Next Meeting

Fall 2014 Netgurus Meeting

NANOG will provide to NetGurus, a room to be set in closed Board style conference seating to support 20-30 R&E Operators on Sunday, February 9, 2014 before NANOG 60 begins. We will have a projector and screen available if needed by participants. The actual room assignment will not be known for a few more weeks.  When established, we will share the room location/information.

Hotel

The NANOG Room Block at the Westin PeachTree will accommodate the early arrival of NetGurus.  They will find hotel reservation information on the web.

Tentative Agenda

Attendance

Contact Jeffry Handal (jhandal@lsu.edu) to RSVP and for topics you wish to discuss during the meeting. Attendance limit is 25.

Discussion Topics and Notes

• Perfsonar and SRX firewalls
  • Recommended placement: inside and outside network; AL2S network.
  • Bandwidth test: sets up blocks; shuffles bits on memory to NIC; there is no hard drive; almost line rate. Can overrun buffers on routers/firewall.
  • Check out perfclub.org. Group open to all.
  • New patches/upgrades coming for perfsonar.
  • Recommended boxes for using perfsonar: qbox, nerada, udroids, beagle.
  • Uses of perfsonar:
    • Before/after snapshots of performance
    • Data can be used by researchers
    • Jason and Ely working on workshop to help users understand perfsonar. Feel free to provide feedback. Contact Celeste.
  • Web10g coming. People need to comment.
  • LSU using v6sonar used for a sensor network for testing ipv4/ipv6 performance.
  • Lots of Asia pacific members use perfsonar.
    • They do not know the community function and they have it closed. Celeste can hook you up.
    • They create tools that allow cool stuff. Example: create network map from perfsonar connections.

• Creating a telepresence friendly campus.
  • Use an SBC
    • Application based firewall for video and voice
    • It will help identify packet loss.
    • Only telepresence sent through it.
    • Recommend using SIP line from provider and a vlan on the internet as backup.
  • Check out bluejeans from NET+.
  • Examples from campuses:
    • Set QoS and do not place behind campus FW.
    • Allow 1723 in from anywhere so people can dial in/out.
    • Cisco VCS express: inside and outside firewall only applies to tanderg (only video not voice).

• IPv6 measurement sharing
  • Deepfield project to replace portal.internet2.edu.
  • I2 issues with measurement:
    • v6 flows are behind - have to mirror traffic, juniper gear having issues (netflow v9 not available with certain hardware).
    • Atlas project may have some statistics that you can reference.
  • Farmer offers to community colleges the following option:
    • v6 free and v4 has a cost. Ipv6 adoption great!
    • Keeps costs down because he peer with HE for free.
    • Similar model Chinese colleges follow.
  • Security:
    • Tools not keeping up.
    • Traffic is there whether you deploy or not.
  • Issues:
    • Monitor both: v4 and v6 to find problems.
    • Recommend nagios/mrtg/cacti.
    • Monitor health of session for bgp.
    • Peering issues still exist.

• Multicast: test sources, future of ipv4 multicast
  • Negative:
    • Pacific wave: multicast not allowed; only p2p; little requests for ipv6 multicast.
    • Most campuses do not worry about it.
  • Positive:
    • Replication of wireless config with ipv4 multicast; next version may work for ipv6. (Cisco controller)
    • People using it, NOC getting more complaints on ipv4 multicast more in one month than in the last 1.5 years.
    • Netcast for on campus only. They would like digital signage.
    • Voip phones have a paging service over multicast.
  • Issues:
    • iptv with multicast: leaves are an issue; continues streaming and consuming bandwidth.
    • Cisco 3750 with ttl 0 problems arise from flooding.
  • Private industry out pacing us. Example: multicast LTE; financial center. Education missing content to offer.

• Engaging with researchers
  • Begin discussions with researchers. Invite them to technical meetings.
  • HPC tends to lead it. Hard part is people leave by the time the grant comes along. Always find a representative. People interested by what they will do is different.
  • Recommend nanog/geni/I2 provide research forum. Researchers need feedback. Allow publications.
  • Research issues topic of interest: interdomain routing for openflow.

• Long Range Ethernet alternatives
  • phybridge: does poe and Ethernet. Cas provided an example: ip phones work just fine.
  • 2900 module for 4g celular: instead of using as a backup, use as main connection. Check for data plan; they may be cheaper than renting dark fiber.
  • Consider bidirectional transceivers: 10,20, 30 km flavors. 1gig only. (Allied Teleson makes them.)

• Replacement of cacti graphs for a dynamic solution
  • Check out statseeker:
    • Install on a dell server with 8gig.
    • Nexus does like it on the fabic extenders.
    • v6 support coming.
    • Focuses on statistics only.
    • Very fast.
  • Consider mrtg instead of cacti.
    • Create scripts to scan periodically.
    • Clemson uses it mostly core devices.
    • Can be used for weather maps.
    • Script maker allows flexibility.
  • Check out router stud.cgi by steve shipway. Has book.
  • Check out snapp used by the I2 noc.
    • Dynamic front end. Very slick.
    • Free!
  • Other:
    • php weather map
    • drraw used for creating maps from same data.
    • whatsup

• Backbone upgrade plans (e.g. speeds, multi-vendor)
  • Examples of campuses:
    • Penn state: Moved from 6500 to Brocade mlx-E; 10gig to start with a few 100gig to computer routers; deployed 2 weeks ago and waiting to see what they will learn.
    • USF: looking at same decision as Penn but with 100gig only.
    • Georgia Tech: VSS at the core; Nexus bug for creating a multicast storm; dual 10gig, planning for 100gig.
      Also considering a 6904 with adapter for 10gig interfaces; OSI tend to work when keyed for Cisco.
    • Clemson: on their CCNIE, went with 40gig interfaces.
    • Other campuses: everyone seem to be happen with Nexus 7k but not happy with 100gig support.
  • Brocade fears:
    • Firmware upgrade
    • Documentation not great.
    • Forward error correction issue still pending.

• Data Center Interconnects (e.g. HA, L2 or L3)
  • Campus stories:
    • Clemson: L2 connectivity but may move all to L3 to avoid broadcast storms.
    • USF: L3 between data centers; recommend GLSB to move stuff around.
  • Recommendation:
    • Hit vendors with not allowing L2 movement of traffic.
    • Do not vmotion L2; DCs should be independent.
    • General consensus: kill L2 and do not let server guys say otherwise.
    • Use load balancers.
    • Check outL gtm = global traffic manager.
  • Kevin Miller has document showing why L2 is bad. (Sent to list)

• Other topics
  • Everyone buying 3rd party optics.
  • In Data Center: openstack applications dominating.
  • CCNIE grant results: Clemson created one flat L2 network and using big switch as controller. Have learned lessons; they were encouraged to share in the near future.
  • Outage notification systems:
    • Email notifications when network is down is dumb. A better notification system needed.
    • Example: blackboard connect. Clients decide decide how to get contacted.
  • Servers people use:
    • DHCP servers dominated by ISC (failover and load balancers are common deployment scenarios).
    • Radius servers: Free radius mostly. Some considering going to radiator. (Radiator written in perl but scales.)
  • eduroam:
    • Service expanding.
    • Campuses like Clemson and LSU using it as the only network to offer. Clients must get used to setting usernames as xxxx@xxx.edu.
    • Suggest eduroam for nanog
    • Question asked: Does eduroam use tls? Ideal when AD changes passwords.

• Future venue for Netgurus
  • Nanog and GENI still welcomes us.
  • Another possibility is the Technology Exchange in Indianapolis.
  • Quarterly call with one topic in mind.
  • Future still uncertain.

Dinner Options

TBD