Presentation on theme: "2 nd Gigapop Geeks BOF Jon-Paul Herron & Dan Magorian, Cochairs Welcome!! The idea is to recreate a forum where gigapop operators (aka regional service."— Presentation transcript:
2 nd Gigapop Geeks BOF Jon-Paul Herron & Dan Magorian, Cochairs Welcome!! The idea is to recreate a forum where gigapop operators (aka regional service providers) can rant, rave, and be politically incorrect about current hot technical topics. 1 st meeting was in HI, on NLR fanout. Plan is to keep having them at each Jt Techs workshop. Not oriented at administrative, management, or enterprise folks. However, we decided to skip IOS/JunOs testing at door or closed meetings and rely on Geekhood Self Determination. If you aren’t a GP Geek, what are you doing here?
So maybe we have the Right Folks: This is Your Meeting Hopefully lots of good discussion about gigapops’ unique set of technical issues. We aren’t pushing any particular agendas. There's plenty of time for multiple topics and free flowing discussion. Thanks to Internet2 folks for providing the food & drink. What are your ideas for what you’d like to hear discussed? Suggestions for topics? How about volunteering to present one? Where should it go? What this meeting isn’t: Rick & Steve’s Abilene Connectors, Heather’s Gigapop Administrators, Ken’s Routing WG, or “the Dan and JP show”.
Tonight’s Topic : MPLS and GMPLS: Are they in your future? Why or why not? Background: Feel free to discuss as go along! MPLS has been used for years by national ISPs for traffic engineering across their backbones, with IP at the edges. Until recently, it hasn't been used by campuses or regional service providers (aka gigapops). But today, because of the advantages it offers, it's starting to move in past the campus edge. Despite its increased complexity and atm-like characteristics.
So, what are MPLS’ advantages? Most Service Providers (SPs) consider them to be (from IETF MPLS drafts): –Supporting end-to-end services requiring QoS guarantees (aka, for networks too slow for best effort to support stuff like video & VoIP) –Performing network resource optimization (aka, filling bandwidth to avoid costs & peaks) –Providing fast recovery (aka, failover paths without losses due to routing convergence)
What do these mean in I2 context? QoS hasn’t done well because of I2 & gigapop approach of massive bandwidth overprovisioning for advanced apps. Eg 3-10x over sustained peaks vs SPs running 2x or less. Best effort in uncongested I2 context is often good enough. Inexpensive fiber gige regional bandwidth doesn’t have to be optimized carefully like carrier sonet. Most gigapops are single-attached w/o multiple path protection eg to Abilene, so no recovery. Does this mean there’s no advantages to MPLS that are useful to gigapops?
No! There are a different set of (related) advantages Solving policy “fish problems” via multiple routing tables (aka, RFC 2547 VPNs). Usually people still work around these or use other approaches, tho. Local traffic engineering. Even though bandwidth is inexpensive, gigapops are starting to need TE like larger national nets. And campuses who have traditionally used stuff like OSPF metrics for multi- exit TE are starting to move to MPLS. Overlay networks. MAX uses RFC2547 to deliver different ISP services to subsets of customers. We also stitch cust LSPs for route & traffic separation (poor man’s lambdas). Private nets are big growth!
So what are the disadvantages? Increased complexity. More protocols to run, have to figure out what they’re doing, trickier configs. But hey, that’s why you get paid big $$. Router memory limits # vpns depend on routes. Lack of visibility into what happening in tunnels. Need lsp-ping and better tools to see tunnels up. Additional peerings increase workload and add complications at edges like when multiple vpns dump into single CE routers (eg no ping return). Separate lambdas are cleaner (simple hardware setup vs complex software) but cost $$.
Beyond MPLS, GMPLS… Blatant ad: come hear my talk on GMPLS/ MPLS MAX/Abilene lsp testing, Mon 11am. Remember that “classical” MPLS is only over packet-switched networks. Generalized Multiprotocol Label Switching expands mpls to control fiber, lambda, sonet, and packet switching between routers, sonet tdms & optical crossconnects (OXCs). Gmpls is far more complex, nasty to set up today, still under development, and vendor interop is tricky. Why would you run it? If you have lambda or fiber switches that use it!
GMPLS is in your future! If you are a RON, optical switching is in your future, and will use gmpls. Right now optical paths are patched and resources are static, but large body of research is proceeding on dynamically switched resource protocols. Lots of researchers working on it, & production folks will start to need switched benefits esp oversubscription soon given traffic growth rates. Fiber costs $$. Look at what switching did historically for ethernet vs shared hubs. Chances are very good that we’ll be using optical switching in production sometime soon.
Dynamic resource switching This will be big change for the industry from static resource allocation as it gets deployed in next few years. Right now, not a lot of economic benefit from optical switching: if you’ve paid to provision a path, why not leave it up? Protocols like BGP expect stability, and get unhappy when resources shifted under their feet. Lots of protocol development needed, as well as “bandwidth brokers”. So why not experiment with it today, both large- scale (NLR) and local testing, so we’ll know what we’re doing tomorrow?
Discussion Your MPLS/GMPLS projects and thoughts What tech topics are bugging you these days? Beer-driven brilliant inspirations, Politically Incorrect Flames, Whatever. Talk about it! Your future Gigapop Geek BOF ideas Feel free to contact us: –Jon-Paul Herron, firstname.lastname@example.org,email@example.com –Dan Magorian, firstname.lastname@example.org. Thanks for coming!