Category: design

The “bring Amazon Web Services mentality back home” blog post generated the expected comments, from “developers have no clue about networking or network services” to “we went through the whole thing and failed badly.”

Well, even though it might have seemed so, I didn’t advocate letting the developers go unchecked, I was just pointing out that double standards make no sense.

In the first half hour of the Infrastructure for Private Clouds workshop at last week’s Interop Las Vegas I focused on business aspects of private cloud design: defining the customers, the services, and the level of self-service you’ll offer to your customers.

Nick Martin published a great summary of these topics @ SearchServerVirtualization; I couldn’t have done it better myself (they want to get your email address, but this article is definitely worth it).

During one of the ExpertExpress engagements I helped a company implement the BGP Everywhere concept, significantly simplifying their routing by replacing unstable route redistribution between BGP and IGP with a single BGP domain running across MPLS/VPN and DMVPN networks.

They had a pretty simple core site network, so we decided to establish an IBGP session between DMVPH hub router and MPLS/VPN CE router (managed by the SP).

Once you decide to use BGP as the routing protocol in your DMVPN network, you face a few more design choices:

• Should you use IBGP or EBGP?
• Should you use a unique AS number for every DMVPN site, or the same AS number on all spoke sites?

The BGP Routing in DMVPN Access Networks ExpertExpress case study describes these dilemmas in more details.

Jason Edelman wrote a great blog post after watching Ethan Banks struggle with yet another multi-vendor IPsec deployment. Some of his ideas make perfect sense (wiki-like web site documenting working configurations between vendor X and Y for every possible X and Y), others less so (tunnel broker – particularly in view of recent Tor challenges), but let’s step back a bit and ask ourselves “Why is IPsec so complex?”

One of the Expert Express sessions focused on an MPLS/VPN-based WAN network using OSPF as the routing protocol. The customer wanted to add DMVPN-based backup links and planned to retain OSPF as the routing protocol. Not surprisingly, the initial design had all sorts of unexpectedly complex kludges (see the case study for more details).

Having a really smart engineer on the other end of the WebEx call, I had to ask a single question: “Why don’t you use BGP everywhere” and after a short pause got back the expected reply “wow… now it all makes sense.”

Microseconds after VXLAN was launched at VMworld 2011, someone started promoting it as a data center extension solution. Even though layer-2 DCI doesn’t make much sense (even to server people) and VXLAN is really not a DCI solution, the lure of misusing a technology was irresistible.

VMware has a fantastic-looking cloud provisioning tool – vCloud director. It allows cloud tenants to deploy their VMs and create new virtual networks with a click of a mouse (the underlying network has to provide a range of VLANs, or you could use VXLAN or vCDNI to implement the virtual segments).

Needless to say, when engineers not familiar with the networking intricacies create point-and-click application stacks without firewalls and load balancers, you get some interesting designs.

One of the answers you get from some of the vendors selling you data center fabrics is “you can use any topology you wish” and then they start to rattle off an impressive list of buzzword-bingo-winning terms like full mesh, hypercube and Clos fabric. While full mesh sounds like a great idea (after all, what could possibly go wrong if every switch can talk directly to any other switch), it’s actually the worst possible architecture (apart from the fully randomized Monkey Design).

Last week Juergen Brendel published an interesting blog post describing how you can use vCider to implement high-availability clusters with multi cloud strategy, triggering the following response from one of my readers: “I hadn't heard of vCider before but seeing stuff like this always makes me doubt my sanity – is there really a situation where the only solution is multi-site L2?”

We’ve seen several interesting data center fabric solutions during the Networking Tech Field Day presentations, every time hearing how the new fabric technologies (actually, the shortest path bridging part of those technologies) allow us to shed the yoke of the Spanning Tree monster (see Understanding Switch Fabrics by Brandon Carroll for more details). Not surprisingly we wanted to know more and asked the obvious question: “and how would you connect the switches within the fabric?”

Chris sent me an interesting question that I haven’t covered in any of my DMVPN webinars: “How would you migrate a part of a Phase-1 DMVPN network to a Phase-2 or Phase-3 network if you can only migrate one spoke site at a time? Can I just upgrade the spokes that need spoke-to-spoke connectivity?”

While it might be theoretically possible to have a mixed Phase-1/Phase-2 DMVPN tunnel (and I just might be able to get it to work in a lab), such a solution definitely violates the KISS principle.

In the Redundant DMVPN Design, Part 1 I described the options you have when you want to connect non-redundant spokes to more than one hub. In this article, we’ll go a step further and design hub and spoke sites with multiple uplinks.

Public IP Addressing

Fact: DMVPN tunnel endpoints have to use public IP addresses or the hub/spoke routers wouldn’t be able to send GRE/IPsec packets across the public backbone.

One of my students was watching the Building IPv6 Service Provider Core webinar and wondered whether he should use 6PE or native IPv6 transport:

Could you explain further why it is better to choose 6PE over running IPv6 in the core? I have to implement IPv6 where I work (a small ISP) and need to fully understand why I should choose a certain implementation.

Here’s a short decision tree that should help you make that decision:

I hope you never believed the “OpenFlow networking nirvana” hype in which smart open-source programmable controllers control dumb low-cost switches, busting the “networking = mainframes” model and bringing the Linux-like golden age to every network. As the debates during the OpenFlow symposium clearly illustrated, the OpenFlow reality is way more complex than it appears at a first glance.

To make it even more interesting, at least four different models for OpenFlow deployment have already emerged: