1. Cooperative technique to optimize total allocation time in heterogeneous domains in an Optical Access Network
NTT Access Network Service Systems Laboratories
Shunsuke Kanai*, Akira Inoue and Hiroshi Uno
Thank you, Mr.Takano. Good morning, everyone.
Today, I will introduce a technique to optimize total allocation time in heterogeneous domains in an optical access network.
2019/1/2

2. Outline 1. OAN configuration and Characteristics in I-ODN
2. Current issues and goal
3. Identifying causes and proposed technique
I’ll first talk about the OAN configuration and introduce its characteristics in an I-ODN.
I’ll then identify current issues and our goal.
Finally, I’ll identify causes and propose a new cooperative technique.

3. OAN(Optical access network)
Overview
OAN(Optical access network)
Intra-office
CR1
J-CR
SE(PON)
PKG:1
PKG:1
ONU
PKG:2
CR2
PKG:3
Splitter
An OAN has three equipment categories that do administration.
The first is an outside optical distribution network (O-ODN) that administers between the ONU at the customer’s home and intra-office. The second is the service equipment　(SE) that provides optical service. The last is an intra-office optical distribution network (I-ODN) that administers optical fibers or racks between the cable rack and SE. Each equipment category has domains as OSS. We defined these OSSs for administration and allocation as O-ODN-D, SE-D, and I-ODN-D. These domains support the allocation of fiber routes in the OAN.
There are three steps in the process to allocate fiber route in an OAN. The first allocates the fiber route from the ONU in the customer’s home to a port in the CR through the O-ODN-D. The second allocates the PKG to provide optical services through the SE-D. The third allocates the fiber route from a port in the CR to the PKG in the SE through the I-ODN-D.
We made several assumptions regarding the OSS. A splitter has to be installed in the Junction-CR for the passive optical network service. Installing a splitter in the J-CR is an optimum configuration that can allocate fibers immediately. Therefore, the conventional cooperative technique for allocation assumes that a splitter has been installed in the J-CR.
However, such an assumption is only based on reducing the total allocation time. In particular, I-ODNs decrease cost efficiency if there is an emphasis on the allocation policies that reduce total allocation time. Therefore, we have to analyze whether these assumptions can be adopted at any time to solve this trade-off between cost efficiency and allocation time.
ONU
O-ODN-D
I-ODN-D
SE-D

4. Comparison-1 Cost Analysis Limited demand Medium demand SE CR1 CR2
PKG:1
PKG:2 SE
PKG:3
CR1
CR2
J-CR1
J-CR1 SE
CR1
PKG:1
Type-1
CR2
Splitter
Splitter
J-CR1
PKG:1
PKG:2 SE
PKG:3
PKG:4
CR1
CR2
J-CR1 SE
This diagram shows two types of configuration that have had their splitters installed in different positions.The splitter installed in the J-CR is Type 1, while the splitter installed in the CR is Type 2.
When demand is limited, Type 1 does not need to construct many fibers between the CR and J-CR. However, when demand is medium, this type requires numerous fibers between the CR and J-CR since the PON service shares a fiber between the PKG and the input port of the splitter, and this fiber is distributed by the splitter. Therefore, more fibers are required between the output port of the splitter and the CR.
Conversely, When demand is limited, Type 2 needs more splitters and PKGs than Type 1, because it is necessary to install splitters for each CR and PKG in the SE. There is a low rate of splitter use for this type because Type 1 can only connect one splitter in the J-CR, but Type 2 has to connect two splitters in each CR. However, the number of fibers between the CR and J-CR does not increase more than Type 1.
As a result, when demand increases, Type 1 needs more fibers than Type 2 and the Type 2 needs more splitters and PKGs installed than the Type 1.
CR1
PKG:1
Type-2
PKG:2
CR2

