System and method for fine-grained power control management in a high capacity computer cluster
One embodiment of the present invention provides a system for a fine-grained power management. The system receives, by a first server, a task assigned based on a global power state, wherein the first server includes one or more drives, a drive includes one or more channels, and a channel includes one or more integrated circuits. The system places an inactive drive into a power-saving mode. The system places an inactive channel of an active drive into the power-saving mode by using a power switch associated with each integrated circuit in the inactive channel. The system places an inactive integrated circuit of an active channel into the power-saving mode by using a power switch associated with the inactive integrated circuit. The system updates a data structure storing the global power state based on a current power state of the first server, the drives, the channels, and the integrated circuits.
1. A computer-implemented method for facilitating a fine-grained power management of a high-density computer cluster, the method comprising:
receiving, by a first server in a plurality of servers associated with a rack, a task assigned based on a current global power state associated with the servers,
wherein the current global power state comprises a global power map stored in a data structure which indicates a current power state of the plurality of servers associated with the rack, each of a plurality of drives in a respective server, each of a plurality of channels in a respective drive, and each of a plurality of integrated circuits in a respective channel;
placing an inactive drive of the first server into a power-saving mode to obtain an updated power state of the drives of the first server;
placing an inactive channel of an active drive of the first server into the power-saving mode by using a power switch associated with each integrated circuit in the inactive channel, to obtain an updated power state of the channels of the drives;
placing an inactive integrated circuit of an active channel of an active drive of the first server into the power-saving mode by using a power switch associated with the inactive integrated circuit, to obtain an updated power state of the integrated circuits of the channels;
obtaining an updated power state of the first server based on the updated power state of the inactive drive, the inactive channel, and the inactive integrated circuits; and
updating the data structure storing the global power map based on the updated power state of the first server, the drives, the channels, and the integrated circuits.
2. The method of claim 1 , wherein inactive servers are placed into the power-saving mode, and wherein the data structure is updated based on a current power state of the servers associated with the rack.
3. The method of claim 1 , wherein the task is received by a task-distributing component of the first server based on a first objective function which optimizes the global power state associated with the servers and a second objective function which improves a performance parameter, including a Quality of Service, and
wherein the task-distributing component updates the data structure.
4. The method of claim 1 , wherein placing a server, drive, channel, or integrated circuit into the power-saving mode further comprises:
adjusting a voltage of a power supply to the server, drive, channel, or integrated circuit by reducing a dynamic power of the power supply and a static power of the power supply.
5. The method of claim 1 , wherein a drive includes one or more power islands, wherein a power island is associated with one or more integrated circuits, wherein the method further comprises:
placing an inactive power island into the power-saving mode by using a power switch associated with the inactive power island to shut off one or more integrated circuits associated with the inactive power island.
6. The method of claim 1 , wherein the plurality of servers is further associated with a plurality of racks.
7. The method of claim 1 ,
wherein a wake-up signal turns on a high-frequency power module,
wherein a transition period occurs between turning on the high-frequency power module and processing data by the associated integrated circuit, and
wherein the data is processed by the associated integrated circuit in a time period between the end of the transition period and turning off the high-frequency power module.
8. A computer system facilitating a fine-grained power management of a high-density computer cluster, the system comprising:
a processor; and
a memory coupled to the processor and storing instructions, which when executed by the processor cause the processor to perform a method, the method comprising:
receiving, by a first server in a plurality of servers associated with a rack, a task assigned based on a current global power state associated with the servers,
wherein the current global power state comprises a global power map stored in a data structure which indicates a current power state of the plurality of servers associated with the rack, each of a plurality of drives in a respective server, each of a plurality of channels in a respective drive, and each of a plurality of integrated circuits in a respective channel;
placing an inactive drive of the first server into a power-saving mode to obtain an updated power state of the drives of the first server;
placing an inactive channel of an active drive of the first server into the power-saving mode by using a power switch associated with each integrated circuit in the inactive channel, to obtain an updated power state of the channels of the drives;
placing an inactive integrated circuit of an active channel of an active drive of the first server into the power-saving mode by using a power switch associated with the inactive integrated circuit, to obtain an updated power state of the integrated circuits of the channels;
obtaining an updated power state of the first server based on the updated power state of the inactive drive, the inactive channel, and the inactive integrated circuits; and
updating the data structure storing the global power map based on the updated power state of the first server, the drives, the channels, and the integrated circuits.
