Electrical computers and digital data processing systems: input/ – Intrasystem connection – Bus access regulation
Reexamination Certificate
1999-08-11
2001-06-26
Auve, Glenn A. (Department: 2181)
Electrical computers and digital data processing systems: input/
Intrasystem connection
Bus access regulation
C709S212000
Reexamination Certificate
active
06253271
ABSTRACT:
BACKGROUND OF THE INVENTION
The invention relates to computer networks, and more particularly, to communications between data storage devices and clients in computer networks.
The growth of client/server based computing, where applications and other resources are distributed among a network of computers having one or more servers and clients, is driven by a need for fast, reliable and secure access to vast amounts of shared data in organizations. As the number of clients serviced by the server grows, along with increased client data processing capability made possible by advances in microprocessor technology, conventional servers can be overloaded with requests from clients.
In response, a number of architectures have been developed to address server overloads, including distributed processor architecture, symmetric multi-processor architecture, parallel processor architecture and hypercube architecture. However, some of these architectures still direct file requests coming over the network to a host processor, which in turn makes appropriate requests to an input/output (I/O) processor controlling data storage devices such as disks in satisfaction of the request from a client. Data from disks is buffered in a memory section before being transmitted to the requester under the control of the host processor. Such micro-management by the host processor can result in delays and can be wasteful of the host processor's capabilities. Additionally, the host processor, if overloaded, may itself become a bottleneck in the transfer of data between the disk and the requesting client. Further, when the processor is involved in the data processing loop, host bus bandwidth may be reduced, as data from disks is typically transferred over a bus and stored in a host memory before being sent to the requesting client over the bus, thus reducing available bus bandwidth.
To minimize this problem, certain architectures shift the I/O load to dedicated sub-systems that handle disk requests. However, even such sub-systems deploy local processors to handle requests arriving over the network and to supervise disk controllers in transferring data. These processors still are intimately involved with the file transfer process. Further, the bus contention problem still exists with respect to local buses supporting the local processor. Thus, the potential bottleneck still exists whenever the local processor or the host processor is directly involved with the disk I/O.
As intimate involvements of the processor in disk I/O handling are inefficient and uneconomical, what is needed is an apparatus in a file server that can directly transfer data from data storage devices to the client over the network with minimal involvement by any processor. Moreover, what is needed is an apparatus which minimizes bus contention and provides high data availability and fast access to shared data with minimal overheads.
SUMMARY OF THE INVENTION
The present invention relates to a bridge in a file server which provides a direct link to data storage devices in satisfaction of data requests. The file server has one or more function-specific processors, including network processors (NPs) and file storage processors (FSPs), all operating in parallel and communicating over an interconnect bus. To minimize local bus contention between data storage devices and network communications, the FSP provides separate internal buses. One internal bus connects to the interconnect bus for network communications. The other internal bus connects to disk controllers for performing I/O operations on data storage devices connected thereto. The bridge provides a path between the FSP's internal buses so-that, for disk access requests, data from a particular data storage device may be sent by the disk controller via the bridge over the interconnect bus to the NP servicing the request with minimal involvement of a local FSP processor.
During operation, client requests are received and analyzed by NPs, and if acceptable, relayed to one of the FSPs which manages a virtual file system of mass storage devices connected to the FSP. The local FSP processor determines the location of a buffer on the requesting NP for storing data to be transmitted and instructs the disk controller so that data retrieved by data storage devices is directly deposited into the buffer on the NP over the interconnect bus via the bridge. After issuing the instruction to the disk controller, the local FSP processor is finished and thus incurs only a minimal overhead for each I/O request.
Upon receipt of the instruction, the disk controller causes data storage devices to retrieve the requested data and sends the result directly to the buffer of the requesting NP via the bridge. The requesting NP in turn packetizes the result and sends the packets to the requesting client.
Advantages of the system includes the following. In place of conventional overheads associated with local FSP processor/disk controller management scheme where data from disk is sent to the local FSP's RAM for eventual transmission to the requesting NP, the bridge provides a direct path between the client and file storage processors which eliminates unnecessary intermediate data routing. As the processor in the FSP need not be involved, the bridge avoids contention on internal buses of the FSP for data transmissions. Further, since the processor is no longer intimately involved with disk transfer operations, the bridge minimizes the local FSP processor overhead. The decoupling of the data storage devices from the local FSP processor enhances system performance. The bridge is also flexible in that it can be bypassed for transactions that require local FSP processor handling.
The resulting server with the bridge is powerful, scalable and reliable enough to allow users to consolidate their data onto one high performance system instead of scores of smaller, less reliable systems. This consolidation of data resources onto a powerful server brings a number of advantages to the client-server environment, including increased ease of administration and quicker and more reliable data access than conventional client-server architectures.
Other features and advantages will be apparent from the following description and the claims.
REFERENCES:
patent: 5163131 (1992-11-01), Row et al.
patent: 5355453 (1994-10-01), Row et al.
patent: 5541853 (1996-07-01), Croft
patent: 5634099 (1997-05-01), Andrews et al.
patent: 5787459 (1998-07-01), Stallmo et al.
patent: 5790816 (1998-08-01), Fok et al.
patent: 5802366 (1998-09-01), Row et al.
patent: 5819292 (1998-10-01), Hitz et al.
patent: 5931918 (1999-08-01), Row et al.
patent: 5941969 (1999-08-01), Ram et al.
Auspex Systems, “The NS 7000 NetServer Family Pocket Guide”, Doc. No. 300-MC063.
Auspex Systems, “NS 7000/250 Series NetServers”, Product Information, Doc.No. 300-DS046.
Auspex Systems, “NS 7000/250 Series NetServers”, Product Information, Doc.No. 300-DS047.
Abraham Sara
Gajjar Kumar
Popelka Paul Lester
Ram Tamir
Syu Syang Edward
Auspex Systems, Inc.
Auve Glenn A.
Fish & Richardson P.C.
LandOfFree
Bridge for direct data storage device access does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Bridge for direct data storage device access, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Bridge for direct data storage device access will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2459691