High-availability computer system, working method and the use thereof
AssigneeFujitsu Limited
InventorBernd Winkelsträter
Life of this patent
Estimating the term…
Abstract
A high-availability computer system includes a plurality of computer nodes for parallel data processing, wherein each computer node has at least one data processing component and a main memory. The computer system furthermore includes a plurality of storage nodes, wherein each storage node has at least one non-volatile mass storage device. The plurality of computer nodes are partitioned into a plurality of computer groups, wherein each of the plurality of computer groups has at least one redundant computer node. Each computer group has a local connection structure allocated to the respective computer group, the structure connecting the storage nodes allocated to the respective computer group to the computer nodes of the respective computer group.
Claims17
A high-availability computer system comprising:
- a first number (n) of computer nodes for parallel data processing of a second number (s) of data segments (d) of data (D) to be processed, wherein the second number (s) being greater than or equal to two, and each computer node has at least one data processing component and a main memory;
- a plurality of storage nodes, wherein each storage node has at least one non-volatile mass storage device:
- and
- at least one data network that accesses a plurality of computer nodes, wherein
- the plurality of computer nodes is partitioned into a third number (m) of computer groups, wherein the third number (m) is greater than or equal to two, and the second number (s) is less than or equal to the difference (n−m) between the first number (n) and the third number (m);
- each of the plurality of computer groups has at least one redundant computer node;
- each storage node of the plurality of storage nodes is allocated to precisely one computer group of the plurality of computer groups;
- and
- each computer group has a local connection structure allocated to the respective computer group, said structure connecting the storage nodes allocated to the respective computer group to computer nodes of the respective computer group.
A working method for a high-availability computer system that includes
- a first number (n) of computer nodes for parallel data processing, wherein each computer node has at least one data processing component and a main memory;
- a plurality of storage nodes, wherein each storage node has at least one non-volatile mass storage device:
- and at least one data network that accesses the plurality of computer nodes, wherein
- the plurality of computer nodes is partitioned into a plurality of computer groups;
- each of the plurality of computer groups has at least one redundant computer node;
- each storage node of the plurality of storage nodes is allocated to precisely one computer group of the plurality of computer groups;
- and
- each computer group has a local connection structure allocated to the respective computer group, said structure connecting the storage nodes allocated to the respective computer group to computer nodes of the respective computer group;
- the method comprising:
- dividing data (D) to be processed in parallel into a second number (s) of data segments (d), and partitioning the first number (n) of computer nodes into a third number (m) of computer groups, wherein the second number (s) is less than or equal to the difference (n−m) between the first number (n) and the third number (m) and the third number (m) is greater than or equal to two.
Dependencies are inferred from the claim text.