Electrical computers and digital processing systems: multicomput – Computer-to-computer protocol implementing – Computer-to-computer data transfer regulating
Reexamination Certificate
1998-05-11
2001-09-11
Meky, Moustafa M. (Department: 2153)
Electrical computers and digital processing systems: multicomput
Computer-to-computer protocol implementing
Computer-to-computer data transfer regulating
C709S236000, C709S250000
Reexamination Certificate
active
06289386
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to the field of computer networking, and in particular to a fast implementation of a buffer-credit based system for high speed serial communication.
2. Description of the Related Art
Internetworking of high-performance computers has become the focus of much attention in the data communications industry. Performance improvements in processors and peripherals, along with the move to distributed architectures such as client/server, have spawned increasingly data-intensive and high-speed networking applications, such as medical imaging, multimedia, and scientific visualization. However, the interconnections between the systems and their input/output devices cannot keep up with the blinding data rates, nor can they provide the distances needed for local area networks spanning campus-wide areas.
According to “Amdahl's Law”, a megabit per second of input/output (I/O) capability is needed for every MIPS of processor performance. Current communications standards top out at just over 100 megabits per second, not nearly fast enough, as technical computing applications already demand processors exceeding 1,000 MIPS. The deficiencies in current transmission rates results in the communications channel becoming a bottleneck to system performance.
A new protocol known as Fibre Channel is 10 to 250 times faster than existing networks, transmitting at rates exceeding 1 Gbps in both directions simultaneously. It defines standard media and signaling conventions for transporting data in a serial fashion, it provides an error correcting channel code and a frame structure for transporting the data, it sets out a flow control methodology, creates some common services, and supports interfaces to existing higher level protocols such as SCSI (small computer system interface). The Fibre Channel protocol can be applied to various network topologies including point-to-point, ring, and switched. The Fibre Channel protocol is being proposed as an ANSI (American National Standards Institute, Inc.) standard, and a multitude of reference material is readily available at http://www.fibrechannel.com.
Of interest to the instant application, the Fibre Channel protocol sets out a buffer-credit flow control methodology. Under this methodology, before a first network node (initiating node) can transmit a data frame to a second network node (responding node), the initiating node must first receive a buffer-credit from the responding node, indicating that the responding node has available space in its receive buffer for a data frame. Due to the high data rates, minimum latencies are required. It is undesirable for the responding node to determine an available number of buffer credits by dividing the receive buffer capacity by the maximum frame size. Further, predetermination of the number of buffer credits is not possible due to the programmability of the maximum frame size, and assuming a worst case scenario (as is necessary for predetermination) results in extremely inefficient use of the receive buffer. A low-latency implementation of buffer-credit determination is needed.
SUMMARY OF THE INVENTION
Accordingly, there is provided herein a network interface unit which implements a low latency algorithm for buffer credit calculation. In one embodiment, a system network is provided with a remote node coupled to a local node by a serial communications link. The remote node is configured to transmit a data frame to the local node only if the remote node receives a buffer credit which indicates that the local node has available receive buffer space for a data frame. The local node includes a network interface unit for coupling to the serial communications link, and the network interface unit includes a receive buffer, a receive controller, a transmit controller, and a buffer credit manager. The receive controller stores incoming data frames in the receive buffer until they can be forwarded to the i/o bus of the local node. The transmit controller is configured to send buffer credits to the remote node in response to a credit signal from the buffer credit manager. The buffer credit manager is coupled to the receive controller to receive a receive flag for each data frame which is completely received by the receive buffer, and is also coupled to the transmit controller to receive a transmit flag for each buffer credit sent. The buffer credit manager iteratively determines the additional committed buffer space (i.e. the space committed but not yet occupied by receive data) by counting the number of receive and transmit flags, and generates the credit signal using a comparator which compares the sum of additional committed buffer space and the maximum frame size to the available receive buffer storage capacity and asserts the credit signal if the available buffer space is greater than or equal to the sum of the maximum frame size and the additional committed buffer space.
REFERENCES:
patent: 5610745 (1997-03-01), Bennett
patent: 5748613 (1998-05-01), Kilk et al.
patent: 5825748 (1998-10-01), Barkey et al.
patent: 6005849 (1999-12-01), Roach et al.
patent: 6006283 (1999-12-01), Hsieh et al.
patent: 6081847 (2000-06-01), Lin
patent: 6092167 (2000-07-01), Cheung et al.
patent: 6097698 (2000-08-01), Yang et al.
patent: 6160813 (2000-12-01), Banks et al.
patent: 6163540 (2000-12-01), Cheung et al.
Conley Rose & Tayon PC
Kivlin B. Noäl
Kupstas Tod
LSI Logic Corporation
Meky Moustafa M.
LandOfFree
Implementation of a divide algorithm for buffer credit... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Implementation of a divide algorithm for buffer credit..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Implementation of a divide algorithm for buffer credit... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2469544