5. Comparison-2 Analysis of total allocation time Limited demand
J-CR
CR1
CR2
CR3
Medium demand
PKG:1
PKG:2 SE
PKG:3
J-CR SE
CR1
PKG:1
Type-1
CR2
CR3
Port occurred demand
Fault
Success
J-CR SE
J-CR
CR2
CR3
CR1
PKG:1
PKG:2 SE
PKG:3
This diagram also shows the two types with different installation positions for the splitter.
Type 1 can finish allocating in only one database retrieval even if demand occurs in any CR. Moreover, it can maintain the total allocation time even if demand increases. Because it can be connected between the output port of the splitter and the CR as the splitter is installed in the J-CR ..
Conversely, Type 2 increases the total allocation time with increasing demand. This is because the SE-D allocates PKGs from the smallest number. The SE needs to give priority for use and allocation from the smallest number of PKGs to improve cost efficiency. An PKG is allocated by the SE-D and the fiber between the CR and PKG is allocated by the I-ODN-D. Therefore, Type 2 needs many allocation and sending and receiving processes in each domain if allocation fails many times. Moreover, such a long total allocation time is sustained after increasing demand. In the other words, the rate of success in allocation depends on where the CR has been allocated by demand.
As a result, when demand increases, Type 1 can finish allocating in a minimum time, but Type 2 needs more time.
CR1
PKG:1
Type-2
CR2
PKG:2
PKG:3
CR3
Port occurred demand
Fault
Success

6. Comparison-3 Comparison of results [Comparison of construction costs]
2000
5000
Type 1
Type 2
Type 1
Type 2
1500
4000
Total allocation time
3000
1000
Total cost
2000
500
1000
These graphs have the results of comparison.
This compares construction costs against the number of demands for Types 1 and 2.
We extracted the amount of equipment for each demand. We then calculated the total cost using the unit cost for each piece of equipment.
Type 1 is cost effective in the early stages. However, after this, cost efficiency decreases with increasing demand because its fiber costs increase more than for Type 2.
Although Type 2 needs higher initial costs, its cost efficiency is better with increasing demand and it can maintain lower fiber costs.
This shows the results of total allocation time against the number of demands for Types 1 and 2.
We found that allocations in Type 1 can be successfully done by installing a splitter in the J-CR even if demand occurs in any CR. Conversely, We found that allocations in Type 2 needs numerous database retrievals with increasing demand.
As a result, when demand increases, Type 1 has low cost efficiency, but its total allocation time is very short. In contrast, Type 2 has high cost efficiency and its total allocation time is very long.
100
200
300
400
500
100
200
300
400
500
Number of demands
Number of demands
[Comparison of construction costs]
[Comparison of total allocation time]

7. Current issues and goal
Trade-off between Types 1 and 2
1000
2000
3000
4000
5000
100
200
300
400
500
Number of demands
Total allocation time
short
long
Type 1
Type 2
Cost
Time
low
long
Type 2
short
Type 1
We therefore found that there was a trade-off between Types 1 and 2 in terms of cost and total allocation time.
We established a goal to resolve these issues. We can see that Type 1 has a problem where many fibers are needed when demand increases. To solve this, we had to find out how to reduce fiber costs. We therefore analyzed new cooperative techniques in heterogeneous domains to reduce both the total allocation time for Type 2 when PON demand increases.
high
short

8. Identifying causes-1 … … … … Start Decision Most Least Fault Success
O-ODN-D
I-ODN-D
SE-D
Decision
Reject
PKG:2
PKG:P …
Most …
PKG:M
Least
PKG:1
ONU
CR:P
CR:P
Splitter
Finish
Accept
CR1
J-CR SE
Now, Let’s look at how we established a technique to reduce total allocation time.
This shows the conventional cooperative technique in heterogeneous domains for allocation in a PON service.
There are some processes of allocation.
First, the process of allocation begins and the order is sent to SE-D, O-ODN-D.
Second, PKG is allocated in the SE-D, and CR is allocated in the O-ODN-D.
Third, the both results of allocation are sent to the I-ODN-D
Fourth, a fiber route (CR <-> Splitter <-> J-CR <-> PKG) is allocated in the I-ODN-D
Fifth, when allocation fails, the order of reallocation is received in the SE-D from the I-ODN-D. The next candidate is allocated in the SE-D
Last, when allocation has succeeded, finish allocating
The conventional method is based on the assumption that the splitter has been installed in the J-CR. We therefore decided to establish one PKG candidate to allocate once in the SE. However, although this had the least allocations, the send and receive information on the PKG candidate had to be transmitted many times to allocate if allocation failed many times.
CR2 …
CR:P …
CR:M
Fault
Success

