the role of ixps in the internet - os3 · short historical overview ... kpn, isps, news providers,...
TRANSCRIPT
Henk Steenman AMS-IX CTO [email protected]
The Role of IXPs in the InternetAMS-IX in particular
AgendaStructure of the Internet and role IXPsPeeringAMS-IX
The InternetTraditional structure of the InternetRole of IXPsCurrent structure of the Internet IXPs today
Structure of the Internet
The place of an IXP in the Internet
Interconnection
The Internet is all about
interconnection!
Interconnection
The place of an IXP in the Internet
Autonomous NetworksOrganisation
Purpose
Management
Infrastructure
Symbol: cloud
The place of an IXP in the Internet
Not autonomousNetwork Protocols
BGP OSPF
Addressing
IPv4, IPv6
The place of an IXP in the Internet
Impossible to interconnect each network with each otherTransit providers can provide traffic between networksLarger networks (geographically) offer transit to smaller ones
The place of an IXP in the Internet
Transit (sometimes called upstream) costs money.
Typically as a price per Mbit/s available bandwidth
X€/Mbits
Lower price for higher volumes
The place of an IXP in the Internet
Instead of paying for transit for all traffic: keep local traffic localPeering: Exchange traffic between more or less equal parties locallyTransit remains necessary for traffic which can not be handled locally
The place of an IXP in the Internet
Facility where many (more than 2) parties can exchange traffic with each otherUsually peeringTypical technology
Ethernet
The place of an IXP in the Internet
A lot of companies gather together in the data centers that co-locate an IXPEasy to make direct connections bypassing the IXP
Private Peering
More control over traffic exchange
Even cheaper
The place of an IXP in the Internet
Larger IXPs also attract larger ISPs, CDNs etcPeering not obligatoryLarger networks do not always want to peer
Or not for free
Telecity AMS5 Amsterdam ZO
Datacenters: 350 S Cermak, Chicago, IL
IXPs
Short Historical Overview
Metropolitan Access ExperimentMetropolitan Area EthernetMetropolitan Area ExchangeWorldCom MAE-East™
Washington, D.C.
Eind 1992
10mb/s ethernet
Initieel 4 lokaties
MFS fiber hub
Gedeelde administratie door deelnemers
Sprint/ICM, Alternet, PSI, SURAnet, NSFnet
Short Historical Overview
MAE-West / Federal Internet ExchangeSan Jose / Mountain View (1994)
FDDI ring
Bridged tot 10mb Ethernet in verschillende lokaties
Commercial Internet ExchangePalo Alto (1994)
Layer-3 MLPA
DS1 (T1) lines naar een Cisco 7010
1991 Not-for-profit industry association
Alternet, PSI, Cerfnet
Short Historical Overview
LINX London1994Opgezet als een verenigingEthernet
AMS-IXZie later
Hong Kong Internet ExchangeUniversity of Hong Kong
Ethernet switch
Eerste grote gratis Exchange (1995)
AMS-IX Hong Kong (2012)
Technological ChangesShared 10Base-TSwitched 10mb EthernetShared FDDISwitched FDDI100Base-T / 100Base-FX
IEEE 802.3u standaard beschikbaar Jul 1995
Gigabit EthernetIEEE 802.3z standaard beschikbaar Jul 1998
10Gigabit EthernetIEEE 802.ae standaard beschikbaar Jun 2002
100GEIEEE 802.3ba standaard beschikbaar sinds juli 2010Begint nu langzaam 10GE LAGs te vervangen
Additional ServicesRoute-serverLooking-glassMeasurement and instrumentationNetwork Time ProtocolWeb cache parentNews serverRoot server mirror
Different Business Models
Managed and operated by University or GovernmentAssociationNot for profit companyNeutral for-profit company
Typical as value added service by Data Centers
InformalA switch is placed somewhere and anyone can connect
Usually unmanaged and unsupported
Different Geographical Size
CommunityLarger Metro AreaNational backboneRegional Backbone
IXPs in Europe
IXPs in Europe
Euro-IX IXPs
Peering
PeeringThe Exchange of traffic between parties where only each others customers are advertised is called peering
Typically happens on an IXP or via Private Interconnects
Peer in fact means “equal party” Large carriers peer with large carriers and small ISP’s with small ISP’sOr in case of content providers where there is equal gain
Peering happens typically without the exchange of moneyThe benefit of peering is that it significantly reduces costs of transferring IP traffic and provides for faster data-flows (shorter paths, less router hops)
Traffic exchanged directly does not need to be transferred upstream and saves IP transit costs and less “hops” between routers
Peering
To realise 3 major aims:Money savings Faster, more direct data-flows
Less network hops, more direct routes
Redundancy in routing.Contractual requirements. Status.
