A Brief Overview…: A Brief Overview… Vanessa Evans


LINX Then…: LINX Then… Established Oct 1994 by 5 member ISPs Objective to keep UK traffic in UK Encourage co-operation amongst competitors Housed in external independent facilities Telehouse North (TFM6) (something of a closed club)


LINX Today…. : LINX Today…. Completely open, the only rules are technical ones Maintained mutual model Not open to Capture 144 members, steady linear growth Increase in GigE, decreasing 100M Sharp increase in avg. traffic on GE late 2003 Still some consolidation, but strong new member prospects 22 members participating in managed PI 18 members participating on Route Server


Membership Growth: Membership Growth


2002-2003 Growth: 2002-2003 Growth


LINX Today…: LINX Today… Head Office in Peterborough, equipment and subsidiary office in London Still committed to providing resilience and redundancy Dual LAN architecture with dual switch vendors (Foundry and Extreme) Own dark fibre with multiple diverse routing between sites PoPs in 8 sites in Docklands & City Conservative approach to engineering configuration, backed up by roadmap look-ahead


Network Architecture: Network Architecture


Traffic: Traffic


Good of the Internet Activities: Good of the Internet Activities K-ROOT Housing for Nominet - .uk master RIPE NCC Routing Information Service Collector Packet Clearing House support cluster Slave ccTLD nameservers Austrian and Belgium I-ROOT Cluster (NetNod) routeviews.org BGP stats collector Team Cymru Bogon route server


LINX Training Programme: LINX Training Programme LINX Accredited Internet Technical (LAIT) Launched in 2000 to help address the critical shortage of skilled Internet engineers and technicians being experienced by ISPs Assists in raising the skills of those involved in installing, operating and maintaining Internet hardware and software Delivered by LINX commercial partner Systems & Network Training Project is therefore financially self sufficient


LAIT Pass Rate: LAIT Pass Rate


Management Structure: Management Structure


Engineering News: Engineering News


Next Gen Switches: Next Gen Switches November 2003 down to two contenders (Foundry & Cisco) Requirements finalised and lab evals completed: High-speed, high-density 10GE switch AC Power Must support 10G and 1G ethernet interfaces XENPAC 10G optics Either GBIC or SFP 1G optics Redundant management modules Field replaceable switch fabric(s) Field replaceable, hot-swappable line cards If no hitless failover today, roadmap commitment for CY 2004 Port density of at least: 2x 10GE ports per linecard 20x GE ports per linecard at least 50 ports of GE and 4 ports of 10GE, with at least 2 empty linecard slots Line rate ports, at 10G and 1G 64-bit interface counters Jumbo-frame support


And the winner is: And the winner is Foundry MG8


Initial Deployment Outline: Initial Deployment Outline February/March Software upgrades on existing platform Enable MRP, to allow rapid failover Late March/Early April Install one MG8 at Telehouse North and insert into 10GE ring (traffic will transit the switch) Ability to take “beta” 10GE customers Phased migration of switch1 GigE customers to MG8


Further Deployment: Further Deployment Then upgrade existing BigIron modules Phased rollout of JetCore modules across other 10GE LINX locations Then take beta customers for trunked GE (as required) Install second BigIron MG8 Location to be decided


Deployment Methodology: Deployment Methodology We’ll adopt our usual approach to this A “slow start”, making gradual changes Control the amount of risk presented at any one time Monitor the first deployments carefully If everything seems stable increase the rate of work Change control freezes where applicable


Future Developments: Future Developments The new hardware will support Sflow Add an Sflow collector, allow us to spot trends and anomalies in traffic patterns Possibility to double ring bandwidth to 20G New 2 port 10GE card for JetCore Transit through 10G at line rate, can add/drop 8G BI4000s in IX and InterXion not upgraded at this time


Pricing and Services : Pricing and Services


Fees: Fees Flat fee structure changed to Traffic Based Charging System Fairer on all Members As costs are split between Members New Members means greater distribution of costs which means lower fees for everyone Initiatives consistently undertaken to reduce fees 20% reduction in all port fees for last 6 months of last year Fees lower this year than any other year


Services: Services Current 100Mb/1Gb ports Multicast and/or IPv6 and/or IPv4 Provision of Rack Space in LINX own space Route Server BENTO Private Interconnect (intrasite and intersite) Between all LINX sites Basic level of support Planned DDOS Protection DNS Caching Trunked 1Gb ports 10Gb


“LINX from Anywhere”: “LINX from Anywhere” Organise marketing efforts with draft-martini ethernet carriers Link up prospects who don’t want to fully build out into London with a suitable carrier No need to deploy equipment locally Provide a means for IXP to IXP connections After three months 7 providers, covering 184 locations in 34 countries http://www.linx.net/joining/linx-anywhere/ Enables LINX to establish “Virtual Sites”


In Asia…: In Asia…


BENTO: BENTO Network Monitoring and Analysis tool Prospecting Set up BGP session with server Obtain report on In a perfect world, how much traffic can be offloaded at the Exchange Cross referenced against “Peering Slut” list and those members connected to Route Server


Looking Ahead: Looking Ahead


Looking Ahead: Looking Ahead Maintain critical mass and continue to provide value Continue efforts to reduce fees and maintain a fair pricing system Attract and secure new members Offer new peering opportunities Lower fees Deploy appropriate next generation switch technology Deal with port capacity issues Offer both 10 GigE and/or trunked ports to members Act on the results of the second member survey


Questions: Questions