# Cloud-Centric and Logically Isolated Virtual Network Environment Based on Software-Defined Wide Area Network

^{1}

^{2}

^{*}

## Abstract

**:**

## 1. Introduction

## 2. Network Architecture and Core Components of KREONET-S

#### 2.1. OpenFlow-Based Software-Defined Wide Area Network Deployment on KREONET-S

#### 2.2. Virtually Converged Network Environment Based on VDN Functionality

## 3. Scale-Aware Virtual Network Generation

#### 3.1. Network Abstraction

- A link is available if it can guarantee the required network bandwidth and it is not pre-allocated by other VDNs.
- A link is available if its remaining network bandwidth is larger than the required bandwidth.

#### 3.2. VDN Generation Algorithms Based on Node Centrality

_{c}with the highest closeness centrality for edge nodes N(H

_{v}) of the required VDN end hosts H

_{v}, and then calculate the shortest paths between the n

_{c}and N(H

_{v}), and then merge the calculated paths into the VDN tree T. The detailed procedure is shown in Algorithm 1.

Algorithm 1. VDN Tree base on n_{c} in $G\prime \left({V}^{\prime},{E}^{\prime}\right)$ | |

0: | Parameter ${G}^{\prime}\left({V}^{\prime},{E}^{\prime}\right),N\left({H}_{v}\right)$ |

1: | Initialize $T$, ${n}_{c}$, sum, min = length of $G\prime \left({V}^{\prime},{E}^{\prime}\right)$; |

2: | for each $v\in $ V‘, and $v$ $\notin N\left({H}_{v}\right)$ |

3: | for each $n\in $ $N\left({H}_{v}\right)$ |

4: | Find ${p}_{v,n}$ on ${G}^{\prime}$ and sum += the length of ${p}_{v,n}$; |

5: | end for |

6: | if ( sum $<$ min) |

7: | min = sum and ${n}_{c}$ = $v$; |

8: | end if |

9: | sum = 0; |

10: | end for |

11: | for each $n\in $ $N\left({H}_{v}\right)$ |

12: | Find ${p}_{n,{n}_{c}}$ on ${G}^{\prime}$ and $T$ = $T\cup {p}_{n,{n}_{c}}$; |

13: | end for |

14: | return $T$ |

_{v}, the condition of selecting a center node n

_{c}is as follows:

_{v}). Therefore, the Algorithm 1 can result in poor performance if the size of G′(V′, E′) becomes large. In order to improve this case, we propose Algorithm 2, which can select n

_{c}in the subgraph consisting of plural shortest paths between N(H

_{v}). Thus, Algorithm 2 has good performance if the size of N(H

_{v}) becomes small regardless of the size of G′(V′, E′). For Algorithm 2, the number of the shortest path calculations can be formulated as:

Algorithm 2. VDN Tree base on ${n}_{c}$ in $G\u2033\left({V}^{\u2033},{E}^{\u2033}\right)$ | |

0: | Parameter ${G}^{\prime}\left({V}^{\prime},{E}^{\prime}\right),N\left({H}_{v}\right)$ |

1: | Initialize $G\u2033,T$, ${n}_{c}$, sum, min = length of $G\prime \left({V}^{\prime},{E}^{\prime}\right)$; |

2: | for each $n1\in $ $N\left({H}_{v}\right)$ |

3: | for each $n2\in $ $N\left({H}_{v}\right)$ |

4: | Find ${p}_{n1,n2}$ on ${G}^{\prime}$ and ${G}^{\u2033}=G\u2033\cup {p}_{n1,n2}$; |

5: | end for |

6: | end for |

7: | Perform Algorithm 1 ($G\u2033\left({V}^{\u2033},{E}^{\u2033}\right)$, $N\left({H}_{v}\right)$) |

_{v}), |V’’| can be determined while performing Algorithm 2. Therefore, we should estimate |V’’| using the graph size based on the graph density or random walk techniques [22,23]. Those estimation methods can also be adjusted based on the network operation experience. However, the methods may estimate an approximate value of |V’’| with a margin of error. Fortunately, it is acceptable because the performance of the two algorithms is within the margin of error.