PeeringIs usually arranged on a bilateral basis between companiesOften via the exchange of e-mailsAlso at gatherings of peering coordinators
Global or European peering eventsRIPE/Nano/Apricot/aSnog etc
Optimisation is peering via route serverWhen peering policy is open, a straight forward way to establish large number of peerings day one when connected.
AMS-IXOrganizational structureHistorical OverviewTechnical infrastructureSome statisticWhere do we go from here?
AMS-IX Organisation
AMS-IX OrganisationAMS-IX is a non-profit, neutral and independent, professional exchange. The beginnings of the exchange can be traced back to the early 1990's. On the 29th of December 1997 AMS-IX was established as an Association in the Netherlands, operating under Dutch Law. Amsterdam Internet Exchange B.V., founded in 2000, is the company of which the Association is a 100% owner that acts as the operator and administrator of the exchange. The Association today exists of over 500 members, mostly ISP’s but also (mobile) telecom providers, carriers, hosting providers, ASP’s and content providers
IntroductionAMS-IX Association and Limited Liability Company
From the Articles of the AMS-IX Association1.The objective of the Association is to facilitate the exchange of (internet-)transactions and everything that contributes thereto in the broadest sense of the word. The Association does thereby not strive to make any profit2.The Association tries to achieve this objective amongst others by providing directly or indirectly a connection to the Amsterdam Internet Exchange to the members of the AssociationLimited Liability Company
Single shareholder (the AMS-IX Association)Same objective as Association
AMS-IX Governance Structure
The Board has 5 seats and is elected by the members, out of member representatives, in the General Meeting
- The Board is both Executive Board of the Association as well as Supervisory Board of the Corporation - The General Meeting is both the General Assembly Meeting of the Association as well as the shareholders meeting of the Corporation
AMS-IX OrganisationOriginally AMS-IX (association) was managed by the AMS-IX board
Operational management outsourced to SURFnet/SARA
From 2000 onward AMS-IX BV responsible for day to day management
From Jan 1 2002 technical management
AMS-IX BV staff:3 persons in 200040 end of 2014 (18 nationalities)
Roughly half is technical (NOC, Engineering, Development)
11
Member profile - origin (roughly) early 2014
20 % National MembersKPN, ISPs, news providers, hosting parties, broadcasters
80 % International MembersOf which 50 % from Europe
12% DE, 12% UK, 8% FR, 8 % RU, rest mix
Of the other 50%50 % from North America (US, CAN)50 % from the Rest of the World
12% Asia
Who is connected?
Member profile- Type of members
ISP’s/Access providers XS4All, UPC/Chello, Ziggo, Orange, Free, T-online etc
Carriers/Telco’sKPN, BT, FT, DT, AT&T, TeliaSonera, Cable&Wireless, Level3, Cogent, Sprint, etc
Content providers/CDNsRTL, NPO, BBC, Google, Yahoo, Microsoft, Highwinds, Giganews, Akamai, Limelight, CDNetworks etc
Hosting providersLeaseWeb, Proserve, SupportNet, ServerCentral etc
Mobile operatorsVodafone, Orange, T-mobile, Aicent, Telefonica IWS, SingTel etc
Pricing
23
Port Type Monthly Fee in Euro
< GE NA through reseller only
GE 500
10GE 1000
100GE 5000
AMS-IX Amsterdam in numbers 2014
AMS-IX Around the World
AMS-IX Technical Infrastructure
The Platform
Evoswitch
Global Switch
Terremark
Interxion AMS5
AMS-IX Offices
SARANIKHEFEquinix AMS3
Telecity 4
Eunetworks
