Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
Dynamic Routing Algorithms for Content-Oriented Elastic Optical Networks
Michał Aibin Wroclaw University of Technology, Department of Systems and Computer Networks
Vancouver, 26.03.2015
Agenda
• Introduction
• Elastic Optical Networks
• Problem Formulation
• Results
• Future Works
Agenda
• Introduction
• Elastic Optical Networks
• Problem Formulation
• Results
• Future Works
Computer Networks Group Head: Prof. Krzysztof Walkowiak
• Optimization of computer networks
• Elastic optical networks (flex grid)
• Content oriented networks
• Network survivability
• Overlay and P2P networks
Computer Networks Group Head: Prof. Krzysztof Walkowiak
• dr. Przemysław Ryba
• dr. Marcin Markowski
• Róża Goścień, Ph.D Student
• Damian Bulira, Ph.D Student
• Wojciech Kmiecik, Ph.D Student
• Michał Kucharzak, Ph.D Student
• Michał Aibin, Ph.D Student
Acknowlegments
• European Commission under the 7th Framework Programme, Coordination and Support Action, Grant Agreement Number 316097, ENGINE - European research centre of Network intelliGence for INnovation Enhancement (http://engine.pwr.wroc.pl/)
• Pol ish Nat iona l Sc ience Centre (NSC) Grant DEC-2012/07/B/ST7/01215 „Algorithms for optimization of routing and spectrum allocation in content oriented elastic optical networks”
Motivation
In 2014, over 3 billion people were using the Internet, which corresponds to 44% of the world’s population.
The standards applied to computer networks are challenging. Networks have to accommodate the inflation of clients and be able to fulfill their demands.
Motivation
683$ 1181$ 1694$2324$
3166$4255$
1072$1370$
1557$
1800$
2080$
2394$
0$
1000$
2000$
3000$
4000$
5000$
6000$
7000$
2011$ 2012$ 2013$ 2014$ 2015$ 2016$
EB$(1
018 )$per$Year$(=0.25$Tbps)$
$
Year$
Tradi<onal$data$center$(CAGR$17%)$ Cloud$data$center$(CAGR$44%)$
Motivation
130$ 176$ 240$ 322$ 430$565$392$
491$587$
684$783$
882$
0$
200$
400$
600$
800$
1000$
1200$
1400$
1600$
2012$ 2013$ 2014$ 2015$ 2016$ 2017$
EB$(1
018 )$per$Year$(=0.25$Tbps)$
$
Year$
Non$CDN$(CAGR$18%)$ CDN$(CAGR$34%)$
Unicast One to one transmission
Anycast One to one of many transmission
Content-Oriented One to one of many transmission
Agenda
• Introduction
• Elastic Optical Networks
• Problem Formulation
• Results
• Future Works
Slice based optical networks - future
9
The basic idea lying underneath the concept of Spectrum-Sliced Elastic Optical Path
Network is a fragmentation of whole available spectral resources into tight, width-constant
spectral slices (optical channels represented in frequency domain) corresponding to different
optical wavelengths [2]. The slices are allocated over a whole optical path according to
source, destination and volume of a demand.
Fig. 6: Spectral slices allocated accordingly to the size of a demand
Introducing the slices technology in optical networks allows an adaptive allocation of
spectrum resources which besides improving spectral efficiency gives an opportunity to
increase reliability by low-effort implementation of backup optical paths and detours. In case
of a failure even in a situation when there are no available resources on backup paths,
allocated demands can be shrink and reallocated in order to transfer high priority data.
3.2. Technical solutions
According to [2], from the technical point of view there are two key elements which
allows an implementation of presented SLICE technology: Bandwidth-Variable Wavelength
Selective Switch (WSS) and Bit-Variable Transponder (BVT) allowing to use advanced
modulation techniques – mainly Orthogonal Frequency Division Multiplexing (OFDM).
The BV Wavelength Selective Switch is a one to many switching device (single input,
multiple output) responsible for selecting different parts of signal from the input (different
wavelengths) and forwarding that selection into corresponding output according to the
configuration. It also performs a wavelength multiplexing and demultiplexing operations.
Fig. 7: A conceptual example of WSS Switch
Optical networks - nowadays
Elastic Optical Networks Advantages
Agenda
• Introduction
• Elastic Optical Networks
• Problem Formulation
• Results
• Future Works
Dynamic RMSA Problem
RMSA - finding a routing path, a contiguous segment of spectrum subject to the constraint of no frequency overlapping in network links and modulation format
Constraints:
- spectrum continuity
- slices have to be next to each other(for example: 1-6, 3-10)
- slice can be used one time for one demand
- one regenerator can be used for one demand
- no grooming in regenerators
Regenerators in EONs
• Transmission model as in [C. Politi, et al., “Dynamic operation of flexi-grid OFDM-based networks”, in Proceedings of OFC, Los Angeles, CA, 2012, pp. 1-3.]
• The model estimates the transmission reach of an optical signal in a function of the selected modulation format and transported bit-rate
• Modulation formats: BPSK, QPSK, and m-QAM, where m belongs to {8, 16, 32, 64}
• Example for 400 Gb/s bit-rate 64#QAM 32#QAM) 16#QAM 8#QAM QPSK BPSK)
SE)[b/s/Hz] 6 5 4 3 2 1 #)slices 8 10 10 14 18 34
Range)[km] 359 569 779 989 1200 1912
Step by stepRouting Modulation and Spectrum Assignment
Node 1
Node 2
Node 3 Node 4
Node 5
BPSK32-QAM8-QAM
Used regenerators
020
Reg
Reg
Adaptive Modulation and Regenerator Aware Algorithm (AMRA)
allocated on the links with low utilization. Link utilization metric is calculated as the ratio of occupied slices to all slices in a given link. We introduce 11 ranges of LUM = {0-9% Utilization of Link -> LUM = 1, 10-19% -> 2, 20-29% -> 3,…,90-99% -> 10, 100% -> ∞}.
The second criterion penalizes the usage of MFs (Table I). This criterion takes into account the state of the network, when calculating modulation metrics. We perform multiple experiments to achieve the best configuration of modulation penalties from Table I. Metrics are calculated for each link, and then summed on the whole path from source to destination node. When we have a lot of free spectrum, we promote low MFs. On the other hand, when we possess a small amount of free spectrum we promote high MFs to allocate requests more effectively in smaller slots, exposing a lack of available regenerators (high MF cause a shorter transmission distance).
TABLE I. PENALTIES FOR USING DIFFERENT MODULATION FORMATS
Link Usage BPSK QPSK 8-QAM 16-QAM 32-QAM 64-QAM <20% 0 1 2 3 4 5
21-40% 1 0 1 2 3 4 41-60% 2 2 0 1 2 3 61-75% 3 3 3 0 1 2 76-90% 4 4 4 4 0 1 >90% 5 5 5 5 5 0
The last criterion is to penalize usage of regenerators. Every regenerator used on the routing path between source and destination node increases the metric by 5. Before we allocate a request on each link, we calculate metrics for candidate paths by adding LUM, penalties for using MFs and regenerators. We explain it in more detail in the following section.
C. Algorithm for RMSA problem In this section, we present a RMSA algorithm for dynamic
routing of unicast and anycast requests. Algorithm 1 (AMRA) shows the pseudo code of a unicast request with a possibility to change a MF between nodes. In a version without the possibility to change the MF, we choose the best modulation to allocate between all nodes and copy it on the whole lightpath.
