09-07-2011 02:55 AM
I'm not familiar with brocade equipment but I'm trying to answer a question being asked of me, if anyone could look at the below and help me I would appreciate it
this is the email I have received
09-07-2011 04:34 AM
With Brocade infrastructure already in place, I would strongly recommend not to mix Brocade and Cisco in the same environment. It's not worth it.
In essense, they offer you a 5-years old platform with (72) 4G badly oversubscribed ports that would require 3000W to power and 7 rack units to mount. What you have today is a modern platform with (80) 8G non-oversubscribed ports that require 300W (yes, that's ten times less) and 2 rack units to mount.
Cisco's line cards are up to 4:1 oversubscribed (although Cisco will sell you 'host-optimized' and 'storage-optimized' fairytale which illustrates why LAN guys shouldn't build SAN switches). Brocade 5300 has no oversubscription (meaning all 80 ports can run 8Gbps full speed).
This equipment is definitely NOT compatible with your Brocade SAN. While Cisco does offer 'compatibility' mode to interoperate with Brocade equipment, this is not supported in the latest Brocade Fabric OS versions and has limited functionality. Don't go for it unless absolutely unavoidable.
In summary, you don't need that MDS because you're already running a modern enterprise-class 8G platform which is way more efficient than MDS9506 proposed.
If SAN needs to be expanded, I'd rather ask that guy to offer something similar from Brocade. Depending on your requirements, another 5300 or perhaps a DCX-4S would fit.
Hope this helps,
09-08-2011 08:41 AM
Oversubscrition of 4:1, in terms of bandwith, means for every 4 userports there is 1 backend port to route traffic to other userports.
Of a 4G platform this results in 16G of traffic havine only 4G off bandwith to reach other ports.
Depending on the architecture Cisco uses it may be possible the platform understand blade and asic locality.
09-08-2011 10:04 AM
Looking at the line cards being proposed:
24-port card has 6 ports per group, 48-port card has 12 ports per group. Each group has 12.8 Gbps bandwidth. Thus in fully populated 48-port card every twelve 4G ports share 12.8 Gbps instead of full 48 Gbps one would expect. This is rougly 48:12 or 4:1 oversubscription.
Again, Cisco will tell you the story about "dedicating bandwidth to ISLs and storage ports, running host ports in shared-bandwidth" which is now a textbook example of how to present poor technology limitations as "competitive advantage".
Contrary to the post above, these cards do not utilize blade or ASIC locality -- all traffic is routed through Supervisor-2 cards.
Not to mention that these cards were introduced in 2006. . . Does it make sense to invest in five-years-old technology?
Hope this helps,