Currently Greg Ferro is duking it out with Cloud toad about MPLS. Is it dead? Will Ivan’s books be put to bed for ever? Find out. I will live blog this event – first attempt at this sort of thing. Lets see how this goes. Follow this link to join
May the best fighter win.
8:20 – Greetings and welcomes. Chats and mic testings.
8:30 – The realization that the cage match is freaking awesome. Still in holding pattern.
8:31 – 300 odd signups and very smart people getting in. Jeff Fry, Lindsey Hill, Nicolas Michel, to name a few.
8:35 – Derick is here. Final checks.
8:48 – Cloud toad is here. Technology sorted. Microphone jack into RJ45 😉
8:50 – Derick’s dulcet tones are resonating. Introductions and Welcomes.
8:51 – Derick’s clogging drains with MPLS tags, Greg “Crusty” Ferro – Python Invoker and Compass SPF wielder.
8:52 – Lets get ready to RUMMMMMMMMMBBBBBLE!
8:53 – Greg is collecting honey and angry and SDN is the way while Derick believes MPLS tool where you can have software written above the network. Ethan jests Greg is a tool too 😉
8:55 – GF: MPLS is NAT! (eww) Essentially nat (label) and then un-nat (pop label). Good call. Duct Tape of networking. Cannot configure MPLS in virtual environment. Cannot configure VRF to Hyper visor. MPLS has no part to play in the DC; therefore dead!
8:58 – DW: Servers increase network functionality. Overlays terminating. The new edge. MPLS doesn’t need to go to hyper visor. MPLS abstracts the network away from provisioning on top. Isolated forwarding path with simple import/export. Build VRF and Overlay. MPLS = build services over the network. Has role at the edge of the DC. L2 spanning DC is a bad idea. Agrees with Ivan P’s thoughts on this.
9:00 – GF: No API support for MPLS. Not programmable.
9:01 – DW: What is the need? What is replacing MPLS? Same work must be done. Abstraction still need to be made.
9:01 – GF: Must have an API due to multiple vendor implementations. How do you product a constant platform over the top?
9:02 – DW: Common problem. Vendors need to fit into abstraction model. You need it no matter if you use SDN or MPLS to abstract!
9:05 – DW and GF : Let’s avoid a Orchestration discussion.
9:06 – DW: Cisco/Juniper translators are required no matter what. Value of puppet, chef, openstack that allows interaction and vendor translators make magic happen.
9:07 – GF: SNMP queries are the only reliable thing to pull interface stats. MPLS needs this. Until then…
9:08 – GF: Stuck in the Dark Forest of Broccoli Despair – looking at MPLS and what information he can get and cannot see unless he makes dirty scripts or manual look up. Extracting route targets and route descriptors.
9:10 – DW: Roll up the sleeves and work with what we have.
9:12 – EB : Answer DW question. Tool of choice today and Globally deployed. How is it even dying?
9:12 – GF: We have a choice. Best answer? Program the TCAM directly with Open-flow Time and money is wasted pursuing MPLS. Keep on bolting on. MPLS-TE, MPLS-TP, MPLS over UDP over GRE, EVPN. show that is is broken. MPLS is now juice less. Dumb stateless Network protocol. Each router has to have a mapping. Overlay propagates tags. Costly to work one way. Costly to groom it. Dead!
9:15 – DW: MPLS is well know, it works. It isn’t as hard as you make it sound. Easy to deploy and lost cost. Growing in popularity. Not convinced it has to be centrally controlled. Open Flow will mutate like MPLS – stuff got bolted on. Derick predicted this in a PP blog. Widely used platform vs SDN which is narrow acute penetration.
9:17 – GF: ATM failed. MPLS will continue to be used in their central control plains. Carriers figure out with massive software deployments to monitor MPLS. They leave the mental comprehension of normal people and rely on machine.
9:19 – DW: You need monitoring for SDN too
9:19 – GF: Yes but MPLS can bite you in the ass no matter the size. Stops people doing dumb things. Tags can get out of hand.
9:20 – DW: Formulating response.
9:20 – GF: Good intentions with “poor” MPLS deployment. Separation by a 20 bit tag MPLS.
9:21 – DW: They are separated still. SDN is the same.
9:22 – GF: Agreed but….
9:22 – DW: Networking in a horrid state. “OH the humanity”. The course of humanity is altered by networks today. Essentially things that Greg is arguing about are orchestration, not MPLS issues.
Side Note –
[9:22:16 AM] Kurt Bales: Derrick’s arguement is that things Greg is complaining about are orchestration problems
[9:22:24 AM] Kurt Bales: *not* MPLS problems
[9:22:34 AM] Anthony Burke: Yeah
[9:22:57 AM] Kurt Bales: the exact same problems will happen with OpenFlow if the orchestration side isnt sorted out
9:23 – DW: version of openflow, operational ability and more. Concedes a point to Greg – driving the hope for SDN, program ability tools that make things easier to people. Why haven’t we built tools for automation processes that adds and subtracts IP’s. Why is it coded? Why do we need graph theory to do automation?
9:27 – GF: Why haven’t we converged around it yet? Why aren’t we using it? SDN and MPLS are a match made in heaven. Same as TRILL and STP.
9:28 – DW: I don’t have psychic powers yet. MPLS may die soon, who knows.
9:29 – GF: Drawing on experience – MPLS has hallmarks of death.
9:29 – DW: CLI isn’t the future. The future is orchestrating network elements.
9:30 – GF: Agrees. Wont be the primary use.
9:31 – EB: You boys are conceding little ones begrudgingly.
9:32 – EB Questions?
Not bad guys. Not bad. Stubborn but not bad 🙂
Plenty of fun to be hard. Read this by Brent Salisbury in the use of MPLS in the Enterprise .
Thanks for the information guys. It was a great format. Maybe time for a Marko vs Ivan ISIS v OSPF Fight night. There would be blood.