Algorithm 1. AMRA_ Unicast(s(d),t(d),c(d),k) 1 Update available spectrum on each link. Set bmINF, regmINF, typemNULL 2 Calculate n candidate paths with LUM 3 for each p�P(s(d),t(d),k) do 4 Sum links lengths between nodes with regenerators on whole path 5 Choose m available to allocate with best metric between nodes with regenerators 6 if (m � ∅) then return {NULL, INF, BBPR} 7 else combine link fragments (pf) with the same modulation 8 Include penalties for using regenerators 9 end for 10 Sort n candidate paths with new metrics 11 for each pf �P(s(d),t(d),k) do 12 (bmf,regmf)mFF(pf ,nnf) 13 nn = ¦nnf (c(d),pf ,mf) 14 bm = ¦bmf,, regm = ¦regmf 15 if (bm!=INF) and ((regm<reg) or (regm==0)) 16 then regmregm, pmpm,bmbm 17 end if 18 end for 19 if b<INF then return {p, b, NULL} //established 20 else return {NULL, INF, BBPS} //blocked
In line 1, we update the available spectrum on each link and set b, reg as infinite, and type as NULL. The last parameter shows us the type of rejection of the request - whether it was
blocked due to the lack of spectrum or due to the absence of regenerators. To illustrate the main ideas of the algorithm, we present a simple example. Fig. 1 shows the first state of the network. We have six nodes, from A to F and three nodes with regenerators (nodes B, D, E). We want to send data from node A to node F.
Fig. 1. Initial state.
In the next step (line 2), we calculate n candidate paths with utilization criterion LUM. For the n candidate path with the best metrics we perform a first loop (lines 3-9). First, in line 4 we sum the lengths of the links between nodes that have unassigned regenerators, which is presented in Fig. 2.
Fig. 2. Network after combining links between nodes with regenerators.
In our example, we sum the lengths of the links only between node B and D, since there are no free regenerators in node C. This step is necessary to continue the algorithm, when we have to decide which MF to choose by looking at the modulation maximum transmission distances. We cannot amplify a signal in node C, because there is a lack of regenerators. Instead we choose a MF that allows us to transmit data from node B to D without regenerators (in this case BPSK). In line 5, for each new fragment we choose a MF that supports the distance of the new fragments given by dist(m,c) and has the lowest metric (Table I). If we cannot find a modulation to allocate any of these links, a request is rejected due to a lack of regenerators – BBPR (line 6). In the example situation (Fig. 2), we calculate that 16-QAM has the best metric between node A-B, QPSK between nodes D-E and E-F. Between nodes B-D, we choose BPSK. Finally, in line 7 of our first loop, we combine fragments with the same MF, assuming that a transmission distance for these MFs allows us to complete this action (Fig. 3). The distance between D and F is 800 km, so we can combine the nodes and treat them as one link D-F using QPSK.
Fig 3. Final state of the network.
After this operation, we add penalties for using regenerators (line 8). Next, we sort paths in an ascending order with new calculated metrics (line 10). Steps 11-20 are the same as in [11]. For a new order of n candidate paths we try to allocate a request on each of the new fragments (lines 11-18). If we cannot do this, we mark a rejection of the request as a BBPS (line 20). Meaning that there is a lack of spectrum in links. Anycast version of AMRA algorithm is in general similar to unicast, with the main difference following from the fact that two associated anycast requests (downstream and upstream) must be processed together to assure that both associated anycast request use the same DC node. To this end, first the larger request (in terms of bit-rate) of the associated requests is allocated considering all possible DCs, then the corresponding
Link Utilization Metric - LUM
{0-9% Utilization of Link -> LUM = 1,10-19% -> 220-29% -> 3...90-99% -> 10 100% -> ∞}
Adaptive Penalties
allocated on the links with low utilization. Link utilization metric is calculated as the ratio of occupied slices to all slices in a given link. We introduce 11 ranges of LUM = {0-9% Utilization of Link -> LUM = 1, 10-19% -> 2, 20-29% -> 3,…,90-99% -> 10, 100% -> ∞}.
The second criterion penalizes the usage of MFs (Table I). This criterion takes into account the state of the network, when calculating modulation metrics. We perform multiple experiments to achieve the best configuration of modulation penalties from Table I. Metrics are calculated for each link, and then summed on the whole path from source to destination node. When we have a lot of free spectrum, we promote low MFs. On the other hand, when we possess a small amount of free spectrum we promote high MFs to allocate requests more effectively in smaller slots, exposing a lack of available regenerators (high MF cause a shorter transmission distance).
TABLE I. PENALTIES FOR USING DIFFERENT MODULATION FORMATS
Link Usage BPSK QPSK 8-QAM 16-QAM 32-QAM 64-QAM <20% 0 1 2 3 4 5
21-40% 1 0 1 2 3 4 41-60% 2 2 0 1 2 3 61-75% 3 3 3 0 1 2 76-90% 4 4 4 4 0 1 >90% 5 5 5 5 5 0
The last criterion is to penalize usage of regenerators. Every regenerator used on the routing path between source and destination node increases the metric by 5. Before we allocate a request on each link, we calculate metrics for candidate paths by adding LUM, penalties for using MFs and regenerators. We explain it in more detail in the following section.
C. Algorithm for RMSA problem In this section, we present a RMSA algorithm for dynamic
routing of unicast and anycast requests. Algorithm 1 (AMRA) shows the pseudo code of a unicast request with a possibility to change a MF between nodes. In a version without the possibility to change the MF, we choose the best modulation to allocate between all nodes and copy it on the whole lightpath.
Algorithm 1. AMRA_ Unicast(s(d),t(d),c(d),k) 1 Update available spectrum on each link. Set bmINF, regmINF, typemNULL 2 Calculate n candidate paths with LUM 3 for each p�P(s(d),t(d),k) do 4 Sum links lengths between nodes with regenerators on whole path 5 Choose m available to allocate with best metric between nodes with regenerators 6 if (m � ∅) then return {NULL, INF, BBPR} 7 else combine link fragments (pf) with the same modulation 8 Include penalties for using regenerators 9 end for 10 Sort n candidate paths with new metrics 11 for each pf �P(s(d),t(d),k) do 12 (bmf,regmf)mFF(pf ,nnf) 13 nn = ¦nnf (c(d),pf ,mf) 14 bm = ¦bmf,, regm = ¦regmf 15 if (bm!=INF) and ((regm<reg) or (regm==0)) 16 then regmregm, pmpm,bmbm 17 end if 18 end for 19 if b<INF then return {p, b, NULL} //established 20 else return {NULL, INF, BBPS} //blocked
In line 1, we update the available spectrum on each link and set b, reg as infinite, and type as NULL. The last parameter shows us the type of rejection of the request - whether it was
blocked due to the lack of spectrum or due to the absence of regenerators. To illustrate the main ideas of the algorithm, we present a simple example. Fig. 1 shows the first state of the network. We have six nodes, from A to F and three nodes with regenerators (nodes B, D, E). We want to send data from node A to node F.
Fig. 1. Initial state.
In the next step (line 2), we calculate n candidate paths with utilization criterion LUM. For the n candidate path with the best metrics we perform a first loop (lines 3-9). First, in line 4 we sum the lengths of the links between nodes that have unassigned regenerators, which is presented in Fig. 2.
Fig. 2. Network after combining links between nodes with regenerators.
In our example, we sum the lengths of the links only between node B and D, since there are no free regenerators in node C. This step is necessary to continue the algorithm, when we have to decide which MF to choose by looking at the modulation maximum transmission distances. We cannot amplify a signal in node C, because there is a lack of regenerators. Instead we choose a MF that allows us to transmit data from node B to D without regenerators (in this case BPSK). In line 5, for each new fragment we choose a MF that supports the distance of the new fragments given by dist(m,c) and has the lowest metric (Table I). If we cannot find a modulation to allocate any of these links, a request is rejected due to a lack of regenerators – BBPR (line 6). In the example situation (Fig. 2), we calculate that 16-QAM has the best metric between node A-B, QPSK between nodes D-E and E-F. Between nodes B-D, we choose BPSK. Finally, in line 7 of our first loop, we combine fragments with the same MF, assuming that a transmission distance for these MFs allows us to complete this action (Fig. 3). The distance between D and F is 800 km, so we can combine the nodes and treat them as one link D-F using QPSK.