9. Identifying causes-2 … … … … Start Most Decision Least Fault Success
ONU
CR:P
PKG:1
PKG:2 …
PKG:M
PKG:P
O-ODN-D
I-ODN-D
SE-D …
PKG:M
Most
Finish
Accept
Decision
CR:P
Splitter
Least
CR1
J-CR SE
Conversely, if the number of allocation candidates was set to the number of all PKGs installed in the SE, the number of times that send and receive information on the PKG candidate or the order to reallocate were both minimized because they were sent only once. However, many allocations were required, since all PKGs were assigned additional PKGs if allocation succeeded in the Pth number of candidates.
As a result, our solution to the problem is two fold. The first is how many times send and receive information was transmitted, and the second is how many PKGs were allocated.
CR2 …
CR:P …
CR:M
Fault
Success

10. Number of times to send & receive Number of times to allocate for PKGs
Identifying causes-3
Number of times to send & receive
　Number of times to allocate for PKGs
Times
Most
Least
Most
Least x
Optimum number of candidates 1
All PKGs
[Number of candidates for one allocation in SE]
There was a trade-off in the number of allocation candidates between one and all. When we adopted one PKG as the candidate, it caused an increase in the number of times send and receive information was transmitted. Conversely, when we adopted all PKGs as candidates, it caused an increase in the number of PKGs that were allocated.
This means that there is an optimum number of candidates that minimizes the total allocation time in heterogeneous domains.
We analyzed a new cooperative technique, taking the minimization of total allocation time into consideration.
PKG candidates =ONE
PKG candidates =ALL
Number of times to send & receive
Number of times to allocate for PKGs

11. Proposed technique … P … I-ODN-D SE-D N N Success Fault
Order of allocation
Receive
Send
Allocation
Receive
Fault
Success CR
Allocation N …
PKG P
Send
Receive
Send & receive (1st) …
Allocation
Send
Receive
Send & receive ( P^ th) N
PKG
Send
Finish
：time to send and receive once (NW) 　 ：time to allocate once (I-ODN-D) ：time to allocate once (SE-D)
：time to send and receive once (I-ODN-D) ：time to send and receive once (SE-D)
Allocation time
Send & receive time
We defined the allocation procedure in heterogeneous domains.
The order of allocation is transmitted from a remote computer with an operator via the NW to the SE-D, holding the number of candidates for allocation once.
When allocation is completed in the SE-D, the PKG candidates are transmitted to the I-ODN-D via the NW.
When allocation fails in the I-ODN-D, the order of re-allocation is transmitted to the SE-D and executed.
Such processes are continued until allocation is successful in the I-ODN-D.
When this occurs, information on successful allocation is transmitted to the remote computer via the NW from the I-ODN-D.
Total allocation time consists of allocation time and send and receive time.
To calculate the total allocation time, we need to identify the allocation times (P) and send and receive times (P^) in heterogeneous domains.
P is the number of average allocation times from allocation start to finish in the I-ODN-D. It is possible to calculate P from expectations of probability and these can be calculated with PKG data. P^ is the number of transmissions (send and receive) in heterogeneous domains.
The total allocation time in the I-ODN-D, NW and SE-D as β,α and γ can be calculated using P and P^ and the number of candidates (N) .
Therefore, the total allocation time in the OSS can be calculated with T(N). In other words, we can identify the optimum number of candidates (N) that minimize T(N) in heterogeneous domains.
As a result, it is possible to improve the efficiency of the new cooperative technique in heterogeneous domains. This shortens the allocation, sending, and receiving times through an optimum number (N).
Total allocation time

