This the final report of Avans University of Applied Sciences WLAN optimization phase 1. Avans and 7signal completed the first phase of optimization. This example report summarizes all five rounds of changes and shows results from the project. As a saparate items, there are also PoC report and itnerim optimization report available.

1. About this material
• This is an example of 7signal optimization project and associated reporting
• After baseline collection, 7signal provides a detailed optimization plan
• Based on the this plan, the first changes are implemented. Network
configuration is changed one change at the time. Impact of each change is
analyzed. The following change is decided after impact of previous change
is properly understood
• This is the final report of optimization phase 1 and summarizes
impacts of changes implemented in five rounds over about three
weeks time. Results are presented here as an example
• After the optimization, network performance is proactively maintained and
managed based on the SLAs, performance alarms and QoE trending
• This material is shared with a generous permission from Avans University of
Applied Sciences (Netherland) Network Administrator Kees Pronk.
1

2. Avans University of
Applied Sciences
Optimization report,
Phase 1 with five rounds of changes completed
June 20, 2013

3. Background
• Avans University of Applied Sciences Learning Center Network
optimization was performed for two floors
• Areas were covered with two Sapphire Eye units
• Number of users in network reduces over the period
– Lectures ended June 7th, however exams continued during the optimization
work
– Covered area is Learning center, so students remained there as active
users
– Data volumes in the covered area remained even or even increased slightly
despite lower number of users
• This report outlines the overall results in both areas and bands on daily
level. There are hourly and AP level dynamics that are not presented in
detail here.
• Each daily result point includes significant amount of data.
• Separate detailed reports are available for each phase.
3

4. Optimization flow
Implemented changes
• Change 1, May 29th
– 4-channel plan on 2.4 GHz (Learning Center floors 0, 1, 2)
• Change 2, June 3rd
– Beacon interval from 100ms to 300ms
– Remove 802.11b support and disable MCS 0, 8 and 16
• Change 3, June 6th
– Adjust AP radio power levels
• Change 4, June 11th
– Complete campus 4-channel plan on 2.4 GHz
– Add more channels on 5 GHz (Learning Center)
• Change 5, June 16th
– At 5 GHz, turn on HT40+
– At 2.4 GHz Balance client counts by adjusting AP power level settings
4

5. Managed Access Points on Learning Center
2.4 GHz Channels
Ground Floor
1st Floor
- AP000621
- AP000627
- AP000631
- AP000633
- AP001076
- AP001077
- AP000632
- AP000636
- AP001078
- AP001080
- AP001094
#13
#5
#13
#9
#1
#5
#9
#1
#5
#13
#1
5

6. 2.4 GHZ OVERALL RESULTS
6

7. Radio attach success rate
• Performance remained roughly on the same level
• Note: 7signal limits attach time to 30s. If attach is not completed during that time, it is
recoded as failure in metrics. End user may get connected by waiting longer.
1 2 3 4 5
Some APs refused
connections, rebooted
7

8. FTP (TCP) downlink throughput
• 40% improvement in ground floor, from about 21 Mbit/s to 30 Mbit/s
1 2 3 4 5 ~ 40% improvement
8

9. FTP (TCP) downlink throughput, hourly
• Clear improvement in the lowest hourly values
• The last change impacts negatively and need to be revisited
1 2 3 4 5
Clear improvement.
The last change
negative
9

10. FTP (TCP) uplink throughput
• 40% improvement in ground floor, from about 22 Mbit/s to 30 Mbit/s
1 2 3 4 5
~ 40% improvement
10

11. FTP (TCP) uplink throughput, hourly
• Clear improvement in the lowest hourly values
• The last change impacts negatively and need to be revisited
1 2 3 4 5
Clear improvement.
The last change
negative
11

12. HTTP downlink throughput
• About 25% improvement
1 2 3 4 5 ~ 25% improvement
12

13. Voice Quality (MOS), downlink
• Some improvement over the period
1 2 3 4 5
13

14. Voice Quality (MOS), downlink, hourly
• Some improvement in the lowest hourly values
• The last change impacts negatively and need to be revisited
1 2 3 4 5
14

15. Voice Quality (MOS), uplink
• Some improvement over the period
1 2 3 4 5
15

16. Voice Quality (MOS), uplink, hourly
• No clear change
1 2 3 4 5
16

17. Jitter
• A clear improvement
1 2 3 4 5 ~ 40% improvement
Last change in floor 1
had negative impact
17

18. Packet loss
• 80% reduction in level (1.5% to 0.3%)
1 2 3 4 5
80% reduction,
from 1.5% to 0.3%
18

19. Network latency (Ping RTT)
• 50% improvement in the ground floor from 7.25ms to 3.5ms,
• The last change has some what negative impact in 1st floor
1 2 3 4 5
Negative impact with
the last changes on
averages
50% improvement
19

