Transcript ppt - Terena

EARNEST Technical Study
Kevin Meynell
TERENA
Bratislava
27 April 2007
Technical Sub-Study Areas
• Transmission Technologies
– Equipment evolution, next-generation standards, transmission
protocols & fibre provisioning.
• Control Plane Technologies
– Switching & routing matrices (optical & IP), multicasting, IPvX,
QoS provisioning.
• Operations and Performance
– End-to-end performance, network management (optical & IP),
VPN provisioning & PERT.
• Middleware
– Authentication and authorisation infrastructures, mobility, PKI,
support for network infrastructure, virtual organisations.
The EARNEST Foresight Study
2006 - 2007
2
Technical Sub-Study Panel
Lars Fischer (Nordunet) – Transmission
John Graham (Indiana University) - Transmission
Otto Kreiter (DANTE) - Transmission
Gigi Karmous-Edwards (MCNC) - Control Plane (Optical)
Alexander Gall (SWITCH) - Control Plane (IP routing)
Stig Venaas (Uninett) - Control Plane (Multicast)
Dimitra Simeonidou (University of Essex) – Operations & Performance
(Optical)
Luca Deri (University of Pisa/Netikos) - Operations & Performance (IP)
Simon Leinen (SWITCH) - Operations & Performance (IP)
Diego Lopez (RedIRIS) - Middleware
Milan Sova (CESNET) - Middleware
Klaas Wierenga (SURFnet) - Middleware (Mobility)
The EARNEST Foresight Study
2006 - 2007
3
Completed and Scheduled
Interviews
• Completed:
– 22/01/07 MERLIN Radio Astronomers, Jodrell Bank, UK
– 30/01/07 IBM, Teleconference
– 07/02/07 Alcatel-Lucent, Teleconference
– 09/02/07 Sun, Teleconference
– 1-2/03/07 Juniper, Sunnyvale, USA
– 27-28/03/07 Cisco, San Jose, USA
– 29/03/07 Force10, San Jose, USA
– 25/04/07 Liberty Alliance, Brussels
• Scheduled:
– 28/04/08 SxIP, Teleconference
– 11/05/07 Calient, Paris
– Alcatel-Lucent, Nortel, Ciena, DTU-COM (tba)
The EARNEST Foresight Study
2006 - 2007
4
Preliminary Findings
Caveat:
So far only interviewed router & ethernet
switching vendors.
Interviews with carrier-class vendors still to
come, and may tell different story.
The EARNEST Foresight Study
2006 - 2007
5
Preliminary Findings
• Currently only a few OC-768 (40 Gbps) customers, mostly
in oil and gas industries
• Reluctance to upgrade transport network to support 40
Gbps, as expensive (x20 the cost of 4 x 10 GE) and seen as
interim step before higher speed standards.
• Running into problems with n x 10 Gbps, due to link
aggregation and load-balancing performance.
• Cisco, Juniper and Force10 pushing for 100 Gigabit Ethernet
standard.
– Little interest in separate SDH-compatible WAN-PHY variants
(<5% of sales)
– 100 GE standard expected by 2009, with implementations by
2010.
– How to implement: 16 x 6.25, 10 x 10, 4 x 25, or 1 x 100
Gbps?
– Copper standard for 100 GE being considered.
– 10 x 100 GE linecards and serial 100 GE possible by 2011.
The EARNEST Foresight Study
2006 - 2007
6
Preliminary Findings
• Sun backing 40 GE, presumably due to PCI-X and
Infiniband limitations.
• Some interest in SDH-compatible 80 Gbps (OC-1536) and
160 Gbps (OC-3072) standards.
• No mention of OC-256 (13 Gbps) or OC-384 (20 Gbps).
• 1 Tbps standard possible by 2020.
• Cisco pushing IP over DWDM.
– Little need for traditional traffic aggregation (data and
telephony) as services increasing provided by IP.
– IP can provide much of the management and fault
tolerance.
– Reduction of equipment and processing needs, reducing
per-port cost by up to 40%.
The EARNEST Foresight Study
2006 - 2007
7
Preliminary Findings
• Equipment consolidation
– Multi-terabit platforms with virtualised routing.
– 5000-7000 watts for typical 12 x 100/120 x 10 GE
chassis.
– Tunable transponders on router/switch interface cards.
– Wider use of Reconfigurable Optical Add/Drop
Multiplexers (ROADM).
– 10 GE expected to cost 1.5-2K per port by 2010.
The EARNEST Foresight Study
2006 - 2007
8
Preliminary Findings
• Routing scalability becoming problematic (again).
– Huge rise in number of hosts, fragmentation of service
provider hierarchy, and amount of traffic.
– Global routing table now >200,000 entries, which is
causing memory and processing problems (0.5-1 GB
memory required).
– Other reasons – more multihoming, traffic engineering,
plus IPv6.
– Proposed to split IP addresses into identifiers and
locators. [Possible implications for AAA as well]
• Improvements to TCP for sustained high-bandwidth
transmissions.
• Juniper pushing (G)MPLS, but Cisco less interested.
The EARNEST Foresight Study
2006 - 2007
9
Preliminary Findings
• In Europe different technologies used for higher
education federations:
–
–
–
–
Liberty (ID-FF)
Shibboleth (SAML-based)
PAPI
A-Select
• In US:
– Mainly Shibboleth
• Good news: SAML2.0 makes all of them interoperable
– Shib SAML2 version should be released within 6 months
The EARNEST Foresight Study
2006 - 2007
10
Preliminary Findings
• Identity Management is big area of interest for
vendors
• Different approaches to implement federations:
– Identity Federations: Liberty Alliance and SUN
– User centric-model
• Fairly new concept, implemented by Microsoft and
OpenID
– Abstract identity framework (Higgins, IBM)
– Some alliances between vendors
• Probably to compete with Microsoft
• Trust is a big concern for vendors
– The user centric approach seems to guarantee more
privacy to the users
The EARNEST Foresight Study
2006 - 2007
11
Remaining Work
• End-May:
Outline report table-of-contents.
• Early-June: Interviews completed.
• June/July: Produce draft report.
• July:
further
• August:
Technical Panel to meet to consider
additions, modifications, and
recommendations.
Final report published.
The EARNEST Foresight Study
2006 - 2007
12