Electrical computers and digital processing systems: multicomput – Computer-to-computer protocol implementing – Computer-to-computer data transfer regulating
Reexamination Certificate
2000-08-04
2003-12-02
Jean, Frantz (Department: 2155)
Electrical computers and digital processing systems: multicomput
Computer-to-computer protocol implementing
Computer-to-computer data transfer regulating
C709S239000, C710S038000, C710S316000
Reexamination Certificate
active
06658478
ABSTRACT:
TECHNICAL FIELD OF THE INVENTION
The present invention relates generally to the field of data storage and, more particularly, to a data storage system.
BACKGROUND OF THE INVENTION
In the context of computer systems, enterprise storage architectures provide mass electronic storage of large amounts of data and information. The frenetic pace of technological advances in computing and networking infrastructure—combined with the rapid, large-scale sociological changes in the way the way these technologies are used—has driven the transformation of enterprise storage architectures faster than perhaps any other aspect of computer systems. This has resulted in a variety of different storage architectures, such as, for example, direct attached JBODs (Just a Bunch Of Disks), SAN (Storage Area Network) attached JBODs, host adapter RAID (Redundant Array of Inexpensive/Independent Disks) controllers, external RAID controllers, redundant external RAID controllers, and NAS (Network Attached Storage) Each of these storage architectures may serve a special niche, and thus may differ significantly in terms of functionality, performance, cost, availability, scalability and manageability.
Typically, any given business has a variety of data storage needs, such as, for example, database storage, home directories, shared application executables, and data warehouse storage. In general, no single one of the previously developed architectures is capable of addressing all of the storage needs of a business. Thus, businesses are forced to use a number of different architectures to provide the functionality and performance which are desired. This results in fragmented data storage which limits the sharing of data resources, erects static boundaries in data, necessitates redundant training for staff, and requires additional management resources. For example, excess storage space in one type of architecture generally cannot be used to ease congestion in an architecture of another type. Nor can storage architectures of different types be used as backup/redundancy for each other.
Previously developed data storage architectures suffer in other respects as well. For example, data storage architectures typically use computer-memory complexes (e.g., central processing unit (CPU) and associated memory) to control access into the devices which actually store data (e.g., disk drives). In previously developed architectures, all data transfers are routed through the internal buses of the computer-memory complexes. Because these internal buses generally have relatively low bandwidth, bulk data transfers significantly slow the operation of the computer-memory complexes which, in turn, negatively impacts the performance of the overall architectures.
Enterprise architectures may utilize disk storage systems to provide relatively inexpensive, non-volatile storage. Disk storage systems have a number of problems. These problems include the following. Disk systems are prone to failure due to their mechanical nature and the inherent wear-and-tear associated with operation. Accesses (i.e., reads and writes) into disk systems are relatively slow, again due to their mechanical nature. Furthermore, disk storage systems have relatively low bandwidth for data transfer because the effective bandwidth is limited by “platter speed” (i.e., the rate at which data bits move under a disk head).
Various efforts have been made to reduce the problems associated with disk storage systems. One exemplary system resulting from such efforts employs a “node” to control the access of data/information into a number of disk drives. In such previously developed system, the node stores redundant data (e.g., parity information or a duplicate copy of the data itself) to multiple disk drives so that if one disk drive fails, the redundant data can be used to reconstruct the data. The node includes a main computer system having system memory into which data can be cached to reduce the slow seek time of associated with disk drives. Furthermore, the node may store data across multiple disk drives in a technique known as “striping” so that the effective data storage bandwidth is the aggregate bandwidth of the individual disk drives. In addition, multiple nodes may be used within a system to provide redundancy.
Nonetheless, the previously developed system utilizing a storage node suffers from its own problems. The data storage bandwidth through the node is still relatively narrow due to limitations of the main computer system. The memory for caching data at a node is typically volatile, and hence, data may be lost if the node fails. Furthermore, the node can be a single point of failure for the system—i.e. if the node fails, all of the data on disk drives connected to the node is unavailable. Even if multiple nodes are provided, communication between nodes is typically slow, and thus performance of the system is less than optimal.
SUMMARY OF THE INVENTION
The disadvantages and problems associated with previously developed storage systems and techniques have been substantially reduced or eliminated using the present invention.
Among other things, the present invention provides a high performance, scalable, flexible, cost-effective storage system architecture which is particularly well suited for communication-intensive, highly-available data storage, processing or routing. This architecture is capable of addressing the entire range of a business's storage needs. It is scalable both in storage capacity and performance, including latency, bandwidth, and performance stability in the event of localized congestion or failures. The architecture incorporates redundancy in every component, thus making it highly reliable.
According to an embodiment of the present invention, a data storage system includes a plurality of nodes for providing access to a data storage facility. Each node has a computer-memory complex to provide general purpose computing for the node, a node controller to control data transfers through the respective node, and a cluster memory to buffer data for the data transfers. A plurality of communication paths interconnect the nodes, with a separate communication path provided for each two nodes of the data storage system.
According to another embodiment of the present invention, a data storage system includes a plurality of system boards for providing access to a data storage facility. Each system board has an interface slot to connect the system board to the data storage facility, a computer-memory complex to provide general purpose computing for the system board, a node controller to control data transfers through the system board, and a cache memory to buffer data for the data transfers. A backplane interconnects the system boards and supports a plurality of communication paths for transfer of data between the system boards.
A data storage system in accordance with an embodiment of the present invention includes multiple nodes (e.g., up to eight in one implementation). These nodes provide connections for transferring data and information between and among a number of host devices (e.g., servers) and storage devices (e.g., disk drives). Each node is connected to every other node by a number of communication paths, each of which can be a high-speed link. Each node may include a node controller, a cluster memory, and a computer-memory complex. A technical advantage of the present invention includes providing, at each node, a node controller and cluster memory which are separate from the computer-memory complex. A central processing unit (CPU) in the computer-memory complex performs the control functions, setting up the various addresses and lengths required for the data transfer. The actual transfer of data blocks, however, does not go through the computer-memory complex, but rather through the node controller to/from the cluster memory. Since the amount of data in the control is much smaller that the amount of data in the data blocks, the computer-memory complex is relieved of the burden of most of the data bandwidth. With cluster memory, data/info
Broniarczyk David J.
Cameron George C.
Price Jeffrey A.
Singhal Ashok
3PARdata, Inc.
Hsia David C.
Jean Frantz
Patent Law Group LLP
LandOfFree
Data storage system does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Data storage system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Data storage system will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3174383