## 4. Selective VDN Reconfigurations

- (1)
- The updated VDN hosts have to communicate with each other in the updated VDN;
- (2)
- The end-to-end paths between updated VDN hosts should support the required bandwidths;
- (3)
- The communications between previous VDN hosts except for the removed hosts should not be affected by the VDN reconfiguration;
- (4)
- The process of VDN reconfiguration should be performed within the range of service tolerance.

_{r}to release the network resources related to H

_{r}from the VDN. It can be decided by comparing the set of requested VDN hosts H

_{n}to the set of previous VDN hosts H

_{p}. Given H

_{v}and H

_{p}, H

_{r}is the difference set of H

_{v}from H

_{p}(H

_{r}= H

_{p}− H

_{v}). For example, as shown in Figure 4, H

_{r}= {h1, h4, h7} because of H

_{p}= {h1, h2, h3, h4, h7}, and H

_{v}= {h2, h3, h5, h6, h8}. If H

_{r}is empty, we do not have to perform step 2 for the removed hosts.

Algorithm 3. Selective VDN Reconfiguration | |

0: | Parameter ${G}^{\prime}\left({V}^{\prime},{E}^{\prime}\right),T,\text{}{n}_{c},\text{}{H}_{p},{H}_{v}$ |

1: | ${H}_{r}$ = ${H}_{p}$ − ${H}_{v}$; |

2: | if (${H}_{r}\mathrm{is}\text{}\mathrm{not}\text{}\mathrm{empty})$ |

3: | Initialize ${T}^{\prime}$ and remove all flow rules related with ${H}_{r}$; |

4: | for each $h\in {H}_{p}$ |

5: | if ($h$ $\notin {H}_{r}$ and $<$ ${n}_{h}$ $\notin N\left({H}_{v}\right)$) |

6: | Find ${p}_{{n}_{h},{n}_{c}}$ on $T$ and $T\prime $ = $T\prime \cup {p}_{{n}_{h},{n}_{c}}$; |

7: | end if |

8: | end for |

9: | $T$ = ${T}^{\prime};$ |

10: | end for |

11: | ${H}_{a}$ = ${H}_{v}$ − ${H}_{p}$; |

12: | for each $h\in $ ${H}_{a}$ |

13: | if (${n}_{h}$ $\notin $ T) |

14: | Find ${p}_{{n}_{h},{n}_{c}}$ on $G\prime $ and = $T\cup {p}_{{n}_{h},{n}_{c}}$; |

15: | end if |

16: | end for |

17: | if (T is changed) |

18: | Reselect ${n}_{c}$ in T |

19: | end if |

20: | return $T$ |

_{r}is found, we eliminate all the flow rules related to H

_{r}from the SDN nodes in the VDN. Moreover, the network resources for H

_{r}should also be released from the VDN. However, this procedure will be unnecessary if an edge node n

_{h}connecting to the removed host in H

_{r}is included in N(H

_{v}) because the edge node will be used for other VDN hosts in H

_{v}. Therefore, in this case, we only remove the flow rules related to the removed host. We can see those cases for h4 and h7 in Figure 4c. Otherwise, we should release the network resources related to the removed host in the VDN such as the edge node of h1 in Figure 4c. For this, we calculate the shortest paths between n

_{c}and N(H

_{p}) for all H

_{p}except for the above hosts in the previous VDN (lines 4–8), and then merge the calculated paths into the updated VDN. This is because the nodes and links related to the remaining VDN hosts can be removed if the nodes and links are included in the paths between the center node and the edge nodes for the removed hosts such as the node of h1 in Figure 4c.

