said first host to said remote host.
3. The computer implemented method of claim 1 , wherein said global I/O load characteristics is taken from a group consisting of I/O throughput, I/O response time, network latency between a corresponding remote host and said local host, I/O load at each of said plurality of hosts, and resource availability.
4. The computer implemented method of claim 1 , further comprising:
determining to process said I/O at said LUN based on said current set of said global I/O load characteristics, wherein said selected host comprises said LUN;
at a multi-pathing driver of said local host, performing local load balancing to assign said I/O to a selected path of one or more local communication paths accessing said LUN.
5. The computer implemented method of claim 1 , wherein said determining a selected host further comprises:
determining said selected host having the best I/O load characteristics for an I/O having a high priority.
6. The computer implemented method of claim 1 , wherein said determining a selected host further comprises:
determining said selected host having the poorest I/O load characteristics for said I/O having a low priority.
7. The computer implemented method of claim 1 , further comprising:
collecting values of said current set of said global statistics in a delayed and amortized fashion for said plurality of hosts.
8. The computer implemented method of claim 1 , further comprising:
providing a plurality of default values for said global statistics.
9. A computer system, comprising:
a processor; and
memory coupled to said processor and having stored therein instructions that, if executed by said computer system, cause said computer system to execute a method for load balancing, comprising:
at a local volume manager associated with a local host, collecting global I/O load characteristics from each remote volume manager of a plurality of remote volume managers, wherein each remote volume manager is associated with a remote host of a plurality of remote hosts in a cluster, wherein said plurality of remote hosts are configured for accessing a LUN of a storage system through a storage network fabric; and
determining a selected host for processing an I/O originating at said local host based on a current set of said global I/O load characteristics and on a rate of received I/O from each remote volume manager for said each corresponding associated remote host of the plurality of remote hosts in said cluster.
10. The computer system of claim 9 , wherein determining the selected host comprises:
determining to ship said I/O from said local host based on said current set of said global I/O load characteristics to said selected host comprising a remote host; and
selecting said remote host from one or more remote hosts based on favorable differential I/O load balancing statistics determined in comparison to said local host; and
redirecting said I/O from said first host to said remote host.
11. The computer system of claim 9 , wherein in said method said global I/O load characteristics is taken from a group consisting of I/O throughput, I/O response time, network latency between a corresponding remote host and said local host, I/O load at each of said plurality of hosts, and resource availability.
12. The computer system of claim 9 , wherein said method further comprises:
determining to process said I/O at said LUN based on said current set of said global I/O load characteristics, wherein said selected host comprises said LUN; and
at a multi-pathing driver of said local host, performing local load balancing to assign said I/O to a selected path of one or more local communication paths accessing said LUN.
13. A non-transitory computer-readable storage medium having computer-executable instructions for causing a computer system to perform a method for load balancing, comprising:
at a local volume manager associated with a local host, collecting global I/O load from each remote volume manager of a plurality of remote volume managers, wherein each remote volume manager is associated with a remote host of a plurality of remote hosts in a cluster, wherein said plurality of remote hosts are configured for accessing a LUN of a storage system through a storage network fabric; and
determining a selected host for processing an I/O originating at said local host based on a current set of said global I/O load characteristics and on a rate of received I/O from each remote volume manager for said each corresponding associated remote host of said plurality of remote hosts in the said cluster.
14. The computer-readable storage medium of claim 13 , wherein determining the selected host comprises:
determining to ship said I/O from said local host based on said current set of said global I/O load characteristics to said selected host comprising a remote host; and
selecting said remote host from one or more remote hosts based on favorable differential I/O load balancing statistics determined in comparison to said local host; and
redirecting said I/O from said first host to said remote host.
15. The computer-readable storage medium of claim 13 , wherein said global I/O load characteristics is taken from a group consisting of I/O throughput, I/O response time, network latency between a corresponding remote host and said local host, I/O load at each of said plurality of hosts, and resource availability.
16. A computer-readable storage medium of claim 13 , wherein said method further comprises:
determining to process said I/O at said LUN based on said current set of said global I/O load characteristics, wherein said selected host comprises said LUN; and
at a multi-pathing driver of said local host, performing local load balancing to assign said I/O to a selected path of one or more local communication paths accessing said LUN.
17. The computer-readable storage medium of claim 13 , wherein said determining a selected host further comprises:
determining said selected host having the best I/O load characteristics for an I/O having a high priority.
18. The computer-readable storage medium of claim 13 , wherein said determining a selected host further comprises:
determining said selected host having the poorest I/O load characteristics for said I/O having a low priority.
19. The computer-readable storage medium of claim 13 , wherein said method further comprises:
collecting values of said current set of said global statistics in a delayed and amortized fashion for said plurality of hosts.
20. The computer-readable storage medium of claim 13 , wherein said method further comprises:
providing a plurality of default values for said global statistics.
21. A computer-implemented method for load balancing comprising:
receiving, at a local load balancer associated with a local host in a cluster, input/output load characteristics of at least one remote load balancer associated with a remote host in the cluster;
measuring a rate of receiving remote input/output requests from the at least one remote load balancer; and
throttling the rate of sending local input/output requests to the at least one remote host based on the input/output load characteristics of the at least one remote host and on the rate of receiving remote input/output requests from the at least one remote host.