Dec 16, 2015
Overview of talk
HEAnet FAQ - what, why, who, where. National and international links
- Gigabit ethernet in the WAN Some services
- Mirroring
- Cisco Distributed Director Some projects IP routing fundamentals
- IP routing in HEAnet using OSPF, IS-IS & BGP
What?
HEAnet is the national network for education and research
Not-for-profit, not-for-loss too. Operating since 1983 Incorporated in 1997 Located in Dublin PoPs in Cork, Dublin, Citywest, Galway, Limerick &
Kilcarbery
Why?
Intercommunication Resource sharing Economies of scale Research collaboration Network technology National strategy
Who?
40 institutions (universities, ITs, …) Owned by the members Small central staff Collaboration with members International associations/partnerships
Where? HEAnet Ltd., Ground Floor, Brooklawn House,
Crampton Ave, Shelbourne Rd, Dublin 4
Tel: +353-1-6609040 Fax: +353-1-6603666
E-mail: [email protected] or [email protected]
GPS Coordinates: 53.3381ºN 6.2403ºW
Web: http://www.heanet.ie
Network Map
DCU Bandwidth Usage
DCU – MRTG (12th –20th March)
Maximum bandwidth available to DCU : 63 Mbps ATM over STM1
Mantova Pic
Gigabit Backbone
Citywest – Galway (March 12th)
Citywest – NOC March 17th –25th
Gigabit Ethernet WAN
Bigger, Better, Faster? Pros
Mass Produced Less Demanding of Hardware Less Complex to Configure Higher Bandwidth Cheaper
Gigabit Ethernet WAN
Cons Slower convergence times More difficult to troubleshoot Next steps may lose some of the
advantages 10GE (add another zero)
Complicates things by adding more framing for long haul.
May cost the same as STM-64.
Gigabit Ethernet WAN
National Backbone Links Router to router, very straightforward.
Client Links HEAnet to TCD, Ratelimited 120Mbps
over GE. Cisco 7200 onsite in TCD. Point-to-point
from there into a Routing-Switch. Allows better visibility of circuit and
development of Routing-Switches makes things cheaper.
International Links
155Mbps: Teleglobe, Abilene & StarTap 622Mbps: GlobalCrossing 1Gbps: JANET, Ca*net4, INEX 2.5Gpbs: Géant
Some Operational Services
National Information Server IP address and Domain registration Listserv Statistics Seminars, training Databases Webhosting Mirroring IPv6 NTP Network security evaluation Hot standby (Cisco DD)
Mirroring - ftp.heanet.ie
2.2 Terabyte Server with dual GigE connectivity IPv6 Capable in HTTP/FTP and RSYNC One of the busiest mirrors in Europe and an official mirror of
RedHat, Debian, FreeBSD Over 60 projects mirrored Developed custom Apache patches to tweak service. [email protected]
Hotstandby Web services
HEAnet offer failover protection for webservers using Cisco “Distributed Director” technology
Router (Cisco 3620/7200) acts as an authoritative DNS server with a TTL of 0
Clients cannot cache the DNS records Router monitors the primary webserver with full TCP
connections on port 80 at specified intervals (eg. Every 10 seconds)
Cisco Distributed Director
Router has IP address of both main webserver and backup “standby server”
If the router cannot connect to the main webserver it will redirect queries to standby
Continues to initiate connections to main webserver and reverts back to it when the webserver recovers
In HEAnet CDD is simply set-up to telnet to webserver on port 80. If it receives a response it considers the webserver to be up but more granularity can be added.
Has limitations
- IPv6 not supported
- https not supported
Some HEAnet service developments
IPv6 (Also a service)
Test traffic project
Multicast
MAN
Honeypots
Videoservices
Netflow
PKI
National Backbone
Next Generation Internet
IP routing Fundamentals
Routing is the process of selecting the interfaces through which a packet should be sent
Forwarding is the process of actually moving this packet once the route has been determined
Classful routing only passes network addresses, classless routing uses both network and subnet addresses
Classless interdomain routing (CIDR) was brought in in a bid to reserve IPv4 addresses and allows blocks of class C addresses to be combined/aggregated eg. 192.168.1.0 /21
More fundamentals …
Variable length subnet masks (VLSM) - greater granularity eg.192.168.0.0/28
Distance Vector protocols use hop count as a metric, link state (eg. OSPF) use cost. Hybrid protocols like EIGRP can use complex rules involving delay, reliability
Convergence is the length of time after a network change (Interface/line down, router added etc.)that routing tables take to update
An Autonomous system (AS) is a group of IP networks operated by one or more network operator/s which has a single and clearly defined external routing policy
Rip v1/v2
RIPv1 – RFC1058 - Classful, Periodic updates, transfers entire routing tables, distance vector, uses holddown timers.