_{a}(lines 11–16 in Algorithm 3). For this, we should first find the set of additional VDN hosts H

_{a}to allocate the network resources for H

_{a}. It can also be decided by comparing H

_{v}and H

_{p}. Given H

_{v}and H

_{p}, H

_{a}is the difference set of H

_{p}from H

_{v}(H

_{a}= H

_{v}− H

_{p}) as shown in Figure 4, where H

_{a}= {h5, h6, h8}. If H

_{a}is empty, we do not have to perform step 3 for additional hosts similar to the procedure for H

_{r}.

_{a}, step 3 is simpler than that for H

_{r}. It finds the new shortest paths on the abstracted network between the center node n

_{c}and N(H

_{a}), and then merges it to the VDN. Furthermore, if an edge node n

_{h}of additional hosts already belongs to T, we do not have to merge the path related to n

_{h}. For this reason, the reconfiguration procedure of H

_{a}should be performed next to that of H

_{r}as shown in h8 in Figure 4d. Finally, if T is changed, we reselect a new center node n

_{c}based on the updated VDN using Equation (1).

_{p})| − |N(H

_{h})| and |N(H

_{a})|, respectively. In the real world, almost all the users reside in the same site (e.g., headquarter, branch, and main partner). Thus, the situation of adding and removing edge nodes is not frequent once VDNs are generated with the edge nodes included, although the characteristics of the nodes may change by the suggested pruning algorithm based on the two conditions in Section 3.1. That is, VDN re-configuration requires calculation of the new paths, otherwise they can be completed by just changing the set of VDN hosts.

## 5. Performance Evaluation and Experimental Results

#### 5.1. Evaluation Setup and Results Based on Mininet

_{r}are included in N(H

_{r}). Accordingly, in the proposed algorithm, we just remove the flow rules of H

_{r}and the H

_{r}information from the set of VDN hosts. Therefore, the VDN update time for selective VDN reconfiguration is much smaller than the update time of the others. Figure 7b depicts the VDN update time when all edge nodes of H

_{r}are not included in N(H

_{r}). Thus, in addition to releasing the network resources (flow rules and information for H

_{r}), the new shortest paths between n

_{c}and N(H

_{r}) have to be calculated. Even though an additional procedure is performed, the VDN update time is still much smaller than the update time of the others. The main reason for the performance difference between the proposed solution and the others is that the others are performed based on the entire network topology in contrast to the proposed solution, which is based on the pre-calculated VDN information. Figure 7c,d show the VDN update time according to the number of additional VDN hosts. In Figure 7c, all edge nodes of Ha already belong to the VDN to be updated in contrast to Figure 7d. Therefore, in Figure 7d, the new shortest path calculations are required for N(H

_{a}). According to the new path calculations, the VDN update time for the proposed solution in Figure 7d increases as the number of VDN hosts grows, compared to the update time of Figure 7c. However, in Figure 7c,d, we can observe that the VDN update time is much smaller than the update time of the others by similar reasons as those for Figure 7a,b. The update time of the selective VDN reconfiguration in Figure 7a,b is larger than that in Figure 7c,d because the former naively releases the network resources related to the VDN hosts.

