Cisco Cisco MDS 9000 48-Port 8-Gbps Fibre Channel Switching Module Libro bianco

 

 
 
 
Fibre channel’s next generation has arrived and now supports 8-gigabit per 
second (Gb/s) transfer rates over one fibre channel  (8GFC) link, effectively 
doubling available bandwidth to 840 megabytes per second (MB/s).  However, 
8GFC technology is priced at a significant premium over 4GFC and depending on 
the configuration, deployment of 8GFC may be up to 2X the cost of using 4GFC. 
In contrast, one appealing factor of 8GFC links is its downward compatibility 
with 4GFC and 2GFC links thus allowing a gradual move to this new technology.   
 
Increased bandwidth capability, the large price differential, and the downward 
compatibility should be the critical factors in determining an imminent proposed 
move to 8GFC.  Aside from these, other less objective factors must also influence 
the ultimate 4GFC versus 8GFC decision.  Indeed, since 4GFC technology was 
introduced in 2004, major enhancements in storage area networking, server, 
application, and storage subsystem performance have been made, making the 
switch to 8GFC even more complicated. 

Storage area networking, particularly switch technology, has seen numerous 
changes in recent years.  Specifically, switches have necessarily become more 
complex to match increasing fabric functionality requirements.  In addition, layers 
of switches and the number of ports have been proliferating to support more 
complex fabrics. Today, core and core to edge switching uses 4GFC, 10GFC or 
inter-chassis copper channels for inter-switch links (ISLs). One area particularly 
suited to 8GFC technology are the ISLs such as between the core networking 
and/or edge storage switch components thereby reducing port counts and cabling 
by at least a factor of two.   
 
Additionally, 10GFC ISLs could conceivably be economically replaced with a 
more general purpose, 8GFC solution.  SAN extensions over metro area networks 
use 10GFC but contrary to the ISL use above, 8GFC metro area networks with FC 
clocking cost significantly more than 10GFC with Ethernet clocking and thus 
continues to justify the status quo.  

Server configurations have experienced several performance-enhancing changes 
in recent years.  One significant change has been the emergence of blade servers 
to manage multiple commodity servers running similar, high-intensity workloads. 
With blade servers, separate infrastructure hardware is now shared across a 
number of server blades within a single cabinet, thereby significantly reducing 
necessary equipment and corresponding power requirements.  Even so, the 
addition of 8GFC to these blade servers may be of questionable benefit depending 
on bandwidth requirements to meet the overall I/O demand.