Transcript
Data Center Fiber Cabling Topologies and Lengths Paul Kolesar IEEE 802.3 HSSG, September 2006
Purpose and background • Provide channel length distributions within Data Centers as guide for setting distance objectives • Focus on Data Center cabling in support of view that >10G applications will initially be largely DC centric • Data gathered from CommScope corporate sales data base of pre-terminated cabling and patch cords
Kolesar, September 2006, IEEE 802.3 HSSG
2 Jim Hulsey
Primer on Data Center Cabling
Kolesar, September 2006, IEEE 802.3 HSSG
3 Jim Hulsey
What is pre-terminated cabling? • Custom-length cabling delivered with factoryinstalled connectors on both ends • Cable is plugged into the back of patch panels MPO Plugs (12-fiber array connectors) OM3-type Cable (aqua color) Kolesar, September 2006, IEEE 802.3 HSSG
4 Jim Hulsey
What is pre-terminated cabling? • Arrays are typically fanned out to present multiple duplex ports at front of panel Front side of panels present duplex ports
or arrays Array cables plug into back of fanouts Kolesar, September 2006, IEEE 802.3 HSSG
5 Jim Hulsey
Why is pre-terminated cabling relevant? • Speeds installation and turn up – Eliminates field termination process – All assemblies factory tested
• Highly advantageous for Data Center builds – Primary Data Center cabling choice for SYSTIMAX Solutions customers
• Length distribution directly applicable to Data Center channels • Array cabling supports parallel transmission Kolesar, September 2006, IEEE 802.3 HSSG
6 Jim Hulsey
TIA TR42 standardized array cabling for parallel applications ANSI/TIA-568-B.1-7 “Guidelines for Maintaining Polarity Using Array Connectors” Approved: January 13, 2006 Addendum to TIA 568-B.1 structured cabling standard
Structured cabling now extends beyond duplex Kolesar, September 2006, IEEE 802.3 HSSG
7 Jim Hulsey
Basic p2p parallel channel topology
PUSH
PULL
PUSH
Patch Panel PULL
PULL
Array Patch Cord PUSH
Array Xcvr
Rx1 Rx2 : : Tx2 Tx1
Channel consists of: 1 cable 2 cords 2 connections PUSH
PULL
Patch Panel PULL
PUSH
Kolesar, September 2006, IEEE 802.3 HSSG
PUSH
Array Patch Cord PULL
Array Xcvr
Rx1 Rx2 : : Tx2 Tx1
Array Cable (Permanent Link)
8 Jim Hulsey
Central cross connect topology Central Cross Connect
2
Channel consists of: 2 cables 3 cords 4 connections
3
Patch Cord
Array Cables (Permanent Links)
Patch Cord
Xcrv Kolesar, September 2006, IEEE 802.3 HSSG
Patch Cord
1
4
Xcrv 9 Jim Hulsey
Why use a central cross connect? • Allows changeable connectivity between any equipment area – The hub of star-wired architecture – Greater flexibility with less cabling than p2p
Equipment Area 1
Equipment Area 2 Central Cross Connect
Equipment Area 3 Kolesar, September 2006, IEEE 802.3 HSSG
Equipment Area 4 10 Jim Hulsey
P2P requires more cabling • For similar connectivity capability relative to central cross connect for star wiring
Equipment Area 1
Equipment Area 2
Equipment Area 3
Equipment Area 4
Kolesar, September 2006, IEEE 802.3 HSSG
11 Jim Hulsey
Why are central cross connects relevant? • Important and popular for managing connectivity, especially in larger data centers • In the following analysis the data on array cable lengths and patch cord lengths must be combined to form the length of complete channels • To derive channel length distributions, one must account for some mixture of central cross connect topologies and basic p2p topologies, but the precise mix is unknown Kolesar, September 2006, IEEE 802.3 HSSG
12 Jim Hulsey
Data Center Cabling Length Data
Kolesar, September 2006, IEEE 802.3 HSSG
13 Jim Hulsey
Array cable length distribution • Many thousands of 12-fiber units Distance between fiber panels in data centers 20%
100%
18%
90%
16%
80%
14%
70%
12%
60%
10%
50%
8%
40%
6%
30%
4%
20%
2%
10%
0%
0% 0-30
31-60
61-90
91-120 121-150 151-180 181-210 211-240 241-270 271-300 >300 ft
73 m Kolesar, September 2006, IEEE 802.3 HSSG
14 Jim Hulsey
Permanent link length distribution • Longer tail emerges Distance of Permanent Link (i.e. w/o cords) assuming ~1/2 are concatenated thru central cross connect 20%
100%
18%
90%
16%
80%
14%
70%
12%
60%
10%
50%
8%
40%
6%
30%
4%
20%
2%
10%
0%
0% 0-30 3160
61- 91- 121- 151- 181- 211- 241- 271- 301- 331- 361- 391- 421- 451- >480 90 120 150 180 210 240 270 300 330 360 390 420 450 480 ft
128 m Kolesar, September 2006, IEEE 802.3 HSSG
15 Jim Hulsey
Patch cord length data • Distilled from data on tens of thousands of cords • Mean Length = ~11 ft Standard Deviation = ~12 ft • Two Standard Deviations covers ~94% within this distribution statistical length of cords in channel number of cords 1 2 3 2 # std dev 34 55 74 11 17 22
4 91 28
ft m
3 cords covers central cross connect topology
Kolesar, September 2006, IEEE 802.3 HSSG
16 Jim Hulsey
Putting it all together • Recommend aiming to cover 90 to 95% of Data Center channel lengths – Precedent: 300 m covers similar percentage of in-building LAN backbones
• Using a simple summation of constituent lengths near this coverage level gives:
Distance Requirements at ~95% Coverage ft m Permanent Link Allocation 420 128 3 Patch Cord Allocation 74 22 Total Channel Allocation 494 150 Recommend setting short reach length objective = 150 m Kolesar, September 2006, IEEE 802.3 HSSG
17 Jim Hulsey
Backup Material
Kolesar, September 2006, IEEE 802.3 HSSG
18 Jim Hulsey
Analytical approach using statistics • Using means and standard deviations for all component lengths Objective Distance as a Function of Concatenated Channel Ratio and Coverage 1 sigma
2 sigma
3 sigma
250 225 200
meters
175 150 125 100 75 50 25 0 0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Concatenated Channel Ratio
Kolesar, September 2006, IEEE 802.3 HSSG
19 Jim Hulsey