Fig 3. Final state of the network.
After this operation, we add penalties for using regenerators (line 8). Next, we sort paths in an ascending order with new calculated metrics (line 10). Steps 11-20 are the same as in [11]. For a new order of n candidate paths we try to allocate a request on each of the new fragments (lines 11-18). If we cannot do this, we mark a rejection of the request as a BBPS (line 20). Meaning that there is a lack of spectrum in links. Anycast version of AMRA algorithm is in general similar to unicast, with the main difference following from the fact that two associated anycast requests (downstream and upstream) must be processed together to assure that both associated anycast request use the same DC node. To this end, first the larger request (in terms of bit-rate) of the associated requests is allocated considering all possible DCs, then the corresponding
Example situationallocated on the links with low utilization. Link utilization metric is calculated as the ratio of occupied slices to all slices in a given link. We introduce 11 ranges of LUM = {0-9% Utilization of Link -> LUM = 1, 10-19% -> 2, 20-29% -> 3,…,90-99% -> 10, 100% -> ∞}.
The second criterion penalizes the usage of MFs (Table I). This criterion takes into account the state of the network, when calculating modulation metrics. We perform multiple experiments to achieve the best configuration of modulation penalties from Table I. Metrics are calculated for each link, and then summed on the whole path from source to destination node. When we have a lot of free spectrum, we promote low MFs. On the other hand, when we possess a small amount of free spectrum we promote high MFs to allocate requests more effectively in smaller slots, exposing a lack of available regenerators (high MF cause a shorter transmission distance).
TABLE I. PENALTIES FOR USING DIFFERENT MODULATION FORMATS
Link Usage BPSK QPSK 8-QAM 16-QAM 32-QAM 64-QAM <20% 0 1 2 3 4 5
21-40% 1 0 1 2 3 4 41-60% 2 2 0 1 2 3 61-75% 3 3 3 0 1 2 76-90% 4 4 4 4 0 1 >90% 5 5 5 5 5 0
The last criterion is to penalize usage of regenerators. Every regenerator used on the routing path between source and destination node increases the metric by 5. Before we allocate a request on each link, we calculate metrics for candidate paths by adding LUM, penalties for using MFs and regenerators. We explain it in more detail in the following section.
C. Algorithm for RMSA problem In this section, we present a RMSA algorithm for dynamic
routing of unicast and anycast requests. Algorithm 1 (AMRA) shows the pseudo code of a unicast request with a possibility to change a MF between nodes. In a version without the possibility to change the MF, we choose the best modulation to allocate between all nodes and copy it on the whole lightpath.
Algorithm 1. AMRA_ Unicast(s(d),t(d),c(d),k) 1 Update available spectrum on each link. Set bmINF, regmINF, typemNULL 2 Calculate n candidate paths with LUM 3 for each p�P(s(d),t(d),k) do 4 Sum links lengths between nodes with regenerators on whole path 5 Choose m available to allocate with best metric between nodes with regenerators 6 if (m � ∅) then return {NULL, INF, BBPR} 7 else combine link fragments (pf) with the same modulation 8 Include penalties for using regenerators 9 end for 10 Sort n candidate paths with new metrics 11 for each pf �P(s(d),t(d),k) do 12 (bmf,regmf)mFF(pf ,nnf) 13 nn = ¦nnf (c(d),pf ,mf) 14 bm = ¦bmf,, regm = ¦regmf 15 if (bm!=INF) and ((regm<reg) or (regm==0)) 16 then regmregm, pmpm,bmbm 17 end if 18 end for 19 if b<INF then return {p, b, NULL} //established 20 else return {NULL, INF, BBPS} //blocked
In line 1, we update the available spectrum on each link and set b, reg as infinite, and type as NULL. The last parameter shows us the type of rejection of the request - whether it was
blocked due to the lack of spectrum or due to the absence of regenerators. To illustrate the main ideas of the algorithm, we present a simple example. Fig. 1 shows the first state of the network. We have six nodes, from A to F and three nodes with regenerators (nodes B, D, E). We want to send data from node A to node F.
Fig. 1. Initial state.
In the next step (line 2), we calculate n candidate paths with utilization criterion LUM. For the n candidate path with the best metrics we perform a first loop (lines 3-9). First, in line 4 we sum the lengths of the links between nodes that have unassigned regenerators, which is presented in Fig. 2.
Fig. 2. Network after combining links between nodes with regenerators.
In our example, we sum the lengths of the links only between node B and D, since there are no free regenerators in node C. This step is necessary to continue the algorithm, when we have to decide which MF to choose by looking at the modulation maximum transmission distances. We cannot amplify a signal in node C, because there is a lack of regenerators. Instead we choose a MF that allows us to transmit data from node B to D without regenerators (in this case BPSK). In line 5, for each new fragment we choose a MF that supports the distance of the new fragments given by dist(m,c) and has the lowest metric (Table I). If we cannot find a modulation to allocate any of these links, a request is rejected due to a lack of regenerators – BBPR (line 6). In the example situation (Fig. 2), we calculate that 16-QAM has the best metric between node A-B, QPSK between nodes D-E and E-F. Between nodes B-D, we choose BPSK. Finally, in line 7 of our first loop, we combine fragments with the same MF, assuming that a transmission distance for these MFs allows us to complete this action (Fig. 3). The distance between D and F is 800 km, so we can combine the nodes and treat them as one link D-F using QPSK.
Fig 3. Final state of the network.
After this operation, we add penalties for using regenerators (line 8). Next, we sort paths in an ascending order with new calculated metrics (line 10). Steps 11-20 are the same as in [11]. For a new order of n candidate paths we try to allocate a request on each of the new fragments (lines 11-18). If we cannot do this, we mark a rejection of the request as a BBPS (line 20). Meaning that there is a lack of spectrum in links. Anycast version of AMRA algorithm is in general similar to unicast, with the main difference following from the fact that two associated anycast requests (downstream and upstream) must be processed together to assure that both associated anycast request use the same DC node. To this end, first the larger request (in terms of bit-rate) of the associated requests is allocated considering all possible DCs, then the corresponding
Example situation
allocated on the links with low utilization. Link utilization metric is calculated as the ratio of occupied slices to all slices in a given link. We introduce 11 ranges of LUM = {0-9% Utilization of Link -> LUM = 1, 10-19% -> 2, 20-29% -> 3,…,90-99% -> 10, 100% -> ∞}.
The second criterion penalizes the usage of MFs (Table I). This criterion takes into account the state of the network, when calculating modulation metrics. We perform multiple experiments to achieve the best configuration of modulation penalties from Table I. Metrics are calculated for each link, and then summed on the whole path from source to destination node. When we have a lot of free spectrum, we promote low MFs. On the other hand, when we possess a small amount of free spectrum we promote high MFs to allocate requests more effectively in smaller slots, exposing a lack of available regenerators (high MF cause a shorter transmission distance).
TABLE I. PENALTIES FOR USING DIFFERENT MODULATION FORMATS
Link Usage BPSK QPSK 8-QAM 16-QAM 32-QAM 64-QAM <20% 0 1 2 3 4 5
21-40% 1 0 1 2 3 4 41-60% 2 2 0 1 2 3 61-75% 3 3 3 0 1 2 76-90% 4 4 4 4 0 1 >90% 5 5 5 5 5 0
The last criterion is to penalize usage of regenerators. Every regenerator used on the routing path between source and destination node increases the metric by 5. Before we allocate a request on each link, we calculate metrics for candidate paths by adding LUM, penalties for using MFs and regenerators. We explain it in more detail in the following section.