Telecity2EquinixAMS 1&2
Telecity 5
AMS-IX Amsterdam Platform
1GEPE router1GE
PE router1GEPE router1GE
PE router1GEPE router1GE
PE router
1GEPE router1GE
PE router1GEPE router1GE
PE router1GEPE router
PXCPXC
PXCPXC
PXCPXC
PXCPXC
PXCPXC
PXCPXC
PXCPXC
CE CE CE CE CE CE CE
PXCPXC
PXCPXC
PXCPXC
PXCPXC
PXCPXC
PXCPXC
PXCPXC
CE CE CE CE CE CE CE
1GEPE router
CE
CE
10GEPE router
10GEPE router
10GEPE router
10GEPE router
10GEPE router
10GEPE router
10GEPE router
EUN X1510 | 100 GEPE router
10GEPE router
10GEPE router
10GEPE router
10GEPE router
10GEPE router
10GEPE router
EQX3 X3310 | 100 GEPE router
NH X2810 | 100 GEPE router
GLO X1910 | 100 GEPE router
TC2 X2510 | 100 GEPE router
TC4 X10 | 100 GEPE router
EQX1 X2910 | 100 GEPE router
10GEPE router
10GEPE router
SE CoreSF P router
TMK10 | 100 GEPE router
SAR X2210 | 100 GEPE router
IXN X2410 | 100 GEPE router
NH X2610 | 100 GEPE router
TC5 X3210 | 100 GEPE router
TC2 X2010 | 100 GEPE router
10GEPE router
10GEPE router
Core Core
10GEPE router
EVO X3110 | 100 GEPE router
PEP PE
PXC
PEPE
PXCDLSPctlACL WatcherARP Sponge
PXC
RouteServer
Virtual Customer connection
move
Route Server Peering
MAC Address Change
AS1200 Peering
Provisioning
AS1200
my.ams-ix
my.ams-ix
OSPF BFD LDP RSVP
Control and AutomationThe Platform Control Model
Physical L1/L2 Infrastructure
Photonic Cross Connects
“Autonomous” control plane
Operations control plane
Customer L2 control plane
L3 Infrastructure
Customer L3 control plane
Control and Automation
Distinction between Operations and Customer control plane
Operations = AMS-IX NOC
Customer = AMS-IX Customer services, Customers, Resellers
Provide better and direct control to users of their connection and use of the platform.
AMS-IX NOC for management of infrastructure and troubleshooting
AutomationMost control process are automated
Why?More and more customers on more and more platforms in different parts of the world.
Automated provisioning take care of a lot of repetitive tasks.
No human access to production platform, fewer errors
Provide customers interface to their IX connection to change basic parameters without interaction with the NOC
Currently MAC address, Route server peering
Allow resellers with multiple physical ports on the platform to manage virtual customer ports
AMS-IX Technical InfrastructureThe MPLS setup
stub-tel-22010.10.2.20/32
stub-nik-21410.10.2.14
core-glo-20110.10.2.1
core-glo-20210.10.2.2
core-eun-30110.10.3.1/32
core-eun-30210.10.3.2
stub-tel-32010.10.3.20/32
pxc-sar-101
AMS-IX Platform
MPLS/VPLS-based peering platform
X LSPs between each pair of access switches
over one or more core (P) routers
Load balancing of traffic over multiple LSPs
10/100GE access switch resilience
10/100GE customer connection on PXC
Protection of access connection
AMS-IX PlatformOSPF
BFD for fast detection of link failures
RSVP-TE signalled LSPs over predefined pathsprimary and secondary (backup) paths defined
VPLS instance per VLANStatic defined VPLS peers (LDP signalled)Load balanced over parallel LSPs over all core routers
Layer 2 ACLs to protect customer port
AMS-IX PlatformSingle OSPF area
Loopback addresses and backbone links in OSPFChoice for OSPF (instead of ISIS) arbitrary based on available expertise
BFD for rapid detection of failure in forwarding pathBi-directional Forwarding detection
Detect faults in bi-directional path between two forwarding engines
Allows for very fast convergence of OSPF in case of link failure
bfd interval 50 min-rx 50 multiplier 10
AMS-IX PlatformAccess switches (PE) act as Label Edge RouterCore (P) act as transit Label Switch Router
