Could you please, in detail, explain the relationship between a hub and a Fibre Channel arbitrated loop.
A hub IS an arbitrated loop. The internal wiring of a hub has all the ports connected as a physical loop. A hub uses the "Fibre Channel arbitrated loop" protocol that defines how the devices connected to the loop arbitrate for access to loop resources and gain control of the loop to communicate with other devices connected to the hub.
As you add a device to the hub, a process called a LIP (loop initialization primitive) function occurs. To keep the story short, the end result is the new device added to the loop looks for a "loop master" on the hub and the device ends up getting a loop ID, called an ALPA address. Each device on the hub gets an ID and each device then arbitrates for access to the loop using its assigned loop ID as its priority. There is a whole bunch of other stuff that happens under the FC-AL protocol to make sure everyone gets a shot at using the loop.
Once the loop is initialized, devices can begin to compete for access to the loop by arbitrating for the resource. A device requesting the loop will transmit an ARB (arbitrate) primitive signal with its ALPA in the operand field. After the device is finished with its "exchange on the loop, it must relinquish control so another device can have a chance to use the loop.
Be that as it may since each device must arbitrate for access. That means only a single device has control of the loop inside the hub at any point in time. This means that access to loop bandwidth is shared between the other devices on the hub. Using 1G-bit Fibre Channel, this is 100MB per second of shared bandwidth. This is why switches are more common today than hubs. By ysing a 1G-bit Fibre Channel switch, everyone gets its own 100MB per second path.
Other hub stuff:
Everyone shares the 100 Megabyte bandwidth
Supports 126 active nodes (NL ports= node to loop ports)
Is a very cost effective solution since hubs are cheaper than switches
One issue when using hubs is SCSI bus resets. If you have a tape drive connected to the hub along with your host nodes and disk drives, every time a tape rewind happens, there is a SCSI bus reset that is seen by all the devices. This can cause weird stuff if you are trying to access your disks at the same time. Always use a separate hub for disks. Keep tapes on their own hub by adding another HBA to your servers. Use one hub-based loop for disk access and the other for tape access.
The other issue is that if you add another node to the hub while there is traffic, another LIP occurs which will not be a good thing for what is currently being transmitted on the loop. You can pull devices from the hub though as hub ports are "self healing", and removing a device will not break the loop.
Other than that, hubs are fairly "plug and play."
Editor's note: Do you agree with this expert's response? If you have more to share, post it in one of our .bphAaR2qhqA^0@/searchstorage>discussion forums.
This was first published in November 2002