#### 5.2. Evaluation Setup and Results Based on KREONET-S

## 6. Conclusions and Future Work

## Acknowledgments

## Author Contributions

## Conflicts of Interest

## References

- Weldon, M.K. The Future X Network: A Bell Labs Perspective; Taylor & Francis Group, LLC.: Boca Raton, FL, USA, 2016; Chapter 5; pp. 160–193. ISBN 978-89-93712-73-5. [Google Scholar]
- Manzalini, A.; Roberto, S.; Cagatay, B.; Prosper, C.; Slawomir, K.; Andreas, G.; Masake, F.; Eliezer, D.; David, S.; Mehmet, U. Software-Defined Networks for Future Networks and Services. In White Paper Based on the IEEE Workshop SDN4FNS; IEEE: Trento, Italy, 2014; pp. 5–14. [Google Scholar]
- What is Software Defined WAN (or SD-WAN)? Available online: https://www.sdxcentral.com/sdn/definitions/software-defined-sdn-wan/ (accessed on 14 November 2017).
- Coady, Y.; Hohlfeld, O.; Kempf, J.; McGeer, R.; Schmid, S. Distributed Cloud Computing: Applications, Status Quo, and Challenges. ACM SIGCOMM Comput. Commun. Rev.
**2015**, 45, 38–43. [Google Scholar] [CrossRef] - Jain, R.; Paul, S. Network Virtualization and Software Defined Networking for Cloud Computing: A Survey. IEEE Commun. Mag.
**2013**, 51, 24–31. [Google Scholar] [CrossRef] - Garcia Lopez, P.; Montresor, A.; Epema, D.; Datta, A.; Higashino, T.; Iamnitchi, A.; Barcellos, M.; Felber, P.; Riviere, E. Edge-centric Computing: Vision and Challenges. SIGCOMM Comput. Commun. Rev.
**2015**, 45, 37–42. [Google Scholar] [CrossRef] - Sun, L.; Ma, J.; Wang, H.; Zhang, Y. Cloud Service Description Model: An Extension of USDL for Cloud Services. IEEE Trans. Serv. Comput.
**2015**, 99, 1–14. [Google Scholar] [CrossRef] - Li, M.; Sun, X.; Wang, H.; Zhang, Y.; Zhang, J. Privacy-aware Access Control with trust management in Web Service. World Wide Web
**2011**, 14, 407–430. [Google Scholar] [CrossRef] - Wang, H.; Cao, J.; Zhang, Y. A Flexible Payment Scheme and its Role based Access Control. IEEE Trans. Knowl. Data Eng. (TKDE)
**2005**, 17, 425–436. [Google Scholar] [CrossRef][Green Version] - KREONET Introductions. Available online: http://www.nisn.re.kr/eng/action.do?menuId=50030 (accessed on 8 December 2017).
- Official KREONET-S Project Website. Available online: http://www.kreonet-s.net (accessed on 14 November 2017).
- Kim, D.; Cho, H.; Kim, Y.; Kim, K.; Yu, K.; Gil, J. User-Oriented Software-Defined Wide Area Network Adopting Virtual Dedicate Networks. ASP Adv. Sci. Lett.
**2016**, 22, 2262–2267. [Google Scholar] [CrossRef] - KREONET-S Implements An SD-WAN Connection From South Korea to The StarLight International/National Communications Exchange Facility In Chicago: Inaugurating Novel Advanced International Communications Services. Available online: http://www.startap.net/starlight/PUBLICATIONS/KREONET-S.html (accessed on 14 November 2017).
- Berde, P.; Gerola, M.; Hart, J.; Higuchi, Y.; Kobayashi, M.; Koide, T.; Lantz, B.; O’Connor, B.; Radoslavov, P.; Snow, W.; et al. ONOS: Towards an open, distributed SDN OS. In Proceedings of the Third Workshop on Hot Topics in Software Defined Networking (HotSDN 2013), Hong Kong, China, 2 January 2013; pp. 1–6. [Google Scholar]
- Open Network Operating System. Available online: http://onosproject.org/ (accessed on 14 November 2017).
- Nguyen, V.G.; Kim, Y.H. SDN-Based Enterprise and Campus Networks: A Case of VLAN Management. J. Inf. Process. Syst.
**2016**, 12, 511–524. [Google Scholar] - Gilani, S.M.M.; Hong, T.; Cai, Q.; Zhao, G. Mobility Scenarios into Future Wireless Access Network. J. Inf. Process. Syst.
**2017**, 13, 236–255. [Google Scholar] - Jung, H.; Kim, N. Multicast Tree Construction with User-Experienced Quality for Multimedia Mobile Networks. J. Inf. Process. Syst.
**2017**, 13, 546–558. [Google Scholar] - Sun, X.; Xie, G.G. An Integrated Systematic Approach to Designing Enterprise Access Control. IEEE/ACM Trans. Netw.
**2016**, 24, 3508–3522. [Google Scholar] [CrossRef] - Akyildiz, I.F.; Anjali, T.; Chen, L.; de Oliveira, J.C.; Scoglio, C.; Sciuto, A.; Smith, J.A.; Uhl, G. A new traffic engineering manager for Diff-Serv/MPLS networks: Design and implementation on an IP QoS testbed. Comput. Commun.
**2003**, 26, 388–403. [Google Scholar] [CrossRef] - OpenFlow
^{®}Switch Specification 1.3.1—Open Networking Foundation. Available online: https://3vf60mmveq1g8vzn48q2o71a-wpengine.netdna-ssl.com/wp-content/uploads/2013/04/openflow-spec-v1.3.1.pdf (accessed on 14 November 2017). - Gross, J.L.; Yellen, J. Graph Theory and Its Applications; CRC Press: Boca Raton, FL, USA, 1999. [Google Scholar]
- Kurant, M.; Butts, C.T.; Markopoulou, A. Graph Size Estimation. arXiv. 2012. Available online: https://arxiv.org/abs/1210.0460 (accessed on 14 November 2017).
- Mininet Website. Available online: http://mininet.org/ (accessed on 14 November 2017).

