1
AS Consumption Patterns
Geoff Huston
APNIC
May 2005
2
AS Numbers
• The 16 bit AS number field in BGP has 64,510 available values to use in the Internet’s public routing space
• Some 38,000 AS numbers have already been assigned by the RIRs
• This number space will be exhausted at some point in the future
3
32 Bit AS Numbers
• Use a 32 bit field for this value• draft-ietf-idr-as4bytes-09.txt describes how• This is proposed for publication as an Experimental RFC
Has been in this state for some years. Unclear whether it has stalled or just moving very slowly through the IETF standards process.
4
The Issue – Transition Planning• At some point we will need to start testing
various transition plans and vendor implementations, set up a new AS number registry, and commence deployment of these extended length protocol objects in BGP
5
When?
• Before we run completely out of 16 bit AS numbers• Need to allow a lead time for testing, deployment
of 4-byte AS BGP implementations and development of appropriate transition arrangements
• Allow some 3-4 years to undertake this smoothly
• So we’d like to know when we have 4 years to go before we run out of AS numbers
6
When?• A number of views can be used to make
forward projections:• The growth of the number of announced AS’s in the BGP routing table
• The rate at which AS number blocks are passed from IANA to the RIRs
• The rate at which RIRs undertake assignments of As’s to LIRs and end users
7
The BGP Routing Table: Announced AS’s
Advertised AS count
0
5,000
10,000
15,000
20,000
25,000
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
Date
Co
un
t
8
The BGP Routing Table: Growth Projections
AS Projections from BGP data
0
8,192
16,384
24,576
32,768
40,960
49,152
57,344
65,536
19
96
19
98
20
00
20
02
20
04
20
06
20
08
20
10
20
12
20
14
20
16
20
18
20
20
20
22
20
24
20
26
Date
AS
Co
un
t
Advertised Exp Projection Linear Projection
Sept 2025July 2014
9
IANA AS block allocations to RIRs
From the IANA AS number Registry
IANA Allocations
0
5,000
10,000
15,000
20,000
25,000
30,000
35,000
40,00019
92
1994
1996
1998
2000
2002
2004
2006
Date
Co
un
t
10
IANA AS Allocation Projection
IANA Allocation Projections
0
8,192
16,384
24,576
32,768
40,960
49,152
57,344
65,536
19
92
19
94
19
96
19
98
20
00
20
02
20
04
20
06
20
08
20
10
20
12
20
14
20
16
IANA Allocations Exp Projection Linear Projection
July 2014June 2011
11
RIR Assignments
From the RIR stats reports
RIR Assignments
0
5,000
10,000
15,000
20,000
25,000
30,000
35,00019
92
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Date
Co
un
t
There are some inconsistencies in the data
12
RIR Allocation ProjectionRIR Data Projections
0
8,192
16,384
24,576
32,768
40,960
49,152
57,344
65,5361
99
2
19
94
19
96
19
98
20
00
20
02
20
04
20
06
20
08
20
10
20
12
20
14
Date
Co
un
t
Assignments Exp-Projection Linear Projection
Aug 2014Sept 2010
13
Combining these viewsAS Projections
0
8,192
16,384
24,576
32,768
40,960
49,152
57,344
65,53619
92
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
Date
Co
un
t
14
Combined View + Differences
AS Status
0
5,000
10,000
15,000
20,000
25,000
30,000
35,000
40,00019
92
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Date
Co
un
t
IANA RIR BGP RIR Pool UnAdvertised
15
Observations• RIRs operate with an allocation buffer of an average of
5,000 numbers
• 12,348 AS numbers (39% of the assigned AS numbers) are not announced in the BGP table.• Is this the result of old AS assignments falling into disuse? • Or recent AS assignments being hoarded?• This pool creates uncertainty in AS number pool exhaustion predictions
16
UnAdvertised and Advertised ASs
Advertised vs UnAdvertised
0
5,000
10,000
15,000
