home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Danny Amor's Online Library
/
Danny Amor's Online Library - Volume 1.iso
/
bbs
/
society
/
society.lha
/
PUB
/
isoc_news
/
1-2
/
n-1-2-040.92a
< prev
next >
Wrap
Text File
|
1995-07-21
|
2KB
|
33 lines
040.92: EARN Regionalization
by Frode Greisen <NEUFRODE@vm.uni-c.dk>
The way EARN (and BITNET) networks function is by a store-and-
forward system with permanent connections and fixed routing
tables in the nodes, which are hosts running the IBM NJE
protocol. Originally, the routing was a pure tree structure.
About 18 months ago, EARN decided on a regionalization plan by
which 13 core nodes were connected essentially all to all. This
was done by setting up virtual NJE over IP links using available
IP lines for research and education in Europe and across the
Atlantic. A similar regionalization took place in BITNET in the
so-called BITNET-II project somewhat earlier.
When traffic conditions are good (i.e., when there are no link
failures or traffic overload on a particular path), then mail and
small files typically have end-to-end transfer times of a few
seconds for international traffic involving half-a-dozen hops.
When traffic conditions are bad, then files are queued at the
intermediate hosts and the transfer can take minutes or longer.
For the last year or so, EARN staff has monitored queue sizes at
the international nodes. It has been very encouraging to note
that from February 1991 to February 1992 the average number of
files queued has been reduced by 67% while there was a slight
increase in total traffic.
The number of files queued is only an indirect service measure
but there is no doubt that the end users have experienced a
significant improvement in transmission times and we largely
assign this improvement to the regionalization described above.