Advances Toward Economic and Efficient Terabit LANs and WANs Cees de Laat Advanced Internet Research...
-
Upload
hester-cross -
Category
Documents
-
view
221 -
download
0
Transcript of Advances Toward Economic and Efficient Terabit LANs and WANs Cees de Laat Advanced Internet Research...
Advances Toward Economic and Efficient Terabit LANs and
WANs
Cees de LaatAdvanced Internet Research Group (AIRG)
University of Amsterdam
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
News Flash:
You are sitting on aquarter terabit Lan!
Check your seats!
You have a network connection of 1 Gigabit/s in
your chair!
250 seats * 1 Gb/s ->0.25 Tb/s
How to define a Terabit/s LAN?• Depends where you put the boundary!
– Backplane of a switch?
• What is the functionality– Many times point to point (dwdm/wss/mems)– Broadcast/multicast capabillity– Packet service versus streaming
• Application requirements– Packet services or streaming
• LAN-WAN interface– Addressing (mac -> ip)– Routing to the edge of the system– Integrate LAN and WAN
Terabits to the Desktop by 2010
• Simplified User View
• Terabit Fiber Connection To The Desktop
• Integrated Photonics And Electronics
• Single Fiber Dense-WDM
• Packets And Flows
• Encryption
• …
“ Ethernet ”
1990 10 Mb
1995 100 Mb
1998 1 Gb
2002 10 Gb
2006 100 Gb
2008 1 Tb
2010 10 Tb
Source: Steven Squires, Chief Scientist HP
OptIPuter is PrototypingThe PC of 2010
• Terabits to the Desktop…
• 100 Megapixels Display – 55-Panel
• 1/3 Terabit/sec I/O– 30 x 10GE interfaces– Linked to OptIPuter
• 1/4 TeraFLOP – Driven by 30 Node
Cluster of 64 bit Dual Opterons
• 1/8 TB RAM• 60 TB Disk
Source: Jason Leigh, Tom DeFanti, EVL@UICOptIPuter Co-PIs
So, is a Terabit LAN just a bigger ethernet switch?
If yes: just wait and buy!
Low power GigaBit/s @ home•5 years ago Gigabit/s networking was a master student project•Don’t forget power consumption www.delaat.net @ 28 Watt
intermezzo
So, is a Terabit LAN just a bigger ethernet switch?
If yes: just wait and buy!
Why no:Grid -> collective services!Anti-statistical behavior!
Towards Hybrid Networking!• Costs of optical equipment 10% of switching 10 % of full routing equipment for
same throughput
– 10G routerblade -> 75-300 k$, 10G switch port -> 5-10 k$, MEMS port -> 0.5-1.5 k$
– DWDM lasers for long reach expensive, 10-50 k$
• Bottom line: look for a hybrid architecture which serves all classes in a cost
effective way ==> map A -> L3 , B -> L2 , C -> L1
• Give each packet in the network the service it needs, but no more !
L1 ≈ 1 k$/portL2 ≈ 5-10 k$/port L3 ≈ 75+ k$/port
Cost of 1 Tb/s LANvery rough estimate sept 2005
• L2– 10 gig nic ~ 1 k– Per L2 switch port ~5 k– Total 6 M
• L1– Per L1 port ~ 1 k– Total 2 M
• OOPS, I forgot the cables and the power bill
SURFnet6
6000 km Dark Fiber
Our National
Laboratory
Dordrecht1
Breda1
Tilburg1
DenHaag
NLR
BT
BT NLR
BT
Zutphen1
Lelystad1
Subnetwork 4:Blue Azur
Subnetwork 3:Red
Subnetwork 1:Green
Subnetwork 2:Dark blue
Subnetwork 5:Grey
Emmeloord
Zwolle1
Venlo1
Enschede1
Groningen1
LeeuwardenHarlingen
Den Helder
Alkmaar1
Haarlem1
Leiden1
Assen1
Beilen1
Meppel1
Emmen1
Arnhem
Apeldoorn1
Bergen-op-ZoomZierikzee
Middelburg
Vlissingen Krabbendijke
Breukelen1
Ede
Heerlen2Geleen1
DLO
Schiphol-Rijk
Wageningen1 Nijmegen1
Hilversum1
Hoogeveen1
Lelystad2
Amsterdam1
Dwingeloo1
Amsterdam2
Den Bosch1
Utrecht1
Beilen1
Nieuwegein1Rotterdam1
Delft1
Heerlen1
Heerlen1
Maastricht1
Eindhoven1
Maasbracht1
Rotterdam4
3XLSOP
IBG1 & IBG2Middenmeer1
AmsterdamCore
location:maximum4 * 72 * 10= 2.8 Tb/s
StarPlane:DWDMbackplaneforDAS-3
CPU’sCPU’s
CPU’s
CPU’s CPU’s
NOC
cdλ
CPU’s
MEMS
Client SURFnet
WS+AAANOC
WS+AAACPL
slide 2
Transport of flows
BWRTT
# FLOWS
For what current Internet was designed
Needs more App & Middleware interaction
C
AB
Full optical future
?
GLIF nowGLIF Future?
Questions ?
More info:
http://www.science.uva.nl/~delaat
slide 0c
iGRID2005 publication opportunity
"Future Generation Computer Systems (FGCS): The International Journal of Grid Computing: Theory, Methods and Applications" will publish a SPECIAL iGRID ISSUE in Spring/Summer 2006.
Guest editors: Larry Smarr, Tom DeFanti, Maxine Brown, Cees de Laat
We can accept around 20-25 papers, Papers will be reviewed
* Maximum paper length is limited to 8 pages* Limit of 1 paper per demonstration.* Describe your iGrid experiences, results and performance measurements.* DEADLINE for submission is ONE MONTH AFTER iGRID -> Oct 31.
Submission must be via the FGCS website. For author guides and submission information, see <http://ees.elsevier.com/fgcs/>.
Contact: Cees de Laat [email protected] (need reviewers :-)
QuickTime™ and aTIFF (Uncompressed) decompressorare needed to see this picture.