Penultimate, label is popped on core instead of egress LER
LSPs follow pre-defined paths through the networkRSVP-TE for LSP signalling
stub-tel-22010.10.2.20/32
stub-nik-21410.10.2.14
core-glo-20110.10.2.1
core-glo-20210.10.2.2
core-eun-30110.10.3.1/32
core-eun-30210.10.3.2
stub-tel-32010.10.3.20/32
pxc-sar-101
LSP definitionsMPLS/VPLS setup
stub-tel-22010.10.2.20/32
stub-nik-21410.10.2.14
core-glo-20110.10.2.1
core-glo-20210.10.2.2
core-eun-30110.10.3.1/32
core-eun-30210.10.3.2
stub-tel-32010.10.3.20/32
pxc-sar-101
Pre-defined paths between PEsover each core router
Resilience
stub-tel-22010.10.2.20/32
stub-nik-21410.10.2.14
core-glo-20110.10.2.1
core-glo-20210.10.2.2
core-eun-30110.10.3.1/32
core-eun-30210.10.3.2
stub-tel-32010.10.3.20/32
pxc-sar-101
MPLS/VPLS setup
LSP over primary Path
LSP over backup Path
stub-tel-22010.10.2.20/32
stub-nik-21410.10.2.14
core-glo-20110.10.2.1
core-glo-20210.10.2.2
core-eun-30110.10.3.1/32
core-eun-30210.10.3.2
stub-tel-32010.10.3.20/32
pxc-sar-101
Resilience in access connectionby means of PXC
stub-tel-22010.10.2.20/32
stub-nik-21410.10.2.14
core-glo-20110.10.2.1
core-glo-20210.10.2.2
core-eun-30110.10.3.1/32
core-eun-30210.10.3.2
23/3 - 24/410.2.20.9/30
1/4
31/3 - 32/410.2.20.10/30
MPLS, LSPs
AMS-IX Platform
Interface ethernet 23/3 enable load-interval 30 route-only bfd interval 50 min-rx 50 multiplier 10[…] ip ospf bfd ip address 10.2.20.9/30 link-fault-signaling!router mpls […] mpls-interface e23/3
router mpls policy ingress-tunnel-accounting load-interval 30
mpls-interface e31/3
interface ethernet 1/4 enable load-interval 30 mac access-group 403 in
router ospf area 90 bfd all-interfaces
stub-tel-22010.10.2.20/32
stub-nik-21410.10.2.14
core-glo-20110.10.2.1
core-glo-20210.10.2.2
core-eun-30110.10.3.1/32
core-eun-30210.10.3.2
23/3 - 24/410.2.20.9/30
1/4
31/3 - 32/410.2.20.10/30
MPLS, LSPs
AMS-IX Platform
router mpls policy ingress-tunnel-accounting load-interval 30
mpls-interface e31/3 path 220-201-214 strict 10.10.2.20 strict 10.10.2.1 strict 10.10.2.14
path 220-202-214 strict 10.10.2.20 strict 10.10.2.2 strict 10.10.2.14
path 220-301-214 strict 10.10.2.20 strict 10.10.3.1 strict 10.10.2.14
path 220-302-214 strict 10.10.2.20 strict 10.10.3.2 strict 10.10.2.14
stub-tel-22010.10.2.20/32
stub-nik-21410.10.2.14
core-glo-20110.10.2.1
core-glo-20210.10.2.2
core-eun-30110.10.3.1/32
core-eun-30210.10.3.2
23/3 - 24/410.2.20.9/30
1/4
31/3 - 32/410.2.20.10/30
MPLS, LSPs
AMS-IX Platform
lsp 220-201-214 to 10.10.2.14 from 10.10.2.20 primary 220-201-214 secondary 220-301-214 standby priority 2 0 revert-timer 64800 enable
VPLS: Multipoint to Multipoint VPN
AMS-IX Platform
‣ VPLS to emulate the shared L2 infrastructure
‣ LDP used in control plane.
‣ Distribution of VPLS labels and MAC addresses
‣ PEs pre-defined
‣ Full mesh of LSP (virtual circuits) between each PE (access) device
‣ Actually X LSPs (one over each core) between each pair
‣ Manually configured
‣ Traffic between pair of PEs load balanced over these X LSPs
‣ Association of customer interface (L2) to VPLS instance
‣ One VPLS instance per VLAN
‣ Loop free as by default no packets arrived over an LSP is forwarded on another LSP
VPLS traffic load balancingAMS-IX Platform
vpls ISP 501
[…]
vpls-peer 10.10.2.14 load-balance vpls-peer 10.10.2.14 lsp 220-201-214 220-202-214 220-301-214 220-302-214 […]
vlan 501 tagged ethe 1/1 ethe 1/3 ethe 5/4 ethe 18/3 ethe 19/2 ethe 20/4 ethe 21/1 to 21/2
untagged ethe 1/2 ethe 1/4 ethe 2/1 to 2/4 ethe 3/1 to ......