**Figure 1.**KREONET Softwarization (KREONET-S) Deployment Status in Korea and the US, composed of three main building blocks (data plane, control plane, and application services).

**Figure 2.**Virtually Converged Network Environment based on virtually dedicated network (VDN)/user-oriented visibility (UoV) Applications.

**Figure 4.**Example of VDN reconfigurations. Hp = {h1, h2, h3, h4, h7}, Hv = {h2, h3, h5, h6, h8}, Ha = {h5, h6, h8}, and Hr = {h1, h4, h7}.

**Figure 9.**TCP and UDP traffic measurement for the VDNs with 1 Gbps and 10 Gbps bandwidths provisioned between Daejeon (DJ) in Korea, Seoul (SL) in Korea, and Chicago (CHI), IL, USA.

Notation | Meaning |
---|---|

P_{i}_{,j} | Shortest path between node i and node j |

n_{c} | center node with the highest closeness centrality |

n_{h} | edge node of a host h |

N(H) | set of edge nodes of a host set H |

H_{v} | set of requested VDN hosts |

H_{p} | set of previous VDN hosts |

H_{a} | set of added VDN hosts |

H_{r} | set of removed VDN hosts |

Case I | Case II | Case III | Case IV | Case V | |
---|---|---|---|---|---|

# of Nodes | 8 | 18 | 28 | 38 | 48 |

# of VDN hosts | 38 | 40 | 42 | 44 | 45 |

# of Unidirectional Links | 154 | 218 | 244 | 308 | 368 |

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Kim, D.; Kim, Y.-H.; Kim, K.-H.; Gil, J.-M. Cloud-Centric and Logically Isolated Virtual Network Environment Based on Software-Defined Wide Area Network. *Sustainability* **2017**, *9*, 2382.
https://doi.org/10.3390/su9122382

**AMA Style**

Kim D, Kim Y-H, Kim K-H, Gil J-M. Cloud-Centric and Logically Isolated Virtual Network Environment Based on Software-Defined Wide Area Network. *Sustainability*. 2017; 9(12):2382.
https://doi.org/10.3390/su9122382

**Chicago/Turabian Style**

Kim, Dongkyun, Yong-Hwan Kim, Ki-Hyun Kim, and Joon-Min Gil. 2017. "Cloud-Centric and Logically Isolated Virtual Network Environment Based on Software-Defined Wide Area Network" *Sustainability* 9, no. 12: 2382.
https://doi.org/10.3390/su9122382