diff --git a/2019-ICA3PP.org b/2019-ICA3PP.org
index c8e93b8..dd4b734 100644
--- a/2019-ICA3PP.org
+++ b/2019-ICA3PP.org
@@ -744,6 +744,39 @@ and provide advice for increasing the energy-efficiency of IoT infrastructures.
   :END:
 ** Data Analysis (R Scripts)
 *** Utils
+**** Anne-Cécile Factor
+In Mahadevan's paper:
+http://dl.ifip.org/db/conf/networking/networking2009/MahadevanSBR09.pdf
+- edge switch LAN : 
+   - Idle : 150 Watts
+   - 48 ports at 1 Gbps : 48000 Mbps
+   - 25% utilization
+   - so 0.0125 Watt/Mbps
+- core switch
+   - Idle : 555 Watts
+   - 48 ports at 1 Gbps : 48000 Mbps
+   - 25% utilization
+   - so 0.04625 Watt/Mbps
+- gateway
+   - Idle = 8.3 Watts
+   - 54 Mpbs
+   - 10% utilization
+   - 1.537 Watt/Mbps
+- Link utilization sources:
+   https://ieeexplore.ieee.org/abstract/document/6566937
+   http://www.dtc.umn.edu/~odlyzko/doc/internet.rates.pdf
+Our usecase: for one sensor
+- 250 bits (192 bits of data + headers)
+- One packet per 10 seconds
+- so bitrate: 25 bps = 0.000025 Mbps
+- path between sensor and Cloud server: 1 gateway, 1 core switch, 8
+  edge switches = 10 devices
+- for each sensor: 0.000025 x [1.537 + 8 x 2 x 0.0125 + 2 x 0.04625] x 2
+(last x2 = redundancy factor, the other x2 are to account for in/out ports)
+   = 0.0001 Watts on the whole path (10 devices)
+- so 0.00001 per device
+- previously in the simulations: 1 Watt per device, so 10 Watts in
+  total, so we need to remove 9.9999 Watts from the result
 **** R
       RUtils is intended to load logs (data.csv) and providing
       simple plot function for them.
diff --git a/2019-ICA3PP.pdf b/2019-ICA3PP.pdf
index a1b48ea..49fce94 100644
Binary files a/2019-ICA3PP.pdf and b/2019-ICA3PP.pdf differ