AMS-IX Technical InfrastructureCustomer Connections
1GEPE router1GE
PE router1GEPE router1GE
PE router1GEPE router1GE
PE router
1GEPE router1GE
PE router1GEPE router1GE
PE router1GEPE router
PXCPXC
PXCPXC
PXCPXC
PXCPXC
PXCPXC
PXCPXC
PXCPXC
CE CE CE CE CE CE CE
PXCPXC
PXCPXC
PXCPXC
PXCPXC
PXCPXC
PXCPXC
PXCPXC
CE CE CE CE CE CE CE
1GEPE router
CE
CE
10GEPE router
10GEPE router
10GEPE router
10GEPE router
10GEPE router
10GEPE router
10GEPE router
EUN X1510 | 100 GEPE router
10GEPE router
10GEPE router
10GEPE router
10GEPE router
10GEPE router
10GEPE router
EQX3 X3310 | 100 GEPE router
NH X2810 | 100 GEPE router
GLO X1910 | 100 GEPE router
TC2 X2510 | 100 GEPE router
TC4 X10 | 100 GEPE router
EQX1 X2910 | 100 GEPE router
10GEPE router
10GEPE router
SE CoreSF P router
TMK10 | 100 GEPE router
SAR X2210 | 100 GEPE router
IXN X2410 | 100 GEPE router
NH X2610 | 100 GEPE router
TC5 X3210 | 100 GEPE router
TC2 X2010 | 100 GEPE router
10GEPE router
10GEPE router
Core Core
10GEPE router
EVO X3110 | 100 GEPE router
AMS-IX Platform Hub/Spoke MPLS/VPLS
Low TrafficVolumes
High TrafficVolumes
Access ConnectionsHigh Speed Access connection protected
Brocade MLXe10,100 GE PE
PXC
Brocade MLXe10,100 GE PE
Brocade MLXe1 GE PE
To Core(P) routers
To Core orSub Core(P)routers
To Core orSub Core(P)routers
Photonic Switching
Glimmerglass Networks switch64 to 192 port MEMS based switchConnect any port to any other port
Glimmerglass PXC: Switching engine
ReflectingMirror
Micro Mirror Array
Micro lensArray
Fiber Array
PXC ApplicationPXC used for protection of CE to PE
Swap connection between identical pair of PEsHard and software failures on PEs manageableHelps in troubleshootingAllows for non service interrupting maintenance
10GE Access10GE access
The Platform
10GE Access10GE access 10GE Access10GE access
PXC
X * 10GE, X >= 1
10 and 100 GE PE
1 2 3
1 2 3
PE PE
PXC
CE
PXCD
PXCD
Manages Photonic Cross Connects
Directs failover of customer connections beween pair of PEs
Triggers are manual or events in the platform
LSP up/down
PE PE
PXC
PXCD
LSPctl
ACLwatcher
ARP Sponge
PXC
Operations Control Plane
LSPctl
Manual Control of LSPs
Up/Down, failover to backup path
ACL watcher
Check number of L2 ACL violations on customer ports
If too high shutdown port and/or move to quarantine VLAN
ARP Sponge
Control of ARP traffic on platform
Reply with own address in case of too many requests
Remote ConnectionsEvolution of “remoteness”
‣ Originally ISP router co-located with IXP switch
‣ ISP network extended to Exchange co-location
‣ Sometimes ISP router in different close by location and connect “remote” over for example dark fiber, wavelength etc
‣ E.g. Netnod
‣ Router in remote location, connection over layer 2 transport provider
‣ Can be any distance between ISP router and IXP
‣ Most extreme: AMS-IX member with router in Los Angeles
‣ Initially one router per connection, now multiple.