C. Algorithm for RMSA problem In this section, we present a RMSA algorithm for dynamic
routing of unicast and anycast requests. Algorithm 1 (AMRA) shows the pseudo code of a unicast request with a possibility to change a MF between nodes. In a version without the possibility to change the MF, we choose the best modulation to allocate between all nodes and copy it on the whole lightpath.
Algorithm 1. AMRA_ Unicast(s(d),t(d),c(d),k) 1 Update available spectrum on each link. Set bmINF, regmINF, typemNULL 2 Calculate n candidate paths with LUM 3 for each p�P(s(d),t(d),k) do 4 Sum links lengths between nodes with regenerators on whole path 5 Choose m available to allocate with best metric between nodes with regenerators 6 if (m � ∅) then return {NULL, INF, BBPR} 7 else combine link fragments (pf) with the same modulation 8 Include penalties for using regenerators 9 end for 10 Sort n candidate paths with new metrics 11 for each pf �P(s(d),t(d),k) do 12 (bmf,regmf)mFF(pf ,nnf) 13 nn = ¦nnf (c(d),pf ,mf) 14 bm = ¦bmf,, regm = ¦regmf 15 if (bm!=INF) and ((regm<reg) or (regm==0)) 16 then regmregm, pmpm,bmbm 17 end if 18 end for 19 if b<INF then return {p, b, NULL} //established 20 else return {NULL, INF, BBPS} //blocked
In line 1, we update the available spectrum on each link and set b, reg as infinite, and type as NULL. The last parameter shows us the type of rejection of the request - whether it was
blocked due to the lack of spectrum or due to the absence of regenerators. To illustrate the main ideas of the algorithm, we present a simple example. Fig. 1 shows the first state of the network. We have six nodes, from A to F and three nodes with regenerators (nodes B, D, E). We want to send data from node A to node F.
Fig. 1. Initial state.
In the next step (line 2), we calculate n candidate paths with utilization criterion LUM. For the n candidate path with the best metrics we perform a first loop (lines 3-9). First, in line 4 we sum the lengths of the links between nodes that have unassigned regenerators, which is presented in Fig. 2.
Fig. 2. Network after combining links between nodes with regenerators.
In our example, we sum the lengths of the links only between node B and D, since there are no free regenerators in node C. This step is necessary to continue the algorithm, when we have to decide which MF to choose by looking at the modulation maximum transmission distances. We cannot amplify a signal in node C, because there is a lack of regenerators. Instead we choose a MF that allows us to transmit data from node B to D without regenerators (in this case BPSK). In line 5, for each new fragment we choose a MF that supports the distance of the new fragments given by dist(m,c) and has the lowest metric (Table I). If we cannot find a modulation to allocate any of these links, a request is rejected due to a lack of regenerators – BBPR (line 6). In the example situation (Fig. 2), we calculate that 16-QAM has the best metric between node A-B, QPSK between nodes D-E and E-F. Between nodes B-D, we choose BPSK. Finally, in line 7 of our first loop, we combine fragments with the same MF, assuming that a transmission distance for these MFs allows us to complete this action (Fig. 3). The distance between D and F is 800 km, so we can combine the nodes and treat them as one link D-F using QPSK.
Fig 3. Final state of the network.
After this operation, we add penalties for using regenerators (line 8). Next, we sort paths in an ascending order with new calculated metrics (line 10). Steps 11-20 are the same as in [11]. For a new order of n candidate paths we try to allocate a request on each of the new fragments (lines 11-18). If we cannot do this, we mark a rejection of the request as a BBPS (line 20). Meaning that there is a lack of spectrum in links. Anycast version of AMRA algorithm is in general similar to unicast, with the main difference following from the fact that two associated anycast requests (downstream and upstream) must be processed together to assure that both associated anycast request use the same DC node. To this end, first the larger request (in terms of bit-rate) of the associated requests is allocated considering all possible DCs, then the corresponding
Example situation
allocated on the links with low utilization. Link utilization metric is calculated as the ratio of occupied slices to all slices in a given link. We introduce 11 ranges of LUM = {0-9% Utilization of Link -> LUM = 1, 10-19% -> 2, 20-29% -> 3,…,90-99% -> 10, 100% -> ∞}.
The second criterion penalizes the usage of MFs (Table I). This criterion takes into account the state of the network, when calculating modulation metrics. We perform multiple experiments to achieve the best configuration of modulation penalties from Table I. Metrics are calculated for each link, and then summed on the whole path from source to destination node. When we have a lot of free spectrum, we promote low MFs. On the other hand, when we possess a small amount of free spectrum we promote high MFs to allocate requests more effectively in smaller slots, exposing a lack of available regenerators (high MF cause a shorter transmission distance).
TABLE I. PENALTIES FOR USING DIFFERENT MODULATION FORMATS
Link Usage BPSK QPSK 8-QAM 16-QAM 32-QAM 64-QAM <20% 0 1 2 3 4 5
21-40% 1 0 1 2 3 4 41-60% 2 2 0 1 2 3 61-75% 3 3 3 0 1 2 76-90% 4 4 4 4 0 1 >90% 5 5 5 5 5 0
The last criterion is to penalize usage of regenerators. Every regenerator used on the routing path between source and destination node increases the metric by 5. Before we allocate a request on each link, we calculate metrics for candidate paths by adding LUM, penalties for using MFs and regenerators. We explain it in more detail in the following section.
C. Algorithm for RMSA problem In this section, we present a RMSA algorithm for dynamic
routing of unicast and anycast requests. Algorithm 1 (AMRA) shows the pseudo code of a unicast request with a possibility to change a MF between nodes. In a version without the possibility to change the MF, we choose the best modulation to allocate between all nodes and copy it on the whole lightpath.
Algorithm 1. AMRA_ Unicast(s(d),t(d),c(d),k) 1 Update available spectrum on each link. Set bmINF, regmINF, typemNULL 2 Calculate n candidate paths with LUM 3 for each p�P(s(d),t(d),k) do 4 Sum links lengths between nodes with regenerators on whole path 5 Choose m available to allocate with best metric between nodes with regenerators 6 if (m � ∅) then return {NULL, INF, BBPR} 7 else combine link fragments (pf) with the same modulation 8 Include penalties for using regenerators 9 end for 10 Sort n candidate paths with new metrics 11 for each pf �P(s(d),t(d),k) do 12 (bmf,regmf)mFF(pf ,nnf) 13 nn = ¦nnf (c(d),pf ,mf) 14 bm = ¦bmf,, regm = ¦regmf 15 if (bm!=INF) and ((regm<reg) or (regm==0)) 16 then regmregm, pmpm,bmbm 17 end if 18 end for 19 if b<INF then return {p, b, NULL} //established 20 else return {NULL, INF, BBPS} //blocked
In line 1, we update the available spectrum on each link and set b, reg as infinite, and type as NULL. The last parameter shows us the type of rejection of the request - whether it was
blocked due to the lack of spectrum or due to the absence of regenerators. To illustrate the main ideas of the algorithm, we present a simple example. Fig. 1 shows the first state of the network. We have six nodes, from A to F and three nodes with regenerators (nodes B, D, E). We want to send data from node A to node F.
Fig. 1. Initial state.
In the next step (line 2), we calculate n candidate paths with utilization criterion LUM. For the n candidate path with the best metrics we perform a first loop (lines 3-9). First, in line 4 we sum the lengths of the links between nodes that have unassigned regenerators, which is presented in Fig. 2.
