Fiber channel arbitrated loop dynamic loop sizing

Details Fiber channel arbitrated loop dynamic loop sizing Fiber channel arbitrated loop dynamic loop sizing

1997-03-31

2003-01-07

Nguyen, Chau (Department: 2663)

Multiplex communications

Fault recovery

Bypass an inoperative switch or inoperative element of a...

C370S223000, C714S004110

Reexamination Certificate

active

06504817

ABSTRACT:

FIELD OF THE INVENTION
This invention relates in general to redundant systems and data communication architectures and, more particularly, to performance features of a fibre channel arbitrated loop in a redundant system.
BACKGROUND OF THE INVENTION
Fibre Channel (FC) is a high speed serial communications architecture for transmitting data from one communication port to a next (i.e., a point to point communication link). Data transfer rates on the order of 100 mega bytes per second (or 1 giga bits per second) are achievable. Devices, such as hard disks and controllers for Redundant Arrays of Inexpensive Disks (RAID controllers), have one or more communication ports that are attached to the FC links. Each port has an input and an output. The ports can be connected in a daisy chain configuration to form a loop over which communication transmissions are passed. Current standards support up to 126 devices on the loop. In order to maintain order on the loop, a loop protocol, which includes an arbitration protocol, is used to determine which port (device) has control of the loop. This configuration and loop protocol are known as a Fibre Channel Arbitrated Loop (FC-AL).
In FC-AL each port examines the incoming data (as the data is passed around the loop) to determine if any action is appropriate to be taken. Each port simply re-transmits the data if that port is not involved in the current activity. This data “checking” or “snooping” requires time and causes a significant amount of delay to be added between the input and output of the data at the port. These “port” delays in connection with “propagation” delays through the port-to-port communication links add up to a round trip loop delay. The round trip loop delay is the time for data to travel from a port's output around the loop and back to its input. Round trip loop delay is typically dominated by port delays.
To gain control of the loop a port transmits a special packet of information known as an arbitration primitive. If the arbitration primitive can make a complete trip around the loop without a higher priority node substituting its own arbitration primitive, the node wins control of the loop. In other words, if a port receives back the arbitration primitive it sent, it has won arbitration. In addition to gaining control of the loop, there is another round trip of overhead incurred for the port that won arbitration to make contact with the port with which it desires to communicate.
To protect loops from having a failing port (or device) cause a break in the loop, a routing device known as a port bypass circuit (PBC) is used. A PBC is a device that duplicates its received input signal. One copy is sent to a multiplexer inside the PBC and the other is driven on the “local” output. The “local” output of the PBC is connected to the input of a respective port. The output of the port is connected to the “local” input of the PBC which is also connected to the PBC's internal multiplexer. The output of the multiplexer is driven to the PBC's output which is connected to the next PBC on the arbitrated loop.
When a node (device) is active and functional, the multiplexer in the PBC selects the local input from the port to drive the output to the rest of the loop. When the node is dysfunctional or absent, the multiplexer will select the input directly from the input connected to the previous PBC. However, the data stream coming into the PBC is always provided to the port so that it can monitor the loop traffic for messages that may affect it. This is important since in some architectures the port has control over the multiplexer in its associated PBC. When bypassed, however, the port does not drive data onto the loop.
The delay through a port is more than 50 times the delay through a PBC in bypass mode. The port delay is also significantly larger than the delay associated with the interconnecting media between ports in typical array applications. Thus, the more operational devices that are connected to a given FC-AL, the longer it takes for arbitration to occur and the higher the overhead penalty. Disadvantageously, the increased overhead reduces the available bandwidth.
Accordingly, an object of the present invention is to improve performance of a Fibre Channel Arbitrated Loop in a redundant system.
SUMMARY OF THE INVENTION
According to principles of the present invention in a preferred embodiment, in a Fibre Channel Arbitrated Loop communications architecture, dynamic loop sizing includes selectively bypassing operational device ports in the loop in order to minimize overhead associated with loop size. In redundant systems with multiple loops, the system is optimized by judicious distribution of the bypassed ports among the loops. Redundancy is at no point reduced. By bypassing unused or unneeded ports off of a loop, the round trip delay of the loop is appreciably shortened. By minimizing the round trip delay, arbitration overhead and access latency is reduced, and loop bandwidth and overall performance is improved.
According to further principles, dynamic load sharing balances the load between dual loops when using dual ported devices on the loops. Dynamic load sharing is accomplished by bypassing a given subset of devices off each loop to reduce round trip delay; monitoring traffic on the loops; and controlling which devices are attached to which loops in order to balance the load across the loops.
According to yet further principles, dynamic system configuration features provide an optimized system on a discretely continuous basis, including implementing dynamic loop sizing, dynamic load sharing, and managing the interconnection of multiple loops and controllers.
Other objects, advantages, and capabilities of the present invention will become more apparent as the description proceeds.


REFERENCES:
patent: 4393497 (1983-07-01), Cantwell, Jr.
patent: 5084863 (1992-01-01), Guezou et al.
patent: 5729692 (1998-03-01), Qiu et al.
patent: 5732211 (1998-03-01), Efron et al.
patent: 5754549 (1998-05-01), DeFoster et al.
patent: 5757642 (1998-05-01), Jones
patent: 5815489 (1998-09-01), Takatori et al.
patent: 5892924 (1999-04-01), Lyon et al.
patent: 5922077 (1999-07-01), Espy et al.
patent: 5923449 (1999-07-01), Doerr et al.
patent: 0751464 (1997-01-01), None
patent: WO92/04787 (1992-03-01), None
patent: WO97/04384 (1997-02-01), None
patent: WO97/07458 (1997-02-01), None
Search Report
Robert W. Kembel; In-Depth Fibre Channel Arbitrated Loop; 1996/1997, XP-002055479; p. 42-43 and p. 269-282.

Affiliated with

Also associated with

Rating

Fiber channel arbitrated loop dynamic loop sizing does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Fiber channel arbitrated loop dynamic loop sizing, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Fiber channel arbitrated loop dynamic loop sizing will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3065744