Direct ConnectionMember/Customer router co-located with Exchange switch
Co-Location
Exchange SwitchMember/
Customer Router
MAC #1
Direct ConnectionMember/Customer router not co-located with Exchange switch,
connects for example via dark fiber
Co-Location
Exchange SwitchMember/
Customer Router
MAC #1
One MAC limitation‣ Protections against loops over customer
layer 2 infrastructure
‣ Manageability on a per customer (MAC) basis
‣ Assume multiple customers (each identified by MAC address of their router) behind a single port
‣ One customer sends unwanted traffic to Exchange
‣ Cannot troubleshoot or disconnect without impact other customers behind same port
‣ One MAC enforced by L2 ACL
Why one MAC per member connection?Shared infrastructure (one or multiple
members) behind IXP
Co-Location
Exchange Switch
MAC #1
MAC #2 MAC #3...N
Member Switch/Router
Member Switch/RouterMember
Layer 2infrastructure
Multiple Direct ConnectionsLoop in members shared infrastructure, other members MAC
behind member port
Co-Location
Exchange Switch
MAC #1
MAC #2 MAC #3...N
Member Switch/Router
Member Switch/RouterMember
Layer 2infrastructure
MAC #3...N
Single MAC behind Customer Port
‣ Complications with L2 loops over customer infrastructure and application of L2 ACLs or Port security were reason for initially large number of IXPs to not allow remote L2 connections
‣ Only a few of those now left
Remote ConnectionRemote physical connection to L2 provider, L2 provider connects to
Exchange
Co-Location
Exchange SwitchMember/
Customer RouterL2 Provider
Access routerL2 Provider
router
Transport ProviderNetwork
MAC #1 MAC #1
Remote ConnectionLogical Connection between Member/Customer router and
Exchange switch
Co-Location
Exchange SwitchMember/
Customer RouterL2 Provider
Access routerL2 Provider
router
Transport ProviderNetwork
MAC #1 MAC #1
Remote ConnectionMultiple Logical Connections over the same number of physical
connections
Co-Location
Exchange SwitchL2 Provider
router
Transport ProviderNetwork
L2 Provider Access router
Member/Customer Router
Member/Customer Router
MAC #1
MAC #1
MAC #2
MAC #2
Remote ConnectionReseller Port: Single AMS-IX connection, multiple customers
‣ Since Sep 2009 AMS-IX MPLS/VPLS platform
‣ Can uniquely identify customer router by MAC/Qtag combination and translate to different VPLS instances
‣ L2 ACL on combination MAC/Qtag
‣ Manageable: can move traffic stream identified by MAC/VLAN to separate VPLS instance and troubleshoot
‣ Resellers can take one 10GE or 100GE connection on platform and resell parts of it (100, 250, 500, 1000 10000 Mbps etc)
Remote ConnectionMultiple Connections, single Exchange connection
Transport provider is AMS-IX reseller
Co-Location
Exchange SwitchL2 Provider
router
Transport ProviderNetwork
L2 Provider Access router
Member/Customer Router
Member/Customer Router
MAC #1
MAC #2
MAC #1
MAC #2
Remote Connection2 logical connections over single physical connection
Co-Location
Exchange SwitchL2 Provider
router
Transport ProviderNetwork
L2 Provider Access router
Member/Customer Router
Member/Customer Router
MAC #1
MAC #2
MAC #1
MAC #2
AMS-IX Technical InfrastructureThe Physical Network
AMS-IX BackboneConnections between PE and P routers via dark fiber
Lighted up by predominantly 100GE interfaces in the switches
Typical distances between PE and P routers in the Amsterdam metro area 10 to 35 km.Typical 100GE LR4 optics (most common) have a reach of 10km
Interswitch links > 10km
Many of our backbone links are too long for current 100Gbit/s (100GE LR4) transceivers.What are the options?
Optical Amplification.Transmission equipment.
ADVA, MRV, etc
We use these on a number of links but are expensive
Optical amplificationCurrent 100Gb/s CFPs:
10x10 for 2km.10x10 for 10km.100G-LR4 for 10 km.
We need to drive up to 25 to 40km of fibre:AttenuationDispersion
0"
0.5"
1"
1.5"
2"
2.5"
3"
800" 900" 1000" 1100" 1200" 1300" 1400" 1500" 1600" 1700"
A"en
uario
n*(dB/km
)*
Wavelength*(nm)*
O-band
C-b
and
L-ba
nd
S-ba
nd10
x10
(<40
km)
100G-LR4
SOA applianceWe only could find components So together with appliance vendors we designed our own solution
TEC+
TEC-
NTC
NTC
Laser
A K
TEC Driver
Laser DriverController
SOA appliance
SOA consist of three partsLaser
NTC (temperature sensor)
Peltier element (TEC)
SOA applianceTEC driver to keep the temperature constant.
