Hey everyone!
I am setting a standard for a QOS design for all our sites. Before I started, we had a vendor come in and do it for us, and while its a good design I think the table mapping portion has me a bit confused. From the best of my knowledge it allows you to change the values from 0 to 8 (IP Precedence to DSCP values) to, I believe, change the way the data is prioritized (Unfortunately I've looked for details on the table mapping for QOS and Cisco's white papers are meh). We only do QOS at the layer 2 portion of our network, so the core doesn't even do anything in terms of IP Precedence or DSCP (Cores are 6800s). Can someone elaborate on this?
Also, it got me thinking to a conversation I once had with my architect (2x CCIE, CCDE) who did our entire data center design at my last place and he made an interesting point when I asked him why we don't run QOS exclusively. His reasoning was that physical links and switches are getting better at handling larger bandwidth (10G, 40G, etc.), so his solution was to not do QOS but to add more port density (example: 9Ks handle 40Gs, which will allow for a bigger backbone and the 9Ks are capable of supporting it in terms of the Data/Control Planes). What are your guys' thoughts on this? I know that most campus networks (like ours) usually have 10GB ports for basic Catalyst switches, so I can understand how that can apply to a data center with upwards of 100 Gig backbones (very unlikely to ever have that in a campus network at a cost efficient rate right now). Let me know I am willing to hear some opinions on this.
No comments:
Post a Comment