Transcript standby environment

NSMA Annual Conference
May 19 & 20, 2015
Holiday Inn Rosslyn at Key Bridge
Arlington, Virginia
A Study of Propagation in a Difficult
Environment
Two Years in Mississippi
George Kizer , Alcatel-Lucent
Network Overview
top of the state
142 Sites
150 Paths
488 Links
1952 T-Rs
near the gulf
2
Network Path Varied
Maximum Path Length = 32 miles
Minimum Path Length = 3 miles
Median Path Length = 18 miles
Average Path Length = 19 miles
Six paths were at 11 GHz
All others were at lower or upper 6 GHz
17% of the paths were non-diversity
83 % were space diversity
Architecture = rings and spurs
3
Path Performance
System design was in accordance with standard Bell Labs
criteria (below)
Path propagation performance of all 488 simplex paths was
measured over one to two years (as network was created)
Vigants, “Space Diversity Engineering,” Bell System Technical Journal, pp. 103-142, January 1975
Vigants, “Microwave Radio Obstruction Fading,” and Schiavone, “Prediction of Positive Refractivity
Gradients for Line-of-Sight Microwave Radio Paths,“ Bell System Technical Journal, pp. 785 – 822, July –
4
Aug 1981
Multipath Fading
Overall the network measured path performance met
customer specifications: 99.998% one way
5
Multipath Fading
You can drown in a river with average depth of one foot –
or a network with average satisfactory performance
6
Path Performance
Path performance variations were primarily the
following:
IP Network Characteristics
Path Multipath Fading
Path Obstruction Fading
7
IP Network Characteristics
MW network supported a two layer IP network:
a Multiprotocol Label Switching (MPLS) network which
supported another overlay proprietary LMR IP network
Each network used Open Shortest Path First (OSPF)
for packet routing
Each network used Bidirectional Forwarding
Detection (BFD) to monitor route reachability
8
IP Network Characteristics
OSPF validity was tried to BFD
The lower level MPLS network went down after three
consecutive 100 millisecond BFD probe failures
(e.g., after a MW path outage greater than 0.3 secs)
The higher level LMR network went down after three
consecutive 300 millisecond BFD probe failures
(e.g., after a MW path outage greater than 0.9 secs)
9
IP Network Characteristics
MPLS OSPF was restored 15 seconds after BFD
came up
The LMR OSPF was restored 45 seconds after BFD
came up
10
IP Network Characteristics
Individual Fading Events
For router networks, individual outage durations is not the
only criteria
A single one second MW path outage can cause a 15 second
outage in am MPLS protected network which can cause a
45 second outage in the customer LMR router protected
network
Outage time extension in router protected network can be
significant
The outage time experienced by the customer can be two
to three orders of magnitude greater than the radio path
outage time
For router networks, number of outages can be more
significant than individual outage duration
11
IP Network Characteristics
Ring and Mesh networks can provide significantly enhanced
network performance with compared to the performance of
cascaded paths
In the State of Mississippi, the routers switched hundreds
of times a week, but outages due to path propagation were
much less frequent
The weakness of rings is when multiple paths are affected
simultaneously
Spurs without route protection are much more vulnerable to
outage time magnification
12
Path Multipath Performance
Spur performance was limited by cumulative
outage duration but also by individual outage
duration
Both durations were measured.
13
Path Multipath Performance
estimated
observed
Cumulative Outage Duration
14
Multipath Fading
Individual path performance varied widely:
Worst Case Observed Outage / Predicted Outage (secs ratio) = 156
Best Case Observed Outage / Predicted Outage (secs ratio) = 0
Median (50%, typical) Observed / Predicted Ratio = 1.9
Average Observed / Predicted Ratio = 6.4
Ratio Not Exceeded for 33% of Paths = 1
15
IP Network Characteristics
Individual Path Annual Performance Varied Widely
Average Total Path Outage Duration = 114 seconds
Minimum Number of Path Outages = 0
Maximum Number of Path Outages = 659
Median Number of Path Outages = 27
Average Number of Path Outages = 55
16
Path Multipath Performance
Cumulative Outage Events
17
Path Multipath Performance
Individual Outage Duration was relatively stable
Above statistics are for 6 GHz paths; 11 GHz paths experienced no outages.