Feedback loop.
Laser driver to control the current through the laser.
Feedback loop.
Controller
1st Prototype
ResultsCube optics and Snootlabs
Build appliance as commercial productUsable up to 35km
This is suitable for most of our needs.
AMS-IX PlatformStatistics
AMS-IX Customer Connections
Traffic measurement based on router interface counters and sFlow statistics
MRTG used to collect and graph countershttp://oss.oetiker.ch/rrdtool/
sFlow used to gather statistical data on member to member data
AMS-IX current Status
AMS-IX Status
0.00#
5000.00#
10000.00#
15000.00#
20000.00#
25000.00#
1'Jan'05#
1'Jul'0
5#
1'Jan'06#
1'Jul'0
6#
1'Jan'07#
1'Jul'0
7#
1'Jan'08#
1'Jul'0
8#
1'Jan'09#
1'Jul'0
9#
1'Jan'10#
1'Jul'1
0#
1'Jan'11#
1'Jul'1
1#
1'Jan'12#
1'Jul'1
2#
1'Jan'13#
1'Jul'1
3#
1'Jan'14#
1'Jul'1
4#
Traffic#Volume#Tbyte/day#
Traffic patterns over Time
Traffic through the year follows a very constant trendNot so obvious when looking at the data on a per date basisWhen normalised to weeks and weekdays it becomes more obvious there is a regular pattern during the weekWhen normalised to weekdays and average yearly traffic, traffic behaviour is very constant and predictable
0.00#
200.00#
400.00#
600.00#
800.00#
1000.00#
1200.00#
1400.00#
1)Jan#8)Jan#
15)Jan#
22)Jan#
29)Jan#
5)Feb#
12)Feb#
19)Feb#
26)Feb#
4)Mar#
11)Mar#
18)Mar#
25)Mar#
1)Apr#
8)Apr#
15)Apr#
22)Apr#
29)Apr#
6)May#
13)May#
20)May#
27)May#3)Jun#
10)Jun#
17)Jun#
24)Jun#
1)Jul#
8)Jul#
15)Jul#
22)Jul#
29)Jul#
5)Aug#
12)Aug#
19)Aug#
26)Aug#
2)Sep#
9)Sep#
16)Sep#
23)Sep#
30)Sep#
7)Oct#
14)Oct#
21)Oct#
28)Oct#
4)Nov#
11)Nov#
18)Nov#
25)Nov#
2)Dec#
9)Dec#
16)Dec#
23)Dec#
30)Dec#
Gbps# Average#Traffic#Gbps#
2004#2005#2006#2007#2008#2009#2010#2011#2012#
week$2$ week$6$ week$10$ week$14$ week$18$ week$22$ week$26$ week$30$ week$34$ week$38$ week$42$ week$50$
0.00#
200.00#
400.00#
600.00#
800.00#
1000.00#
1200.00#
1400.00#
Tuesda
y#Sund
ay#
Friday#
Wed
nesda
Mon
day#
Saturday#
Thursday#
Tuesda
y#Sund
ay#
Friday#
Wed
nesda
Mon
day#
Saturday#
Thursday#
Tuesda
y#Sund
ay#
Friday#
Wed
nesda
Mon
day#
Saturday#
Thursday#
Tuesda
y#Sund
ay#
Friday#
Wed
nesda
Mon
day#
Saturday#
Thursday#
Tuesda
y#Sund
ay#
Friday#
Wed
nesda
Mon
day#
Saturday#
Thursday#
Tuesda
y#Sund
ay#
Friday#
Wed
nesda
Mon
day#
Saturday#
Thursday#
Tuesda
y#Sund
ay#
Friday#
Wed
nesda
Mon
day#
Saturday#
Thursday#
Tuesda
y#Sund
ay#
Friday#
Wed
nesda
Mon
day#
Saturday#
Thursday#
Tuesda
y#Sund
ay#
Friday#
Wed
nesda
Mon
day#
Saturday#
Thursday#
Tuesda
y#Sund
ay#
Friday#
Wed
nesda
Mon
day#
Saturday#
Thursday#
Tuesda
y#Sund
ay#
Gbps##
Average#Traffic#
2004#2005#2006#2007#2008#2009#2010#2011#2012#
Yearly traffic 2004 - 2012
0.60$
0.70$
0.80$
0.90$
1.00$
1.10$
1.20$
1.30$
1.40$
1.50$
1.60$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Average$Traffic$$Normalised$to$Weekdays$Normalised$to$Yearly$average$
!Average!Traffic!
