dr Nenad Krajnović Elektrotehnički fakultet u Beogradu E

dr Nenad Krajnović
Elektrotehnički fakultet u Beogradu
E-mail: krajko@etf.bg.ac.rs
The ARPANET – početak...
Growth of the ARPANET (a) December 1969. (b) July 1970.
(c) March 1971. (d) April 1972. (e) September 1972.
2
3
Interconnection of Backbone
ISP Networks
POP
POP
POP
POP
POP
POP
POP
POP
NAP
IXP
POP
4
Path Choice
0.5$/MB
POP
POP
POP
POP
POP
POP
POP
NAP
POP
IXP
0.15 + 0 + 0.10 = 0.25 $/MB
POP
5
Commercial Impact on the Global
Network
There is no such thing as global backbone any more!
There is no such thing as Internet core any more!
Backbone of the global network consists of backbone ISP
networks, mutually interconnected at the Internet Exchanges (ex
NAPs).
The core of the networks consists of routers, switches and other
devices, located in the ISP backbones, preventing the global
network from splitting into parts.
Routing policies between major backbone ISPs determine the
ways of traffic flow on the Internet today.
Routing policies don’t assure optimal routing - the main criterion
for the routing policy design is - the cost of the links!
6
Network Access Points and
Internet Exchanges (NAPs & IXs)
1. IXPs in the USA:
•
•
•
•
SIX – Seattle, Salt Lake City, Vancouver
NYIIX – New York International Int. Exch.
DET-IX – Detroit Internet Exchange
DE-CIX, LINX, Equinix, W-IX,...
2. Major European NAPs:
•
•
•
•
•
DE-CIX - Frankfurt am Main
AMS-IX – Amsterdam
LINX – London
Netnod – Stockholm, Malmo, Sundsvall...
SOX – Serbian Open Exchange
https://www.euro-ix.net/ixps/listhttps://www.euroix.net/ixps/listixps/
7
Traffic on DE-CIX
8
Arhitektura DE-CIX-a
9
Traffic on LINX - 01.03.2016.
10
Arhitektura LINX-a (2016.) – I deo
11
Arhitektura LINX-a (2016.) – II deo
12
Traffic on AMS-IX – 01.03.2016.
13
Arhitektura AMS-IX-a (2016.)
14
Traffic on Netnod – 01.03.2016.
15
Traffic on Seattle IX – 29.02.2016.
16
Architecture of Seattle IX
17
Traffic on NYIIX – 29.02.2016.
18
Traffic on SOX – 01.03.2016.
19
Access Networks - Overview
20
Backbone Technologies
LAN technologies (PoPs
(PoPs,, NAPs):
Fast and Gigabit Ethernet (0.1(0.1-1Gbps) – almost ultimate PoP LAN
technologies today.
10Gbps Ethernet – developed, implemented on biggest sites.
100Gbps Ethernet–
Ethernet new standard for high demand sites.
WAN technologies (backbone lines):
Long
Long--haul Gigabit,
Gigabit, 10G Ethernet and 100G Ethernet
Mostly metropolitain area.
On xWDM optical transport networks Gigabit and 10G Ethernet
can be offered in WAN networks.
21
Backbone Interconnectivity
In order to exchange traffic, ISP’s must be connected to each other.
Every ISP should consider:
Buying transit connectivity from one or more larger-sized ISP’s (a must for
business!)
Interconnecting (for free!) its network with other local ISP’s of the same size:
Not a must for business and not always achievable (politics ...).
National ISP’s should be interconnected to other national ISP’s in the same country.
Same goes for regional ISP’s – they should interconnect with other partners in the
region.
Exceptional case are International backbone ISP’s:
Example: UUnet, TeliaSonera, Level3, Combridge, Sprint ...
They don’t buy transit service from anyone (they are transit
transit--free
free).
They interconnect their network (for free!) with other backbone ISP’s.
Interconnection between backbones is usually called peering
peering.
Peering can be:
Public – implementing by bringing backbone routers to the public IX/NAP’s.
Private – usually a leased line or local cable between two backbones.
22
Internet Routing
Protocol responsible for routing is IP (Internet
(Internet Protocol)
Protocol).
Connectionless!
Based on hop
hop--byby-hop paradigm, each router being a “hop”.
Source and destination address – carried within the IP header.