20. Ping success rate
• Some improvement
1 2 3 4 5
20

21. Retransmissions in tests by Eye sensor
• 50% improvement in ground floor
• 10% improvement in 1st floor
1 2 3 4 5
50% improvement
Negative impact with
the last change
21

22. Retransmission by AP towards clients
• 45% improvement in ground floor
1 2 3 4 5
45% improvement
22

23. Retransmission by AP towards clients
• High peak hour values remain
1 2 3 4 5
Target is below 10%
23

24. Retransmissions by Clients toward AP
• Improvement in both areas
1 2 3 4 5
24

25. Retransmissions by Clients toward AP
• High peak hour values remain
1 2 3 4 5
Target is below 10%
25

26. Beacon airtime utilization
• Significant impact in both areas
1 2 3 4 5
Significant positive
impact
26

27. WLAN traffic airtime utilization
• 55% reduction in ground floor, a clear improvement in 1st floor too
• 90% lower WLAN utilization in empty network
1 2 3 4 5
In ground floor
55% reduction
Empty network
during weekends
27

28. Management traffic density from AP
1 2 3 4 5
Significant positive
impact
28

29. Aggregate traffic density towards AP
• Frame density lowers
1 2 3 4 5
29

30. Aggregate traffic density towards AP
• Frame density lowers
1 2 3 4 5
30

31. Uplink payload data volume
• Payload volume varies quite a bit between days
• May have increased. Short baseline time makes assessment a bit inaccurate
1 2 3 4 5
31

32. Downlink payload data volume
• Payload volume varies quite a bit between days
• May have increased. Short baseline time makes assessment a bit inaccurate
1 2 3 4 5
32

33. Number of clients/AP
• Client count lowers over the period
• At the same time traffic volume seem to hold up
1 2 3 4 5
33

34. 5 GHZ OVERALL RESULTS
34

35. Radio attach success rate
• Significant improvement
• Note: 7signal limits attach time to 30s. If attach is not completed during that time, it is
recoded as failure in metrics. End user may get connected by waiting longer.
1 2 3 4 5
From 40% to >90%
35

36. FTP (TCP) downlink throughput
• Significant improvements, 70-270%
1 2 3 4 5
1 2 3 4 5
270% improvement in
1st floor
70% improvement in
ground floor
36

37. FTP (TCP) downlink throughput, hourly
• Significant improvements
1 2 3 4 5
1 2 3 4 5
900% improvement in
1st floor
100% improvement in
ground floor
37

38. FTP (TCP) uplink throughput
• Significant improvements, 85-700%
1 2 3 4 5
700% improvement in
1st floor
85% improvement in
ground floor
38

39. FTP (TCP) uplink throughput, hourly
• Significant improvements
1 2 3 4 5
1 2 3 4 5
900% improvement in
1st floor
100% improvement in
ground floor
39

40. HTTP downlink throughput
• >90% in 1st floor floor and 50% in ground floor
1 2 3 4 5
90%/50%
improvements
40

41. Voice Quality (MOS), downlink
• Improving
1 2 3 4 5
41

42. Voice Quality (MOS), downlink, hourly
• Clearly improving
1 2 3 4 5 +0.25MOS in ground
+0.25MOS in 1st floor
42

43. Voice Quality (MOS), uplink
• Improving
1 2 3 4 5
43

44. Voice Quality (MOS), uplink, hourly
• Clearly improving
1 2 3 4 5 +0.15MOS in ground
+0.8MOS in 1st floor
No drops
44

45. Jitter
• Looks great all the time
1 2 3 4 5
45

46. Packet loss
• Improved
1 2 3 4 5
46

47. Network latency (RTT)
• 50% improvement in the 1st floor
• 25% improvement in ground floor
1 2 3 4 5
50% improvement in
1st floor
47

48. Ping success rate
• Significant improvement in the 1st floor
1 2 3 4 5
From 85% to 99% in
1st floor
48

49. Retransmission in Eye tests
• 85% improvement in the 1st floor
• 50% improvement in the ground floor
1 2 3 4 5
85% improvement in
1st floor
50% improvement
ground floor
49

50. Retransmission by AP towards clients
• Good general improvement
• Turning on HT40 however increases significantly retries
1 2 3 4 5
Lowered by 25% until
HT40 is turned on at
step 5
HT40 increases
significantly
retransmissions from
AP
50

51. Retransmission by AP towards clients, hourly
• High peak hour values remain
1 2 3 4 5
High peak hours
values remain
Target is below 10%
51

52. Retransmission by Clients toward AP
• Good general improvement
• Turning on HT40 however increases significantly retries
1 2 3 4 5
Lowered by 50% until
HT40 is turned on at
step 5
HT40 increases
significantly
retransmissions from
clients
52

53. Retransmission by Clients toward AP
• High peak hour values remain
1 2 3 4 5
High peak hours
values remain
Target is below 10%
53

