Client/server architecture for administering a supercomputer
The invention relates in particular to an administration server (S) of a supercomputer suitable for first loading information on the environment of said supercomputer; receiving an administration task transmitted by an administration client (C1, C2); executing said administration task in collaboration with said information, previously loaded; and transmitting the results of the execution of said administration task to said administration client.
1. A method for reducing latency in performance of administration tasks for a supercomputer, the method comprising:
transmitting, by an administration client, an initial administration task to an administration server associated with environment information of a supercomputer;
subsequent to the transmitting the initial administration task, determining a subsequent administration task on the administration client;
subsequent to the determining the subsequent administration task, transmitting, by the administration client, script code of the subsequent administration task to an administration server associated with the environment information of the supercomputer having attained a power that ranges from some petaFLOPS to an exaFLOP, the environment information previously being loaded into a memory of the administration server before the transmitting the initial administration task, and the environment information of the supercomputer comprises information on a topology of the supercomputer;
directly executing the script code in collaboration with the previously loaded information on the topology of the supercomputer; and
transmitting, by the administration server, the results of the executing the script code to the administration client.
2. The method according to claim 1 , wherein the subsequent administration task is determined by programming of a computer program by a user of the administration client.
3. The method according to claim 1 , wherein the script is a script written in Python computer language.
4. The method according to claim 1 , wherein a process dedicated to the administration client executes the subsequent administration task in collaboration with the environment information.
5. The method according to claim 4 , wherein the environment information is accessible in the context of a dedicated process in accordance with a copy-on-write mechanism.
6. The method according to claim 1 , wherein the environment information is updated according to changes in an environment of the supercomputer.
7. The method according to claim 6 , wherein when the environment information is updated, the subsequent administration task in the process of being executed is not affected by the update.
8. An administration server of a supercomputer for reducing latency in performance of administration tasks for the supercomputer, the administration server configured to:
load environment information of a supercomputer having attained a power that ranges from some petaFLOPS to an exaFLOP, the environment information comprising information on a topology of the supercomputer;
subsequent to the load environment of the supercomputer, receiving an initial administration task;
subsequent to the receiving the initial administration task, receive script code of a subsequent administration task transmitted by an administration client;
directly execute the subsequent administration task in collaboration with the previously loaded environment information; and
transmit results of the execute the subsequent administration task to the administration client.
9. The administration server according to claim 8 , comprising a main process for previously loading environment information of the supercomputer and, on reception of a connection of the administration client, creating a dedicated process, the dedicated process configured to:
receive the script code of the subsequent administration task transmitted by the administration client;
execute the script code in collaboration with the environment information, previously loaded; and
transmit results of the executed script code to the administration client.
10. A system comprising:
an administration server of a supercomputer for reducing latency in performance of administration tasks for the supercomputer, the administration server configured to:
load environment information of a supercomputer having attained a power that ranges from some petaFLOPS to an exaFLOP, the environment information comprising information on a topology of the supercomputer;
subsequent to the load environment of the supercomputer, receiving an initial administration task;
subsequent to the receiving the initial administration task, receive a script code of a subsequent administration task transmitted by an administration client;
execute the script code in collaboration with the environment information of the supercomputer the environment information of the supercomputer having been previously loaded; and
transmit results of the executed script code to the administration client;
the administration server comprising:
a main process for previously loading environment information of the supercomputer and, on reception of a connection of the administration client, creating a dedicated process, the dedicated process configured to:
receive the script code transmitted by the administration client;
directly execute the script code in collaboration with the previously loaded environment information; and
transmit results of the executed script code to the administration client; and
the administration client.