Cell Relay Archive

[Venkatesh Kumar M <venkatesh.kumar@wipro.com>]

HI Davi,

	This is my understanding:
	IMA just takes the "high bandwidth stream" from the upper layers, and
	blindly distributes it among the constituent links present in the
	group.

	In your case, you have 2 E1 links, and the incoming cell stream is
	600 cps. So, IMA takes Cell 1, and puts it on Link 1, and Cell 2 and
	puts it on Link 2, and so on. This pattern will be followed until IMA
	realises that an ICP cell has to be inserted after "M" cells.

	Whenever IMA realises that there are no cells coming from the upper
layers,
	it just inserts FILLER cells, which will be dropped by the receive IMA
	layer.

	So, in effect, IMA does do load balancing.
	
	If you go a bit deep, you will realise that this "load balancing" only
	happens **over a period of time**, and not at particular instants of 
	time. This is because, the cell arrival time, and the link service time
	for a particular link can be different. 

	Consider this:
	
	L0	L1	L0	L1	L0	L1  ==> E1 service time slots
	  C0       C1      C2      C3	   C4  	    ==> Cell arrival time from
							higher layer.

	Here, when I come to service link 0, i find that there is no cell.
Hence,
	a filler is inserted in L0. But, at the time to service L1, there is 
	C0 available, and C0 goes out on L1. Similarly, in this period shown
	above, **ALL** cells move out on L1 alone, and L0 carries **ONLY**
	fillers.
	But, this scenario will reverse when I find that I have to insert an
	ICP/STUFF cell on L1, and the cell traversal shifts to Link 0.
	And, this can continue.....

	So, in effect, you cannot say that EXACTLY 50% load balancing happens,
	but, as you can see, it does happen more or less at that percentage.

	Hope this helps.
-Venki.


Davi Gupta wrote:
> 
> Distribution: world
> 
> Is IMA capable of doing load balancing? For example if I've two
> E1 links in a IMA group and pass a traffic at 256kpbs (around 600 cells
> per second) then whether on each link 300 cells per second will be flowing.
> 
> As per my understaning it will not.
> 
> Thanks in Advance,
> davi

-- 
##### ##### ##### ##### ##### ##### ##### ##### ##### ##### ##### ##### 
Visit http://www.wipro.com/wipima for the latest about our IMA solution
+++++ +++++ +++++ +++++ +++++ +++++ +++++ +++++ +++++ +++++ +++++ +++++
Venkatesh Kumar M,
Sr. Software Engineer,
Wipro Infotech, Global R & D,
Madiwala, Bangalore.
PH: (Off)572 2293/96 Extn: 5430. (Res) 349 0804
##### ##### ##### ##### ##### ##### ##### ##### ##### ##### ##### #####