The router looks the destination address and forwards the packet to the next
hop.
We distinguish between packet routing and packet forwarding:
forwarding
Forwarding – sending the packet from one router to another, based on various
params.
Routing – complex process, includes routing info exchange among routers.
Always follows the same path, regardless of service used.
No guaranted QoS!
Within one backbone, QoS can be controlled using various mechanisms:
On MPLS backbones – by defining various QoS classes and using DiffServ.
On pure IP backbones – queueing, traffic shaping, per-service rate limiting.
Between two different ISP backbones – no guarantees for QoS at all!
23
Need for QoS
Users experience packet loss, delays etc.
Routers – need some time for processing IP packets (forwarding) ...
Routing – always follows the same path, regardless of service used ...
Some Internet services are delay or throughput sensitive:
File-transfer oriented services, like Web, Ftp, Usenet etc. – need throughput!
VoIP, Real video/audio and other interactive services are delay-sensitive!
For voice – delays over 100 ms cause a significant degradation!
Customers need firm QoS agreement with their ISP’s.
Service level agreements (SLA):
(SLA)
Defined between the customer and their ISP.
May include strict boundaries on delay, throuput, (un)availability etc.
Care must be taken to define realistic values, within the laws of physics:
RTT of 80-90 ms over transatlantic trunks is a realistic value.
RTT of 40 ms over transtlantic trunks is not realistic (speed of light !!!).
What can an ISP do about it?
24
Routing Between Backbones
Source
Hot potato routing:
Popular in early days of Internet (NSFNET).
Principle – get rid of the packet ASAP.
Give the packet away at the first possible hop
to the peering partner
Advantage if backbone capacities are a
problem.
Disadvantage, if strict QoS control is needed.
New York
Wash DC
London
Paris
Destin
Source
Cold potato routing:
•
•
•
•
Preferred strategy nowadays
Principle – keep the packet within your own
backbone as long as possible.
Advantage, better QoS control!
Disadvantage, if backbone capacities are a
problem.
New York
Wash DC
Destin
London
Paris
25
Internet Services
Services are:
The main raison d’etre of any telecommunications network.
network
Traffic generators and bandwidth consumers !!!
Drivers of network development and growth ...
Revenue generators for service providers.
From the technical point of view:
All services use TCP/IP protocol suite as the underlying communications engine.
All services follow the same client-server communication principle.
Services are a matter of the contract between the ISP and user.
The ISP reserves the right to restrict some services by AUP, additional fees etc.
So far, we used to talk about classical Internet services:
Email, Web, Telnet – derived from simple applications
Today, we must talk about value
value--added services:
Services for residential customers (home business and fun).
Services for business customers (corporate solutions).
All value-added services based on classical Internet services.
26
Basic Services
Term services here is equal to specific Internet appliances and applications.
Basic services – accessible almost everywhere in the Net.world
Net.world::
Electronic mail (email,
(email, gmail,
gmail, yahoo mail)
mail)
Information store and retrieval (Web,
(Web, ftp, cloud)
cloud)
Chat and interactive games (Viber
(Viber,, Twitter)
Twitter)
OnOn-line discussion ((Usenet
Usenet,, mailing lists, blogs)
blogs)
Unified communications (Skype,
(Skype, GTalk,
GTalk, Webex,…
Webex,…))
Net.broadcasting (YouTube, Netflix, webTV
webTV))
Interactive access to remote hosts (telnet
(telnet,, ssh)
ssh)
Value
Value--added services:
Telecommuting (access to corporate servers and services from home)
Distance learning (broadcasts of academic lectures, vendor courses etc.)
Online shopping
Small LAN (SOHO) interconnection
Computer--telephony and computer
Computer
computer--TV integration (Internet access from TV
etc.)
etc.)
Value--added services are based on one or more basic services
Value
(e.g. EE-commerce is mostly based on Web
Web))
27
C&W global cable systems
28
C&W USA cable systems
29
C&W – Europe cable systems
30
C&W – UK cable systems
31
C&W – Asian network
32
SPRINT – Europe network
33
SPRINT – USA network
34
SPRINT – Asia network
35
Cogent USA – world network
36
Global Internet Map
october 2009.
37
dr Nenad Krajnović
Elektrotehnički fakultet u Beogradu
E-mail: krajko@etf.bg.ac.rs