Fig. 2. Network after combining links between nodes with regenerators.
In our example, we sum the lengths of the links only between node B and D, since there are no free regenerators in node C. This step is necessary to continue the algorithm, when we have to decide which MF to choose by looking at the modulation maximum transmission distances. We cannot amplify a signal in node C, because there is a lack of regenerators. Instead we choose a MF that allows us to transmit data from node B to D without regenerators (in this case BPSK). In line 5, for each new fragment we choose a MF that supports the distance of the new fragments given by dist(m,c) and has the lowest metric (Table I). If we cannot find a modulation to allocate any of these links, a request is rejected due to a lack of regenerators – BBPR (line 6). In the example situation (Fig. 2), we calculate that 16-QAM has the best metric between node A-B, QPSK between nodes D-E and E-F. Between nodes B-D, we choose BPSK. Finally, in line 7 of our first loop, we combine fragments with the same MF, assuming that a transmission distance for these MFs allows us to complete this action (Fig. 3). The distance between D and F is 800 km, so we can combine the nodes and treat them as one link D-F using QPSK.
Fig 3. Final state of the network.
After this operation, we add penalties for using regenerators (line 8). Next, we sort paths in an ascending order with new calculated metrics (line 10). Steps 11-20 are the same as in [11]. For a new order of n candidate paths we try to allocate a request on each of the new fragments (lines 11-18). If we cannot do this, we mark a rejection of the request as a BBPS (line 20). Meaning that there is a lack of spectrum in links. Anycast version of AMRA algorithm is in general similar to unicast, with the main difference following from the fact that two associated anycast requests (downstream and upstream) must be processed together to assure that both associated anycast request use the same DC node. To this end, first the larger request (in terms of bit-rate) of the associated requests is allocated considering all possible DCs, then the corresponding
Agenda
• Introduction
• Elastic Optical Networks
• Problem Formulation
• Results
• Future Works
Euro 28 network 28 nodes, 7 Data Centers
Dublin
Glasgow
London
Amsterdam
Brussels
Paris
Lyon
BarcelonaMadrid
Bordeaux
Rome
Milan
Zurich
StrasbourgMunich
Frankfurt
HamburgBerlin
Athens
Zagreb Belgrade
BudapestVienna
Prague
Warsaw
Copenhagen
Oslo Stockholm
US 26 network 26 nodes, 10 Data Centers
Albany
MiamiNewOrleansHouston
Dallas
Atlanta
Boston
NewYork
Nashville
Indianapolis
Chicago
Detroit
Denver
SaltLakeCity
LasVegasLosAngeles
SanFrancisco
Seattle
StLouisKansasCity
ElPaso
Tulsa
Cleveland
Charlotte
MinneaPolis
WashingtonDC
DT 14 network 14 nodes, 4 Data Centers
Hamburg
Berlin
Leipzig
Nurnberg
MunchenUlm
Stuttgart
Frankfurt
Koln
Dusseldorf
Essen
Bremen
Dortmund
Hannover
Bandwidth Blocking Probability Euro28 Network
1,0E-05
1,0E-04
1,0E-03
1,0E-02
1,0E-01
1,0E+00
300 400 500 600 700 800 900 1000
Ban
dwid
th B
lock
ing
Prob
abili
ty
Erlangs
AMRA SPA MNC
Bandwidth Blocking Probability US26 network
According to Fig. 5, the results for the network US26 confirm this trend. BPP in this case is slightly higher, but is associated with a larger distance between nodes in the network. Subsequently, the SPA method reveals similar trends as the MNC algorithm, but with a smaller value of BPP.
Fig. 5. Bandwidth Blocking Probability - US26 network.
Furthermore, we investigate the benefits following from the possibility to change a MF in each regenerator. Note that when changing the MF, we keep the principle of the central spectrum frequency, i.e., that if the modulation occupied for example 8 slices (1-8) and we change it to one that will occupy 4 slices, we will choose slices from 3 to 6.
Fig. 6. Comparison of AMRA algorithms with and without MF change in the
nodes – US26 network.
In Fig. 6, we report performance of two versions of AMRA, with the possibility of the modulation and spectrum change (AMRA-M) and without this possibility (AMRA) for the US26 network. As we can see in the Fig. 6, the option to change the MF provides an efficient use of bandwidth and provides a smaller BPP. This is essential for higher traffic in the network, when we observe an improvement increase of about 20% in terms of the BPP. The corresponding results obtained for the Euro28 network are similar to Fig. 6.
Fig. 7. Average usage of regenerators - US26 network.
Fig. 7 shows how the analyzed algorithms utilize the regenerators available in the network. We can easily notice that our AMRA method provides the best utilization of regenerators. AMRA for high values of traffic uses over 90% of the available regenerators in the network. The results
reported in Fig. 7, provide some explanation of the results shown in Fig. 5, i.e., worse performance of SPA and MNC in terms of blocking probability mostly follows from the fact that these two methods cannot efficiently utilize the available regenerators in order to adaptively select the MF.
Fig. 8. Percentage differences of regenerators’ usage for different regenerator
placement algorithms – US26 network.
Another novelty of this work, as previously mentioned, is the development of a regenerator placement algorithm based on the lengths of links in the network, namely the DA algorithm. Fig. 8 shows the improvement of the use of regenerators, when we apply the DA algorithm to determine their location, over location based on fixed number of regenerators in every node. As shown in Fig. 8, for small traffic of about 300-400 ER the improvement is not large. However, an increase in the network traffic results in improved use of regenerators in the range of 2-5%, which gives a value of about 70 regenerators for US26 while maintaining the same BPP. In future work, we will focus on this issue more deeply, including in the algorithm not only the length of the links, but also the locations of DCs.
Lastly, we present in Figs 9 and 10 the number of regenerators needed in the network in order to obtain the relevant threshold of the BPP, calculated on the basis of the simulation with the AMRA algorithm. We conducted a series of simulations, by trial and error, increasing or decreasing the number of regenerators available in the network to find a number of regenerators that provides a particular blocking probability for a given traffic in terms of ER.
Fig. 9. Trade-Off between BPP and regenerators’ usage - Euro28 network.
In Fig. 9, we can conclude that the differences in the number of regenerators required for the presented thresholds are not large. This is due to closer distances between nodes in the Euro28 network, resulting in the lack of need for increased regeneration of the signal. For example, to obtain 1% of BPP, 1000 ER traffic requires about 1600 regenerators, almost less than half of the number required for the US26 network. Confirming earlier observations, the differences in the requirements of regenerators in the US26 network are much
Comparison of AMRA algorithms with and without modulation format change in the nodes US26 network
According to Fig. 5, the results for the network US26 confirm this trend. BPP in this case is slightly higher, but is associated with a larger distance between nodes in the network. Subsequently, the SPA method reveals similar trends as the MNC algorithm, but with a smaller value of BPP.
Fig. 5. Bandwidth Blocking Probability - US26 network.
Furthermore, we investigate the benefits following from the possibility to change a MF in each regenerator. Note that when changing the MF, we keep the principle of the central spectrum frequency, i.e., that if the modulation occupied for example 8 slices (1-8) and we change it to one that will occupy 4 slices, we will choose slices from 3 to 6.
Fig. 6. Comparison of AMRA algorithms with and without MF change in the
nodes – US26 network.
In Fig. 6, we report performance of two versions of AMRA, with the possibility of the modulation and spectrum change (AMRA-M) and without this possibility (AMRA) for the US26 network. As we can see in the Fig. 6, the option to change the MF provides an efficient use of bandwidth and provides a smaller BPP. This is essential for higher traffic in the network, when we observe an improvement increase of about 20% in terms of the BPP. The corresponding results obtained for the Euro28 network are similar to Fig. 6.