20,000
25,0001
99
6
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
Date
Co
un
t
Advertised UnAdvertised
17
UnAdvertised : Advertised AS’s
Unadvertised / Advertised Ratio
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
Date
Rat
io
18
Trend: UnAdvertised : Advertised Ratio
UnAdvertised / Advertised Ratio Projection
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
20
06
20
07
20
08
20
09
20
10
20
11
20
12
Date
Rat
io
19
UnAdvertised / Advertised Distribution by Date
Unadvertised and Advertised ASes
0
100
200
300
400
500
600
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Allocation Date
Co
un
t
Unadv Adv
20
Distribution by AS Number Range
AS Status - April 2005
0
32
64
96
128
160
192
224
256
0
40
96
81
92
12
28
8
16
38
4
20
48
0
24
57
6
28
67
2
32
76
8
36
86
4
40
96
0
45
05
6
49
15
2
53
24
8
57
34
4
61
44
0
AS Block (256)
ADV UNADV RIR IANA IETF
21
UnAdvertised / Advertised Relative Proportion by Date
Unadvertised ASs (% of Allocated) by Date
0
10
20
30
40
50
60
70
80
90
100
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Allocation Date
Per
cen
t
22
Observations• AS numbers age out and disappear
• 5% attrition rate per year
• Old (low) AS number ranges have the highest unannounced / announced ratios
• Recent assignments take some 4 months to be advertised• LIR staging point factors
• Projections of AS number consumption should include a factor for Unadvertised / Advertised ratio that has a linear best fit (negative slope)
23
Selecting a Best Fit to the Data
• A Linear growth model will have a constant first order differential
• An exponential growth model will have a linear best fit to the log of the data
• The data set for the best fit is to a smoothed (moving average) time series of the cumulative sum of RIR AS allocations
24
Linear Model Fit
Allocations
23,000
24,000
25,000
26,000
27,000
28,000
29,000
30,000
31,000
32,000
2003 2003.5 2004 2004.5 2005
Date
Allo
cati
on
s
25
Linear Model FitAS Allocations - 1st order differential
4
6
8
10
12
14
16
18
20
03
20
04
20
05
Date
AS
's p
er d
ay
26
Exponential Model FitAllocations
23,000
24,000
25,000
26,000
27,000
28,000
29,000
30,000
31,000
32,000
2003 2003.5 2004 2004.5 2005
Date
Allo
cati
on
s
27
Exponential Model Fit
1st Order Differential of Log(Allocations)
1
2
3
4
5
6
7
2003 2003.5 2004 2004.5 2005 2005.5
Date
d lo
g(a
lloca
tio
ns)
/dt
28
Combining Allocation and Advertised AS Data Projections
AS Comsumption Projections
0
8192
16384
24576
32768
40960
49152
57344
65536
19
92
19
94
19
96
19
98
20
00
20
02
20
04
20
06
20
08
20
10
20
12
20
14
20
16
20
18
20
20
20
22
20
24
20
26
Date
Assigned Advertised Unadvertised Linear-Alloc Linear-Adv Linear-UnAdv Exp-Alloc Exp-Adv Exp-UnAdv
29
Current AS Use Projections• The available AS number pool will exhaust in the timeframe of 2010
- 2020 if current AS use trends continue
2010• No significant reclamation in old AS number space• No coordinated effort to increase utilization density of AS numbers• Increasing consumption trend• Most likely outcome as a best fit to post 2003 consumption data
2020• Extensive reuse of idle / unused AS numbers• High use rate maintained• Assuming a continued constant rate of demand
30
Planning Considerations (again)
• Need to allow a lead time for testing, deployment of 4-byte AS BGP implementations and development and testing of appropriate transition arrangements
• Allow some 3-4 years to undertake this smoothly
• So we’d like to know when we have around 4 years to go before we run out of AS numbers
• In the most likely consumption projection that advance planning date looks like being in 2006
• In a more tempered growth environment this may extend out to 2010