12. First simulation Total allocation time CR PKG ：2 ：10 ：5 ：1 ：1 J-CR SE
：2 ： ：5 ：1 ：1 １
J-CR ２ SE
n1=9 ・・ ・・ 9
n1=Number of CRs with Splitter
100
T(N)＝α＋β＋γ
Total allocation time
We simulated the effect the new cooperative technique had.
In the first simulation, we could specify the number of candidates that minimized total allocation time, when the number of allocation candidates (N) changed sequentially within a specific equipment configuration. In the second simulation, we could specify the total allocation time for demand.
This graph shows the transition in total allocation time using the new cooperative technique from N=1 to N=9. We can see that the number of candidates that minimizes total allocation time is N= 5. 75
Optimum number of candidates 50 1 2 3 4 5 6 7 8 9
Number of candidates (N)

13. Second simulation Total allocation time Number of demands 120000
Type 2 (Conventional method)
Type 2 (Proposed method)
29% Down
80000
Total allocation time
40000
Our goal was to shorten the total allocation time for Type 2. We therefore proposed a new cooperative technique in heterogeneous domains.
In this simulation, We calculated the total allocation time for each demand for each type. We identified the total allocation time for each type by referring to the results of first simulation,
This graph shows the results of evaluating the proposed method. After comparing the transition for the “Proposed Method” and “Conventional Method”, we can see that the proposed method could allocate 29% faster than the conventional method.
100
200
300
400
500
Number of demands

14. Conclusion and future work
Cost efficiency in an I-ODN improves by positioning the splitter at installation with increasing demand for PON service.
Total allocation time can be reduced by changing the number of PKG candidates
Future Work
Improve efficiency of allocation algorithm in heterogeneous domains
Here, we emphasized the need for a cooperative technique in heterogeneous domains that enables quick and cost-effective allocations. The problem with the conventional cooperative technique has not been taken into consideration which enables improved cost efficiency in I-ODN. We described a cooperative technique that satisfies this need. It identifies the number of optimum candidates by using the allocation times (P) and the number of transmissions (P^).
We clarified the effect the proposed cooperative technique had through simulation. We found that it reduced the total allocation time by 29% compared with the conventional cooperative technique. As a result, it enabled us to deliver quick and cost-effective services.
In future work, we intend to increase the efficiency of the allocation algorithm in heterogeneous domains, and reduce the allocation time there.
Thank you very much for your kind attention.

15. Cooperative technique to optimize total allocation time in heterogeneous domains in an Optical Access Network
NTT Access Network Service Systems Laboratories
Shunsuke Kanai*, Akira Inoue and Hiroshi Uno
2019/1/2

16. …. …. Reference Cost Analysis O-ODN-D I-ODN-D Service Type=A SE-D1
Service Type=B
SE-D3
Service Type=C
This diagram is shown two types that differs installation position for splitter.
We defined splitter is installed in CR is Type 1, conversely splitter is installed in J-CR is Type 2.
When demand is limited, Type 1 is not necessary to construct many fibers between the CR and J-CR. However, when demand reaches medium, this type requires numerous fibers between the CR and J-CR since the PON service shares a fiber between the PKG and the input port of the splitter, and this fiber is distributed by the splitter. Therefore, more fibers are required between the output port of the splitter and the CR.
When demand is limited, Type 2 needs more splitters and PKGs than Type 1, because it is necessary to install splitters for each CR and PKG in the SE with increasing demand. The rate of splitter use is low for this type. When demand occurs each CR (CR1 and CR2), for Type 1 can only connect one splitter in the J-CR for each demand, but Type 2 has to connect two splitters in each CR for each demand. However, as a splitter has been installed in each CR, the number of fibers between the CR and J-CR does not increase more than it does for Type 1.
As a result, when demand increases, Type 1 needs more fibers than Type 2 and the latter needs more splitters and PKGs installed than the former. …. ….