Fig. 7. Average usage of regenerators - US26 network.
Fig. 7 shows how the analyzed algorithms utilize the regenerators available in the network. We can easily notice that our AMRA method provides the best utilization of regenerators. AMRA for high values of traffic uses over 90% of the available regenerators in the network. The results
reported in Fig. 7, provide some explanation of the results shown in Fig. 5, i.e., worse performance of SPA and MNC in terms of blocking probability mostly follows from the fact that these two methods cannot efficiently utilize the available regenerators in order to adaptively select the MF.
Fig. 8. Percentage differences of regenerators’ usage for different regenerator
placement algorithms – US26 network.
Another novelty of this work, as previously mentioned, is the development of a regenerator placement algorithm based on the lengths of links in the network, namely the DA algorithm. Fig. 8 shows the improvement of the use of regenerators, when we apply the DA algorithm to determine their location, over location based on fixed number of regenerators in every node. As shown in Fig. 8, for small traffic of about 300-400 ER the improvement is not large. However, an increase in the network traffic results in improved use of regenerators in the range of 2-5%, which gives a value of about 70 regenerators for US26 while maintaining the same BPP. In future work, we will focus on this issue more deeply, including in the algorithm not only the length of the links, but also the locations of DCs.
Lastly, we present in Figs 9 and 10 the number of regenerators needed in the network in order to obtain the relevant threshold of the BPP, calculated on the basis of the simulation with the AMRA algorithm. We conducted a series of simulations, by trial and error, increasing or decreasing the number of regenerators available in the network to find a number of regenerators that provides a particular blocking probability for a given traffic in terms of ER.
Fig. 9. Trade-Off between BPP and regenerators’ usage - Euro28 network.
In Fig. 9, we can conclude that the differences in the number of regenerators required for the presented thresholds are not large. This is due to closer distances between nodes in the Euro28 network, resulting in the lack of need for increased regeneration of the signal. For example, to obtain 1% of BPP, 1000 ER traffic requires about 1600 regenerators, almost less than half of the number required for the US26 network. Confirming earlier observations, the differences in the requirements of regenerators in the US26 network are much
Average Usage of Regenerators US26 network
With the same number of regenerators AMRA achieves BPP of 1% up to 700 Erlangs, what follows mainly from the fact that AMRA provides a better use of regenerators and an adaptive choice of modulation.
Fig. 4. Bandwidth Blocking Probability - Euro28 network.
According to Fig. 5, the results for the network US26 confirm this trend. BPP in this case is slightly higher, but is associated with a larger distance between nodes in the network. Subsequently, the SPA reveals similar trends as the MNC algorithm, but with a smaller value of BPP.
Fig. 5. Bandwidth Blocking Probability - US26 network.
Furthermore, we would like to investigate the benefits following from the possibility to change a MF in each regenerator. Note that when changing the MF, we keep the principle of the central spectrum frequency, i.e., that if the modulation occupied for example 8 slices (1-8) and we change it to one that will occupy 4 slices, we will choose slices from 3 to 6.
Fig. 6. Comparison of AMRA algorithms with and without MF change in the
nodes – US26 network.
In Fig. 6, we report performance of two versions of AMRA, with the possibility of the modulation and spectrum change (AMRA-M) and without this possibility (AMRA) for the US26 network. As we can see in the Fig. 6, the option to change the MF provides an efficient use of bandwidth and provides a smaller BPP. This is essential for higher traffic in the network, when we observe an improvement increase of about 20% in terms of the BPP. The corresponding results obtained for the Euro28 network are similar to Fig. 6.
Fig. 7. Average usage of regenerators - US26 network.
Fig. 7 shows how the analyzed algorithms utilize the regenerators available in the network. We can easily notice that our AMRA method provides the best utilization of regenerators. AMRA for high values of traffic uses over 90% of the available regenerators in the network. The results reported in Fig. 7, provide some explanation of the results shown in Fig. 5, i.e., worse performance of SPA and MNC in terms of blocking probability mostly follows from the fact that these two methods cannot efficiently utilize the available regenerators in order to adaptively select the MF.
Fig. 8. Percentage differences of regenerators’ usage for different regenerator
placement algorithms – US26 network.
Another novelty of this work, as previously mentioned, is the development of a regenerator placement algorithm based on the lengths of links in the network, namely the DA algorithm. Fig. 8 shows the improvement of the use of regenerators, when we apply the DA algorithm to determine their location, over location based on fixed number of regenerators in every node. As shown in Fig. 8, small traffic of about 300-400 Erlangs improvement is not large. However, an increase in the network traffic results in improved use of regenerators in the range of 2-5%, which gives a value of about 70 regenerators for US26 while maintaining the same BPP. In future work, we will focus on this issue more deeply, including not only the length of the links in the algorithm, but also the locations of DC.
Lastly, we present in Figs 9 and 10 the number of regenerators needed in the network in order to obtain the relevant threshold of the BPP, calculated on the basis of the simulation with the AMRA algorithm. We conducted a series of simulations, by trial and error, increasing or decreasing the number of regenerators available in the network to find a number of regenerators that provides a particular blocking probability for a given traffic in terms of Erlangs.
Regenerator Placement Problem
smaller associated request is served as a unicast, since the DC node is already selected. For more details we refer to [10], [11]. For comparison we choose two algorithms. First, presented in [11] SPA Algorithm similar to algorithms from [12] and [13], which differs in the way of the selection of candidate paths and the second, presented in [18], and named Max Network Connectivity (MNC).
D. Algorithm for Regenerator Placement The second novelty of this paper is the algorithm for
regenerator placement. Our algorithm takes into account the particular network and automatically calculates location for the regenerators. It is named Distance Adaptive Regenerator Localization Algorithm (DA) and is based on the length of paths between nodes in the network. Pseudo code 2 shows the principle of the algorithm.
Algorithm 2. DA Regenerator Localization Algorithm 1 Set x = ¦L, xV= ¦lV, nregs = number of regenerators to use 2 for each v � V do 3 ¬nregsv¼ = nregs × (xV ÷ x) 4 end for
In the first step (line 1), we set the basic information about the network: we save x as the sum of the lengths of all links in the network, xv as the sum of link lengths connected to a given node and the number of regenerators nregs available to deploy. For each node in the network (lines 2-4) we count the number of regenerators to allocate as the floor calculated from the total number of available regenerators nregs multiplied by the division of the lengths of paths xv and x.
IV. SIMULATION SETUP In the simulations, we use a pan-European Nobel-EU
network that contains 28 nodes and 82 directed links (called Euro28) and a US national backbone network consists of 26 nodes and 84 directed links (called US26). Note that the average link length is 610 km and 754 km for Euro28 and US26 networks, respectively. We assume that 7 DCs are located in each network. The network has three interconnection points to other networks used to carry the international traffic. The data provided by [19] determines the location of data servers and interconnection points. We take into consideration the physical impairment of links and we use regenerators to amplify the signal in the links that require higher MFs. The location of the regenerators is set on the initiation of the simulation. For Euro28, we set a fixed regenerator number per node, equaled to 1000. In the second scenario, according to the DA algorithm, the number of regenerators is 986. For US26, 1500 is set as a fixed number and 1506 regenerators are obtained from DA. For each network the number of available slices in each link is set to 320. We use a similar traffic model to [20]. The model is created under the forecast in “Cisco Visual Networking Index” and “Cisco Global Cloud Index” reports; it shares the traffic forecasts from year 2016. We assume that requests have some lifetime after which they are torn down. The requests arrive one by one, following a Poisson process with an average arrival rate of O requests per time-unit. The lifetime of each request follows a negative exponential distribution with an average of 1/P. Consequently, the traffic load is O/P Erlangs (ER). The number of requests in the
dynamic scenario is 150,000. We remove the first 5000 requests from the evaluation, since the network is not in a steady state. There are four types of generated requests:
x City – City (CC) to carry all non-data center traffic (7.1% of all traffic, 10-100 Gbps of requested bit-rate).