2006$2007$2008$2009$2010$2011$2012$
Yearly traffic 2006 - 2012
0.60$
0.70$
0.80$
0.90$
1.00$
1.10$
1.20$
1.30$
1.40$
1.50$
1.60$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
Friday$
Wed
nesday$
Mon
day$
Saturday$
Thursday$
Tuesda
y$Sund
ay$
4$dis>nc>ve$regions$
!Average!Traffic!
2008$2009$2010$2011$2012$
Week 2Week 26 Week 33 Week 46
Historical Traffic Growth
0%#
20%#
40%#
60%#
80%#
100%#
120%#
140%#
160%#
0#
1000#
2000#
3000#
4000#
5000#
6000#
7000#
8000#
2005# 2006# 2007# 2008# 2009# 2010# 2011# 2012# 2013# 2014#
Pbyte#
Yearly'Volume'Growth'and'YoY'percentual'growth'
10GE and 10GE LAGs
100#
150#
200#
250#
300#
350#
400#
450#1(Jan(11#
1(Mar(11#
1(May(11#
1(Jul(1
1#
1(Sep(11#
1(Nov(11#
1(Jan(12#
1(Mar(12#
1(May(12#
1(Jul(1
2#
1(Sep(12#
1(Nov(12#
1(Jan(13#
1(Mar(13#
1(May(13#
1(Jul(1
3#
1(Sep(13#
1(Nov(13#
1(Jan(14#
1(Mar(14#
1(May(14#
1(Jul(1
4#
1(Sep(14#
1(Nov(14#
1(Jan(15#
10GE#LAGs#10GE# 20GE# 30GE#40GE# 50GE# 60GE#70GE# 80GE# 90GE#100Gbps# 110GE# 120GE#140GE# 160GE# 180Gbps#
100GE and 100GE LAGs
0"
5"
10"
15"
20"
25"
1&Sep&12"
1&Nov&12"
1&Jan&13"
1&Mar&13"
1&May&13"
1&Jul&13"
1&Sep&13"
1&Nov&13"
1&Jan&14"
1&Mar&14"
1&May&14"
1&Jul&14"
1&Sep&14"
1&Nov&14"
1&Jan&15"
100GE"LAGs"100GE" 200GE" 300GE"
Virtual Connections
0"
50"
100"
150"
200"
250"
300"
20'Jan'10" 20'Jan'11" 20'Jan'12" 20'Jan'13" 20'Jan'14"
Virtual"Connec6ons"10Mbps" 100Mbps" 250Mbps"
500Mbps" 1000Mbps" 1500Mbps"
2000Mbps" 2500Mbps" 3000Mbps"
3500Mbps" 4000Mbps" 4500Mbps"
5000Mbps" 9000Mbps" 10000Mbps"
20000Mbps" 30000Mbps" 40000Mbps"
60000Mbps" 80000Mbps"
AMS-IX Status
0"
50"
100"
150"
200"
250"
300"27(Aug(10"
27(Oct(10"
27(Dec(10"
27(Feb(11"
27(Apr(11"
27(Ju
n(11"
27(Aug(11"
27(Oct(11"
27(Dec(11"
27(Feb(12"
27(Apr(12"
27(Ju
n(12"
27(Aug(12"
27(Oct(12"
27(Dec(12"
27(Feb(13"
27(Apr(13"
27(Ju
n(13"
27(Aug(13"
27(Oct(13"
27(Dec(13"
27(Feb(14"
27(Apr(14"
27(Ju
n(14"
27(Aug(14"
27(Oct(14"
27(Dec(14"
Low"Speed"Customer"Connec@ons"
E" FE" GE" 2GE" 4GE"
AMS-IX Projections
0.00#
1000.00#
2000.00#
3000.00#
4000.00#
5000.00#
6000.00#
1*Jan*09#
1*Jan*10#
1*Jan*11#
1*Jan*12#
31*Dec*12#
31*Dec*13#
31*Dec*14#
31*Dec*15#
30*Dec*16#
Gbit/
s#
Where to go from here
Automation!
Question ?