54. Beacon airtime utilization
• Unnecessary beaconing does not consume any more airtime
1 2 3 4 5
54

55. WLAN traffic airtime utilization
• Overall air time utilization is 50% lower
• Idle network bandwidth consumption minimized
1 2 3 4 5
50% reduction in
airtime utilization
Empty network
during weekends
55

56. Management traffic density from AP
• Management traffic density has been reduced significantly
1 2 3 4 5
56

57. Aggregate traffic density towards AP
• Aggregate frame density towards APs is about 30% lower
1 2 3 4 5
57

58. Aggregate traffic density towards AP
• Aggregate frame density from APs is about 40% lower
1 2 3 4 5
58

59. Uplink payload data volume
• Traffic volume seems not to decrease even though there are less clients
• Short baseline makes assessment difficult
1 2 3 4 5
59

60. Downlink payload data volume
• Traffic volume seems not to decrease even though there are less clients
• Short baseline makes assessment difficult
1 2 3 4 5
60

61. Number of clients/AP
• Number of clients lowers over the period
1 2 3 4 5
61

62. NETWORK LEVEL
TRAFFIC VOLUME
62

63. Avg. number of clients/AP, avg. of all areas and bands
• Number clients/AP was reduced at the time the lectures ended
Lectures ended
Learning center
continued in use
63

64. Downlink data volume, avg. of all areas and bands
• Downlink data volume remained roughly on the previous level
Lectures ended
Learning center
continued in use
64

65. Uplink data volume, avg. of all areas and bands
• Uplink data volume remained roughly on the previous level or even increased
Lectures ended
Learning center
continued in use
65

66. SLA
66

67. 2.4 GHz, Learning center, ground, 8am-6pm
• A clear improvement in SLA metrics
• Beacon availability slightly degraded but in very
good level
• 2.4 GHz network would not be good enough for
voice traffic
BEFORE
AFTER
Improved
areas indicated
with black
67

68. 2.4 GHz, Learning center, 1st floor, 8am-6pm
• Some improvement
• 2.4 GHz network would not be good enough for
voice traffic
BEFORE
AFTER
Improved
areas indicated
with black
68

69. 5 GHz, Learning center, ground, 8am-6pm
BEFORE
AFTER
Improved
areas indicated
with black
• Clear improvement in SLA metrics
• Beacon availability slightly degraded but in very
good level
69

70. 5 GHz, Learning center, 1st floor, 8am-6pm
BEFORE
AFTER
• A significant improvement in SLA metrics
• Beacon availability slightly degraded but in very
good level
Improved
areas indicated
with black
70

71. SUMMARY
71

72. Observations
• Very good improvements were achieved in 5 GHz (1st floor) and 2.4 GHz
(ground floor). Ground floor 5 GHz network also improved significantly.
• 1st floor 2.4 GHz network improved but is still lacking capacity still due to
large amount of clients. Some APs have still too many clients to serve them
optimally. APs need to be added and some re-optimization done to increase
capacity to required level.
• The last change in 2.4 GHz 1st floor had some negative impact. AP power
level changes steered too many clients to another AP that became overload
instead of original one.
• Data volumes are likely to increase when network provides faster speeds. It
may be a good idea to limit throughputs/client at 2.4 GHz to share available
capacity more evenly and motivate users moving to 5 GHz band.
• At times there are APs that seem to start degrading by themselves or stop
accepting associations.
• Performance requires continuous attention, number of devices and traffic
volumes are likely to continue grow remarkably.
72

73. Key impacts in daily averages
Key Performance Indicator Ground
2.4 GHz
1st
2.4 GHz
Ground
5 GHz
1st floor
5 GHz
FTP (TCP) downlink throughput +40% 0% +70% +270%
FTP (TCP) uplink throughput +40% 5%** +85% +700%
HTTP downlink throughput +25% +25% +50% +90%
VoIP MOS downlink (MOS) +0.1 +0.1 +0.05 +0.05
VoIP MOS uplink (MOS) +0.1 +0.1 0 +0.07
UDP jitter -40% -40% 0% 0%
UDP packet loss -80% -80% >+200%* -90%
Latency -50% -10%* -25% -50%
AP retransmissions -45% -10% +40%*** +40%***
Client retransmissions -20% -20% -15% -15%
WLAN air time utilization -55% -55% -50% -50%
*=Packet loss increased clearly. Was at 0.3% at the end **= Last change had negative impact. Need to be revisited
***=AP retransmissions increased clearly after enabling HT40 NOTE: Network load varied over the period also
73

74. Want to see how your network performs?
To agree a demo or trial, contact 7signal
www.7signal.com
info@7signal.com
74
Trial, fast and easy!
- All software runs in a laptop provided by 7signal
- Eye sensor is temporarily installed in the area of interest
- Data is collected automatically
- You can access the data as you like
- 7signal provides a trial report with observations