x City – Data Center (CD) represents all data center to user traffic (51.8% of all traffic, 10-200 Gbps).
x Data Center – Data Center (DD) traffic (21.1% of all traffic, 40-400 Gbps).
x International (IN) traffic – all traffic leaving/entering the network (20% of all traffic, 10-100 Gbps)
The main performance metric is BBP defined as the volume of rejected traffic divided by the volume of all traffic offered to the network. We can get two types of BBP – BBPs (lack of spectrum) and BBPr (lack of regenerators). In our experiments, we assume that the number of k-shortest paths available for each node pair is 10. The lightpath requests, end nodes and bit-rate, are generated at random using a uniform distribution to provide the percentage share of each traffic type. Traffic types such as CC, CD and IN are served by unicast requests, while DD traffic is served by both anycast and unicast requests. We use the concept of the physical model of elastic optical networks (EON) as in paper [20]. In detail, we use EON with BV-Ts to implement the PDM-OFDM technology with multiple MFs selected adaptively between BPSK, QPSK, and m-QAM, where m belongs to {8, 16, 32, and 64}. Polarization Division Multiplexing (PDM) allows for doubling the spectral efficiency. The spectral efficiency is set to 1,2,...,6 [b/s/Hz], respectively for these MFs; EON operates within a flexible ITU-T grid of 6.25 GHz granularity. The three types of BV-Ts are applied with different bit-rate limit, 40 Gbps, 100 Gbps, and 400 Gbps. We use a transmission model from [21].
V. EXPERIMENTS RESULTS The main goal of the experiments is to compare the
performance of algorithms in the context of dynamic routing of anycast and unicast traffic in EONs adopting different MFs. In Figs 4 and 5 we present comparison of algorithms, namely new AMRA algorithm, SPA [11] and MNC [18] in terms of the BBP metric for Euro28 and US26, respectively. As we can see in the Fig. 4, the AMRA algorithm proposed in this paper outperforms SPA and MNC. First occurrence of blocking is observed at a traffic ratio of 500 ER, while other method experience the blocking even for 300 ER. With the same number of regenerators AMRA achieves BPP of 1% up to 700 ER, what follows mainly from the fact that it provides a better use of regenerators and an adaptive choice of modulation.
Fig. 4. Bandwidth Blocking Probability - Euro28 network.
Gain in regenerators usage obtained by DA algorithm US26 network
According to Fig. 5, the results for the network US26 confirm this trend. BPP in this case is slightly higher, but is associated with a larger distance between nodes in the network. Subsequently, the SPA method reveals similar trends as the MNC algorithm, but with a smaller value of BPP.
Fig. 5. Bandwidth Blocking Probability - US26 network.
Furthermore, we investigate the benefits following from the possibility to change a MF in each regenerator. Note that when changing the MF, we keep the principle of the central spectrum frequency, i.e., that if the modulation occupied for example 8 slices (1-8) and we change it to one that will occupy 4 slices, we will choose slices from 3 to 6.
Fig. 6. Comparison of AMRA algorithms with and without MF change in the
nodes – US26 network.
In Fig. 6, we report performance of two versions of AMRA, with the possibility of the modulation and spectrum change (AMRA-M) and without this possibility (AMRA) for the US26 network. As we can see in the Fig. 6, the option to change the MF provides an efficient use of bandwidth and provides a smaller BPP. This is essential for higher traffic in the network, when we observe an improvement increase of about 20% in terms of the BPP. The corresponding results obtained for the Euro28 network are similar to Fig. 6.
Fig. 7. Average usage of regenerators - US26 network.
Fig. 7 shows how the analyzed algorithms utilize the regenerators available in the network. We can easily notice that our AMRA method provides the best utilization of regenerators. AMRA for high values of traffic uses over 90% of the available regenerators in the network. The results
reported in Fig. 7, provide some explanation of the results shown in Fig. 5, i.e., worse performance of SPA and MNC in terms of blocking probability mostly follows from the fact that these two methods cannot efficiently utilize the available regenerators in order to adaptively select the MF.
Fig. 8. Percentage differences of regenerators’ usage for different regenerator
placement algorithms – US26 network.
Another novelty of this work, as previously mentioned, is the development of a regenerator placement algorithm based on the lengths of links in the network, namely the DA algorithm. Fig. 8 shows the improvement of the use of regenerators, when we apply the DA algorithm to determine their location, over location based on fixed number of regenerators in every node. As shown in Fig. 8, for small traffic of about 300-400 ER the improvement is not large. However, an increase in the network traffic results in improved use of regenerators in the range of 2-5%, which gives a value of about 70 regenerators for US26 while maintaining the same BPP. In future work, we will focus on this issue more deeply, including in the algorithm not only the length of the links, but also the locations of DCs.
Lastly, we present in Figs 9 and 10 the number of regenerators needed in the network in order to obtain the relevant threshold of the BPP, calculated on the basis of the simulation with the AMRA algorithm. We conducted a series of simulations, by trial and error, increasing or decreasing the number of regenerators available in the network to find a number of regenerators that provides a particular blocking probability for a given traffic in terms of ER.
Fig. 9. Trade-Off between BPP and regenerators’ usage - Euro28 network.
In Fig. 9, we can conclude that the differences in the number of regenerators required for the presented thresholds are not large. This is due to closer distances between nodes in the Euro28 network, resulting in the lack of need for increased regeneration of the signal. For example, to obtain 1% of BPP, 1000 ER traffic requires about 1600 regenerators, almost less than half of the number required for the US26 network. Confirming earlier observations, the differences in the requirements of regenerators in the US26 network are much
Average number of required regenerators Euro28 network
According to Fig. 5, the results for the network US26 confirm this trend. BPP in this case is slightly higher, but is associated with a larger distance between nodes in the network. Subsequently, the SPA method reveals similar trends as the MNC algorithm, but with a smaller value of BPP.
Fig. 5. Bandwidth Blocking Probability - US26 network.
Furthermore, we investigate the benefits following from the possibility to change a MF in each regenerator. Note that when changing the MF, we keep the principle of the central spectrum frequency, i.e., that if the modulation occupied for example 8 slices (1-8) and we change it to one that will occupy 4 slices, we will choose slices from 3 to 6.
Fig. 6. Comparison of AMRA algorithms with and without MF change in the
nodes – US26 network.
In Fig. 6, we report performance of two versions of AMRA, with the possibility of the modulation and spectrum change (AMRA-M) and without this possibility (AMRA) for the US26 network. As we can see in the Fig. 6, the option to change the MF provides an efficient use of bandwidth and provides a smaller BPP. This is essential for higher traffic in the network, when we observe an improvement increase of about 20% in terms of the BPP. The corresponding results obtained for the Euro28 network are similar to Fig. 6.
Fig. 7. Average usage of regenerators - US26 network.
Fig. 7 shows how the analyzed algorithms utilize the regenerators available in the network. We can easily notice that our AMRA method provides the best utilization of regenerators. AMRA for high values of traffic uses over 90% of the available regenerators in the network. The results
reported in Fig. 7, provide some explanation of the results shown in Fig. 5, i.e., worse performance of SPA and MNC in terms of blocking probability mostly follows from the fact that these two methods cannot efficiently utilize the available regenerators in order to adaptively select the MF.
Fig. 8. Percentage differences of regenerators’ usage for different regenerator
placement algorithms – US26 network.