Number of outage events  Total outage seconds
One reason measure outage > predicted outage is
that Vigant’s method predicts outage on an analog
basis while we measure it on an error-second basis
18
Path Multipath Performance
Cumulative Outage Duration
Cumulative outage duration is directly related to
Vigants C factor
Good: 0.5
Average = 1
Difficult = 2
Vigants, “Space Diversity Engineering,” Bell System Technical Journal, pp. 103-142, January 1975
19
Path Multipath Performance
Cumulative Outage Duration
Vigants (1975) C Factor = 2
Measured C Factor = 7
This is consistent with Vigants’ and Barnett’s latter
suggestions for the Gulf coast.
They suggested (1992) the C factor should be 10 for
this area.
Loso, Inserra, Brockel, Barnett and Vigants, “U. S. Army Tactical LOS Radio Propagation
Reliability,” Proceedings of the IEEE Tactical Communications Conference, pp. 109-117, 1992.
20
Path Multipath Performance
Cumulative path outage varied significantly from
the average
R = measured cumulative individual path outage – average outage
 = cumulative path outage standard deviation
10 log () = 7 dB (measured)
10 log () = 10 dB (reported by Bellcore*)
*Achariyapaopan, “A Model of Geographic Variation of Multipath Fading Probability,”
Bellcore National Radio engineer’s Conference Proceedings, pp. TA1 – TA16, 1986. 21
Path Multipath Performance
Individual Outage Duration was relatively stable
95 % of the paths have cumulative outage durations of the
following:
8.2 x network average duration (measured 10 log () = 7)
16.5 x network average duration (Bellcore 10 log () = 10)
Since the typical outage duration  one second, 95% of the
number of outage events average the following:
8.2 x average number of outages (measured 10 log () = 7)
16.5 x network average number of outages(Bellcore 10 log () = 10)
For router networks, the number of outage events
is more significant than the actual outage duration
22
Path Obstruction Fading
Obstruction fading was relatively infrequent but
could be quite troublesome when it occurred
23
Typical Multipath Fading
24
Classical Obstruction Fading
RSLs for Both Receivers
(Space Diversity)
Threshold RSL
Transmit Power
25
Nominal RSL
Slow Speed Obstruction Fading
RSLs for Both Receivers
(Hot Standby Non-diversity)
RSLs for Both Receivers
(Hot Standby Non-diversity)
26
Moderate Speed Obstruction Fading
RSLs for Both Receivers (hot standby)
27
High Speed Obstruction Fading
RSLs for Both Receivers
(space diversity)
Receiver
Overload
28
Receiver
Overload
Receiver
Overload
Obstruction Fading
Unexpected diversity effect
29
Different Paths from the Same Node
Unexpected path diversity effect
30
Path Pairs from Same Node
Path diversity effect
31
Obstruction Fading Predicted vs Measured Outages
Currently airport refractivity measurements are taken
at different times (typically at sun up and sun down)
than when obstruction fading typically occurs (from
about midnight to just before sun rise)
32
Obstruction Fading Lessons Learned
Obstruction fading comes in two forms:
Relatively short duration amenable to space diversity
Relatively long duration indifferent to space diversity
Obstruction fading very localized:
Geographically close paths can be de-correlated
Obstruction Fading estimation unreliable:
Results are unpredictable
Models under estimate outages on some paths
33
What Causes Obstruction Fading
Radio Wave Propagation is a Function of Atmospheric Refractivity
= index of refraction  1.000319
= refractivity = (n – 1) 106
= dry component + wet component
dry component = [ 77.6 p ] / [ 273 + T ]
wet component = [3.73 x 105 eS HR ] / [ 273 + T ]2
p
= atmospheric pressure in millibars
= 1.33 (pressure in mm of mercury)
= 33.9 (pressure in inches of mercury)
T
= temperature in degrees Centigrade
HR
= relative humidity (percent) / 100
n
N
N is a weak function of p
Refractivity increases as HR increases or T decreases
See G. Kizer, Digital Microwave Communication,
Chapter 12, pages 461 to 463.
