Thursday, July 1, 2010

Cisco Live 2010 - Tuesday and Wednesday

I thought I'd be able to keep up with this on a daily basis, but I was wrong. So here's a recap of the bulk of the conference. I'm sitting in a Virtualization Case Study on Thursday while writing this.

Picking up where I left off in my previous post, I attended the LISP session. It was a good refresher, as I had not followed up on the technology since my first introduction at Cisco Live 2009. I'll save my thoughts on the technology for a later blog post.

On Tuesday I attended an Overlay Transport Virtualization Breakout Session. It was a good compliment to my OTV Lab on Sunday. I felt that the presenter glossed over the significant architectural change that introducing Data Center Interconnect (DCI) brings to a network. He (rightly) stated that configuring OTV is fairly trivial, assuming you have the appropriate NX-OS and a pair of Nexus 7000 switches. Once you enable L2 adjacency between disparate DCs, there's no going back. On the plus side, if you need DCI, OTV looks much better than the alternatives (dot1q dedicated links, carrier VPLS, internal VPLS, EoMPLSoGRE, L2TPv3... I'm sure I am missing a couple).

The John Chambers keynote was good as usual. The introduction of the Cisco Cius was probably the highlight. I haven't investigate it yet, and probably won't get around to looking into it in detail. I have to confess that I'm not terribly interested in technology. I don't personally own any Apple products and I've only carried a Blackberry smart phone for three years. Probably the 'coolest' technology that I own is an original Kindle, which meets my nontechnical reading needs. For that matter, I only have a single Cisco router (and no switches) at my home. Somehow I manage to keep up, so don't let anyone convince you that you need racks full of equipment to participate in our industry.

My one complaint is that I always feel like my organization is 5 - 10 years behind when I leave the keynote. For example, the Cisco Live 2008 keynote was centered around complete virtualization. Two years later, we're nowhere near there. I know the point of the keynote is to show the future direction of Cisco, and I wouldn't want anything to change, but I don't like the feeling of being behind. I always leave the session with a sense of urgency to get us back on track. This year I brought my manager, so hopefully he'll feel the same way and we can get things moving a bit faster.

After lunch on Tuesday I attended a second Panel on Experiences with Deploying IPv6. I am finally convinced that my employer needs to begin the transition. This was a great session. Presenters from Comcast, Google, Microsoft and Tata described their IPv6 deployments and gave recommendations. For once, Apple took a bit of a beating (and not from the Microsoft presenter) because no Apple OS supports DHCPv6. This requires the use of SLAAC for acquiring IPv6 addresses in mixed PC environments. It's not a show stopper, but flexibility would be nice. I prefer SLAAC anyway for the simplicity, but I am firmly in the minority, based on a show of hands during the session.

On Tuesday evening I spent a few hours with John Chambers and my ~75 closest CCIE/CCDE friends. This was my first year attending the CCIE/DE NetVet reception. It was quite an experience. John Chambers took questions for about 1.5 hours, and gave very direct responses to some difficult questions. I received some great information that will help me position my organization for upcoming shifts in Cisco strategy. While I'm fairly confident this was information I could acquire elsewhere, when it comes from the CEO, it is reliable.

After the NetVet reception I spent an hour or so attending the CCIE party at the Voodoo Lounge. I didn't stay long, but I did get a picture of the cool (both literally and figuratively) CCIE logo ice sculpture. I would have preferred the original laurel leaf router logo, but I was still impressed.


On Wednesday morning I attended "IS-IS Network Design and Deployment". I was looking forward to this session more than any other, as I have had great difficulty getting this content from other sources. There are a few books that cover IS-IS, but Cisco Live Virtual has never had the audio for this session. The slide deck I found in the past (and the one used during this presentation) lacks sufficient detail to reconstruct the information. The presenters did not disappoint, and I took about four pages of notes (the only notes I've taken at Cisco Live 2010). I intend to write a separate blog post on why IS-IS can no longer be ignored by Enterprise-focused networkers. If this describes you, you'll have a year or two to prepare.

After lunch I attended a session called "Designing Multipoint WAN QoS". This session didn't stand out in the session list when I was registering, but I was tipped off by a friend (the presenter) that it would be a good session. He didn't disappoint. I am almost embarrassed to say that I had never given much thought to the problem this session attempts to solve. It certainly deserves an individual post, but I'll describe it in a few sentences here. The best example of the issue is the Teleworker problem, where the remote user has a small Cisco router with a VPN tunnel to a Corporate headend router. The user also uses the same ISP connection (DSL, cable, etc) for home Internet usage. The problem is how do you ensure that the ISP connection prioritizes real-time Corporate traffic over general Internet web traffic downstream (towards the remote user)? To put it more generally, how do you get outbound QoS from a Service Provider network that is not QoS-aware?

My last session on Wednesday was Enterprise IPv6 Deployment. I've seen this content before, and originally I had scheduled a DCI session (DCI w/ Advanced-VPLS), but I thought it would be best to reinforce the IPv6 content. The presenter (Shannon McFarland) is very good, so I was not bored by seeing some of the information for a second time. His recommendation to begin in the core and work out towards the edge is somewhat unconventional, but he defends it well, to the point that I was convinced. Maybe more contemplation will change my mind. More information to come in the future. The biggest point he wanted to get across was "Dual Stack Where Possible, Tunnel When You Must." Tunnels introduce MTU issues and create difficult-to-troubleshoot network architectures. Maybe he should talk to the TRILL, LISP and OTV groups! :)