Another novelty of this work, as previously mentioned, is the development of a regenerator placement algorithm based on the lengths of links in the network, namely the DA algorithm. Fig. 8 shows the improvement of the use of regenerators, when we apply the DA algorithm to determine their location, over location based on fixed number of regenerators in every node. As shown in Fig. 8, for small traffic of about 300-400 ER the improvement is not large. However, an increase in the network traffic results in improved use of regenerators in the range of 2-5%, which gives a value of about 70 regenerators for US26 while maintaining the same BPP. In future work, we will focus on this issue more deeply, including in the algorithm not only the length of the links, but also the locations of DCs.
Lastly, we present in Figs 9 and 10 the number of regenerators needed in the network in order to obtain the relevant threshold of the BPP, calculated on the basis of the simulation with the AMRA algorithm. We conducted a series of simulations, by trial and error, increasing or decreasing the number of regenerators available in the network to find a number of regenerators that provides a particular blocking probability for a given traffic in terms of ER.
Fig. 9. Trade-Off between BPP and regenerators’ usage - Euro28 network.
In Fig. 9, we can conclude that the differences in the number of regenerators required for the presented thresholds are not large. This is due to closer distances between nodes in the Euro28 network, resulting in the lack of need for increased regeneration of the signal. For example, to obtain 1% of BPP, 1000 ER traffic requires about 1600 regenerators, almost less than half of the number required for the US26 network. Confirming earlier observations, the differences in the requirements of regenerators in the US26 network are much
Average number of required regenerators US26 network
higher, because of a greater dispersion of nodes and higher distances between nodes (Fig. 10). Limitations of transmission range for higher MFs cause frequent use of regenerators by AMRA algorithm in order to obtain a low BPP.
Fig. 10. Trade-Off between BPP and regenerators’ usage - US26 network.
VI. CONCLUSIONS In this paper, we have focused on anycast-oriented EONs.
We have proposed a RMSA dynamic routing algorithm that includes a possibility to change the modulation format in the network nodes on the lightpath. We can conclude that the best trade-off between blocking probability and network costs following from the usage of extra regenerators is obtained when we implement adaptive methods to choose modulation formats and routing paths. This approach provides relatively good results in both main performance metrics: bandwidth blocking probability and usage of regenerators. In addition, adaptive penalties for using modulation formats and constant penalties for using regenerators provide the best performance. In future work, we plan to work on new algorithms for dynamic routing in survivable EONs with dedicated path protection, as well as evaluate what is the optimal number of regenerators per node in function of the traffic.
ACKNOWLEDGMENT This work was supported in part by the Polish National
Science Centre (NCN) under Grant DEC-2012/07/B/ST7/01215 and statutory funds of the Department of Systems and Computer Networks, Wroclaw University of Technology.
REFERENCES [1] Cisco white paper, “Cisco Visual Networking Index: Forecast and
Methodology, 2010-2016” [2] L. Zong, G. N. Liu, A. Lord, Y. R. Zhou and T. Ma, “40/100/400 Gb/s
mixed line rate transmission performance in flexgrid optical networks,” in Proc. of OFC, 2013.
[3] M. Jinno, H. Takara, B. Kozicki, Y. Tsukishima, Y. Sone, S. Matsuoka, “Spectrum-efficient and scalable elastic optical path network:
architecture, benefits, and enabling technologies”, IEEE Communications Magazine, vol. 47, no. 11, pp. 66 – 73, 2009.
[4] M. Jinno, H. Takara, B. Kozicki, A. Hirano, Y. Tanaka, Y. Sone, A. Watanabe, NTT Corporation, “Distance-Adaptive Spectrum Resource Allocation in Spectrum-Sliced Elastic Optical Path Network”, IEEE Communications Magazine, vol. 48, no. 8, August 2010.
[5] J. Armstrong, “OFDM for optical communications,” IEEE/OSA Journal of Lightwave Technology, vol. 27, no. 3, pp. 189-204, Feb. 2009.
[6] I. Albanese, Y. O. Yazır, S. W. Neville, S. Ganti, T. E. Darcie „Big File Protocol (BFP): a Traffic Shaping Approach for Efficient Transport of Large Files”, HPSR, Vancouver, Canada, July 2014.
[7] K. Christodoulopoulos et al., "Elastic bandwidth allocation in flexible OFDM based optical networks", IEEE J. Lightw. Technol., vol. 29, no. 9, pp. 1354 – 1366, May 2011.
[8] M. Klinkowski and K. Walkowiak, "Routing and spectrum assignment in spectrum sliced elastic optical path network", IEEE Commun. Lett. vol. 15, no. 8, pp. 884 – 886, August 2011.
[9] K. Walkowiak, „QoS Dynamic Routing in Content Delivery Network”, Lecture Notes in Computer Science, vol. 3462, pp. 1120-1132, 2005.
[10] K. Walkowiak, A. Kasprzak, M. Klinkowski, “Dynamic Routing of Anycast and Unicast Traffic in Elastic Optical Networks”, ICC, Sydney, Australia, June 2014.
[11] M. Aibin, K. Walkowiak, „Dynamic Routing of Anycast and Unicast Traffic in Elastic Optical Networks with Various Modulation Formats - Trade-Off Between Blocking Probability and Network Cost”, HPSR, Vancouver, Canada, July 2014.
[12] Z. Zhu, W. Lu, L. Zhang and N. Ansari, “Dynamic Service Provisioning in Elastic Optical Networks With Hybrid Single-/Multi-Path Routing”, Journal Of Lightwave Technology, vol. 31, no. 1, January 1, 2013.
[13] L. Zhang, W. Lu, X. Zhou and Z. Zhu, “Dynamic RMSA in Spectrum-Sliced Elastic Optical Networks for High-Throughput Service Provisioning”, ICNC, San Diego, USA, January 2013.
[14] X. Chen, Y. Zhong and A. Jukan, “Multipath Routing in Elastic Optical Networks with Distance-adaptive Modulation Formats”, IEEE ICC, Budapest, Hungary, June 2013.
[15] S. Yang and F. Kuipers, “Impairment-Aware Routing in Translucent Spectrum-Sliced Elastic Optical Path Networks”, 17th ECNC, Barcelona, Spain, June 2012.
[16] M. Klinkowski, “On the Effect of Regenerator Placement on Spectrum Usage in Translucent Elastic Optical Networks”, ICTON, Coventry, England, July 2012.
[17] T. Zami, “Co-Optimizing Allocation of Nyquist Superchannels and Physical Impairments Aware Placementof Regenerators in Elastic WDM Networks”, Journal of Lightwave Technology, vol. 32, no. 16, August 15, 2014.
[18] N. Wang, J. P. Jue, „Holding-Time-Aware Routing, Modulation, and Spectrum Assignment for Elastic Optical Networks”, IEEE Globecom, Texas, USA, 2014.
[19] DataCenterMap. [Online]. Available: http://www.datacentermap.com [20] M. Klinkowski and K. Walkowiak, „On Advantages of Elastic Optical
Networks for Provisioning of Cloud Computing Traffic,” IEEE Network, Vol. 27, No. 6, pp. 44-51, 2013.
[21] C. Politi et al., "Dynamic Operation of Flexi-Grid OFDM-based Networks," in Proc. of OFC, 2012.
[22] J. M. Simmons, “Optical Network Design and Planning”, 2nd Edition, Springer, 2014.
Agenda
• Introduction
• Elastic Optical Networks
• Problem Formulation
• Results
• Future Works
Future works
• Algorithms for calculating the best regenerators’ location
• More sophisticated algorithms for dynamic routing
• Multicast traffic
• Network simulator with availability to choose particular modulations, networks and traffic
• Network orchestration using SDN controllers
Michał Aibin
Thank You For Your Attention
Any questions?