RIPv2- Added classless, authentication, multicast support.
Uses “routing by rumour” as it only sees directly connected routers
Both of these routing protocols are inefficient and are now only used on very small networks. They have been
phased out in favour of more “intelligent”Interior gateway protocols (IGP’s) such as OSPF and IS-IS.
OSPF Overview
Features
- Supports CIDR/VLSM
- Fast convergence
- Uses multicast addresses for hello packets
- Load balancing
- Uses areas to cut down CPU usage Uses hello packets to establish adjacencies initially via
a 6 step process Once adjacency established it uses regular Link State
advertisements (LSA) and Link state updates (LSU) to describe links & adjacencies changes
OSPF
9 types of LSA, 5 of which are used extensively :
1-Router link
2-Network link
3-Network summary
4-ASBR summary
5-AS external.
(other 4 for TOS) Link state databases (LSDB) built from info in LSA’s Each router in an Area has the same LSDB and using this &
the SPF algorithm (Dijkstra) to build its routing table It’s an efficient & quiet routing protocol as it can have separate
LSDB for each area & only sends LSA updates
OSPF
Works under different network topologies differently
- Broadcast multiaccess (LAN)
- Point to Point (eg. Serial connection)
- Point to Multipoint (eg ATM using multiple VC’s)
- Point to Multipoint (Non broadcast multi access (NBMA)) Reacts quickly to network changes (Line down, router death)
- Hello interval every 10 secs
- Dead interval (Link down) is 4 times hello interval Routing descisions based on cost of 100000000/1
- Cost ranges from 0 to 65535 with 0 being the best
- 100Mb = OSPF cost of 1
- OSPF can do load balancing between same cost routes
OSPF in HEAnet
In HEAnet we have only one area (backbone: area 0)although we are investigating the use of multiple areas in the future (or using IS-IS)
In conjunction with static routes and BGP we use OSPF to route traffic throughout our internal network
Routers in DCU have full visibility of all routers running OSPF
in our AS ie. They have the same LSDB as the other routers in the AS
Does have its limitations:
- Currently running OSPF v2 which doesn’t handle IPv6
- OSPF v3 does but isnt considered stable enough yet
- For IPv6 we need to run another routing protocol IS-IS
IS-IS
IS-IS (Intermediate System to Intermediate System). Part of the OSI standard.
ISO/IEC 10589, RFC 1195 and RFC 2763 Not originally designed for IP. Uses Link-State PDUs, similar to OSPF
LSAs. Extensible Protocol, updated to deal with
IPv6 faster than OSPF. Add more Type-Length-Values to the PDUs
to handle more protocols.
IS-IS in HEAnet
Used purely for IPv6 routing within the HEAnet network.
Communication between routers. IPv6 connections to clients.
BGP
BGP (Border Gateway Protocol) is an EGP (External Gateway Protocol) and an IGP (Internal Gateway Protocol).
RFC 1771 Deals with AS’ (Autonomous Systems) Exchanges routing information between
peers. Makes routing decisions based on a number
of different criteria applied to each route.
BGP
These criteria feed into the BGP Best Path Algorithm. Each step is gone through until a definite routing decision can be made.
Invalid paths are ignored. Common values/criteria:
Local Preference AS Path Length Multi Exit Discriminator (MED)
Will eventually choose the path to the router with the lowest IP address.
BGP in HEAnet
We have BGP peerings with all of the networks we directly connect to eg Géant, JANET, Ca*net4, Global Crossing, Teleglobe etc.
A large number of the routers in HEAnet form part of our internal BGP mesh.
Allows for proper route selection to any accessible point in the internet from anywhere in HEAnet.
Questions ?