34
Optical Wave Propagation
Refraction
The electromagnetic wave is
reflected or refracted depending
upon the angle of incidence
Reflection
If refracted, the electromagnetic
wave bends toward region of
higher refractivity
35
Radio Wave Propagation
Radio Wave Propagation is a Function of Atmospheric Refractivity
Without wind or rain atmospheric refractivity a function of weather
When radio wave is inside a slab of high refractivity air, wave bends
down (moves toward region of higher refractivity)
When radio wave encounters a nearby slab of significantly different
refractivity air, wave is reflected if angle of incidence is less than
Brewster’s angle (<< 10 degrees)
36
Obstruction Fading as a Function of Height
Reflective
fading for a
path above
a high
refractivity
layer
Obstruction
fading for a
path inside
a high
refractivity
layer
Layer acts like a large lake with slowly moving waves
Above a layer of different refractivity air, the radio wave is
subjected to reflective (interference) fading
Within a layer of different refractivity air, the radio wave is
subjected significant bending (earth bulge) or trapping (ducting)
37
October 24th Early Morning Outages
38
October 27th Early Morning Outages
39
Weather History
October 24, 2013
Outages Occurred at Times of No Wind, Low Temperature and High Humidity
These conditions are conducive to formation of a dense atmospheric layer
Low Temperature
High Relative Humidity
No Wind
(Static Atmosphere)
Example of Jackson, Mississippi for day of October 24th
(see www.wunderground.com/history)
See local airport for more exact weather.
40
Weather History
October 25, 2013
A typical day had no path propagation outages
Example of Jackson, Mississippi for day of October 25th
(see www.wunderground.com/history)
See local airport for more exact weather.
41
Weather History
October 27, 2013
Outages Occurred at Times of No Wind, Low Temperature and High Humidity
These conditions are conducive to formation of a dense atmospheric layer
Low Temperature
High Relative Humidity
No Wind
(Static Atmosphere)
Example of Jackson, Mississippi for day of October 27th
(see www.wunderground.com/history)
See local airport for more exact weather.
42
Weather History
For a graphical depiction of unusual air flow, go to www.wunderground.com/history.
Select Jackson, MS, Oct 27, View, and then select View Animated Radar Loop.
Watch the cold moist Gulf air come inland in early morning and then blown out later that day.
43
Obstruction Fading
If reflective fading occurs during periods of unusual
weather, we are dealing with “abnormal propagation”
that cannot be predicted accurately.
This fading can significantly expand the fading
season (summer for normal multipath, fall and winter
for abnormal weather layering).
Expect path to have history of outages significantly
longer than one second.
44
Typical Remedies for Obstruction Fading:
Decrease the distance between sites
The effect of distance reduction cannot be predicted
accurately
Increase the antenna heights
The placement of antennas cannot be predicted accurately
without radiosonde measurements from nearby airports
Convert to space diversity
Improvement typically moderate
Convert to adaptive modulation
Improvement can be dramatic
Turn linear routes into rings
Weather anomalies must be localized
Powerful with tall antennas (moderate uncorrelated outages)
45
Lessons Learned
Linear router based paths will experience longer
outages than the underlying microwave paths
Error extensive can be as much as two orders of magnitude
Actual path performance may be significantly
different than estimates
Use architecture (rings or meshes) to tame path
performance.
System performance can significantly exceed the performance of
cascaded paths.
46
Remember
Availability estimates and clearance
guidelines are no guarantee of path
performance !
> Use architecture to overcome propagation surprises <
47
48