9. The computer system of claim 8 , wherein inactive servers are placed into the power-saving mode, and wherein the data structure is updated based on a current power state of the servers associated with the rack.
10. The computer system of claim 8 , wherein the task is received by a task-distributing component of the first server based on a first objective function which optimizes the global power state associated with the servers and a second objective function which improves a performance parameter, including a Quality of Service, and
wherein the task-distributing component updates the data structure.
11. The computer system of claim 8 , wherein placing a server, drive, channel, or integrated circuit into the power-saving mode further comprises:
adjusting a voltage of a power supply to the server, drive, channel, or integrated circuit by reducing a dynamic power of the power supply and a static power of the power supply.
12. The computer system of claim 8 , wherein a drive includes one or more power islands, wherein a power island is associated with one or more integrated circuits, and wherein the method further comprises:
placing an inactive power island into the power-saving mode by using a power switch associated with the inactive power island to shut off one or more integrated circuits associated with the inactive power island.
13. The computer system of claim 8 , wherein the plurality of servers is further associated with a plurality of racks.
14. The computer system of claim 8 ,
wherein a wake-up signal turns on a high-frequency power module,
wherein a transition period occurs between turning on the high-frequency power module and processing data by the associated integrated circuit, and
wherein the data is processed by the associated integrated circuit in a time period between the end of the transition period and turning off the high-frequency power module.
15. A non-transitory computer-readable storage medium storing instructions that when executed by a computer cause the computer to perform a method, the method comprising:
receiving, by a first server in a plurality of servers associated with a rack, a task assigned based on a current global power state associated with the servers,
wherein the current global power state comprises a global power map stored in a data structure which indicates a current power state of the plurality of servers associated with the rack, each of a plurality of drives in a respective server, each of a plurality of channels in a respective drive, and each of a plurality of integrated circuits in a respective channel;
placing an inactive drive of the first server into a power-saving mode to obtain an updated power state of the drives of the first server;
placing an inactive channel of an active drive of the first server into the power-saving mode by using a power switch associated with each integrated circuit in the inactive channel, to obtain an updated power state of the channels of the drives;
placing an inactive integrated circuit of an active channel of an active drive of the first server into the power-saving mode by using a power switch associated with the inactive integrated circuit, to obtain an updated power state of the integrated circuits of the channels;
obtaining an updated power state of the first server based on the updated power state of the inactive drive, the inactive channel, and the inactive integrated circuits; and
updating the data structure storing the global power map based on the updated power state of the first server, the drives, the channels, and the integrated circuits.
16. The storage medium of claim 15 , wherein inactive servers are placed into the power-saving mode, and wherein the data structure is updated based on a current power state of the servers associated with the rack.
17. The storage medium of claim 15 , wherein the task is received by a task-distributing component of the first server based on a first objective function which optimizes the global power state associated with the servers and a second objective function which improves a performance parameter, including a Quality of Service, and
wherein the task-distributing component updates the data structure.
18. The storage medium of claim 15 , wherein placing a server, drive, channel, or integrated circuit into the power-saving mode further comprises:
adjusting a voltage of a power supply to the server, drive, channel, or integrated circuit by reducing a dynamic power of the power supply and a static power of the power supply.
19. The storage medium of claim 15 , wherein a drive includes one or more power islands, wherein a power island is associated with one or more integrated circuits, and wherein the method further comprises:
placing an inactive power island into the power-saving mode by using a power switch associated with the inactive power island to shut off one or more integrated circuits associated with the inactive power island.
20. The storage medium of claim 15 ,
wherein a wake-up signal turns on a high-frequency power module,
wherein a transition period occurs between turning on the high-frequency power module and processing data by the associated integrated circuit, and
wherein the data is processed by the associated integrated circuit in a time period between the end of the transition period and turning off the high-frequency power module.