Impact of flexible mechanism on localized hopby hop routing algorithm

In this paper, we propose a novel localized QoS routing algorithm, which uses hop-by-hop routing (or distributed routing) type to route flows. We perform it with experiments, compare and realize the more considerable performance of this algorithm than other algorithms, which use the same conditions of simulation.

Kỹ thuật điều khiển & Điện tử IMPACT OF FLEXIBLE MECHANISM ON LOCALIZED HOPBY-HOP ROUTING ALGORITHM Tran Minh Anh*, Nguyen Chien Trinh Abstract: Nowadays, the requisition for quality of network is more and more popular and sophisticated to meet the demand of telecommunication (telecom) market To assure the performance of network, Quality of Service (QoS) routing algorithms based on local information have recently been researched as a promising method for the present global QoS routing schemes With locally collected statistics information, the localized routing algorithm can avoid some weakpoints committed by global routing algorithms such as: a large communication overhead, the inexact of global state information at nodes and the out of date information etc In this paper, we propose a novel localized QoS routing algorithm, which uses hop-by-hop routing (or distributed routing) type to route flows We perform it with experiments, compare and realize the more considerable performance of this algorithm than other algorithms, which use the same conditions of simulation From that, we imply that localized hop-by-hop routing algorithm with flexible mechanism works more effectively than other types of routing algorithm Keywords: Algorithm; Hop-by-hop; Localized; QoS; Routing INTRODUCTION Presently, telecom services require very high speed and high quality transmission To solve that problem, the technique of using local information at source node to route flows is an effective method to communicate in network This technique will diminish the blocking probability of flows transmitted through the network as well as improve the performance of network It is showed that this technique is simpler and better than traditional global QoS routing schemes The traditional global routing technique makes routing decisions by using global network state information, hence, it commits some problems, such as: a large communication overhead, the inexact of global state information and the out-of-date information etc., due to the fact that the global network state information needs to be precise, to be maintained and to be exchanged periodically to keep routing effectively The localized QoS routing method researched recently is to use local network state information at source node such as: blocking probability, network statistics etc to route data In researched localized schemes, the approach of localized hop-by-hop routing algorithm is more flexible, because it needs each node to find and maintain only a predetermined set of adjacent nodes that can be used to route flows through towards each possible destination Moreover, due to keeping track of only a set of nodes, source node will occupy less memory than other localized routing algorithms that have to maintain the full set of paths In this paper, we study the type of hop-by-hop routing, some related notions and definitions, and propose a novel routing algorithm, a localized hop-by-hop routing algorithm with flexible mechanism of controlling We also realize better performance against some other localized routing algorithms and a global QoS routing algorithm (WSP), under the same types of topology, traffic patterns and range of traffic loads as in the section 26 T M Anh, N C Trinh, “Impact of flexible mechanism … hop-by-hop routing algorithm.” Nghiên cứu khoa học công nghệ RELATED WORK The parameters of QoS like bandwidth, delay … have been used to estimate the quality of transmission for telecom services in some years recently and there are also a lot of studies of QoS routing which have been published on many different areas as [1-3] and references therein In the future, when the convergence of services becomes stronger, QoS routing will be more important As introduced in the previous section, localized QoS routing algorithm has been proposed recently as an effective alternative to global QoS routing as in [4-5] and [13-14] In localized routing algorithm, nodes make routing decisions using only local state information; so global state information does not need to be exchanged any longer, and this will reduce the overhead computation at nodes We now survey some localized QoS routing schemes as follows The first scheme we mention is the algorithm of Credit Based Routing (CBR) as in [4] The CBR uses a simple routing procedure to route traffic across the network by using crediting scheme for each candidate path that rewards a path upon flow acceptance and penalizes it upon flow rejection The larger path credits, the larger chances for selection The CBR algorithm keeps updating each path's credit upon flow acceptance; keeps monitoring the flow blocking probabilities for each path and conveys the data to the credit scheme to select path The CBR predetermined a set of candidate paths R between each pair of source and destination where = ∪ (Rmin: the minimum hop path set and Ralt: the alternative path set) The CBR selects the largest credit path P.Credits in each set, minimum hop (minhop) paths set Rmin and alternative paths set Ralt upon flow arrival The flow is routed along the minimum hop path that has the largest credit Pmin that is larger than the alternative path that has the largest credits Palt following the formula (1): ≥ Ф , where Ф≤ (1) Otherwise, Palt will be chosen (The parameter Ф in [1] is a system parameter that controls the usage of alternative paths) The blocking probability in crediting schemes is used to improve the performance of the algorithm The path credits are incremented or decremented upon flow acceptance or rejection using statistics of the path blocking probability Besides, the CBR uses a MAX_CREDITS system parameter to determine the maximum attainable credits for each path by computing the blocking probability ≤ ≤ _ (2) The CBR algorithm uses a moving window for a predetermined period of M connection requests It uses for flow acceptance and for flow rejection, dividing the number of 0's by M to calculate each path blocking probability for the period of M connection requests The main problem with CBR is that a path’s credits are only updated each time when that path is selected If a path is selected infrequently, then its credit value will become stale leading to errors in the selection process The CBR scheme uses source routing type It means that the intermediary nodes only transfer data without changing the path decided by source node Other than source routing, the hop-by-hop routing does differently as in [6, 7] Hop-by-hop routing means that routing decisions are made independently at each node This type of routing algorithm is used quite much now, and it will be described in detail in next section Using this type, there are some proposed algorithms such as the global distributed routing algorithm proposed in [7], or the algorithm of Localized Distributed QoS Routing (LDR) proposed in [5] Tạp chí Nghiên cứu KH&CN quân sự, Số 59, 02 - 2019 27 Kỹ thuật điều khiển & Điện tử One popular algorithm using hop-by-hop routing is distributed routing algorithm proposed in [7] It uses global network state and distance vector protocols to maintain that state and it routes flows by the type of distributed method By that, every node will use the Dijsktra algorithm as in [8] to compute routing table Such algorithms often cost the high overhead because of much exchange of update messages for maintaining global network state Likewise, the algorithm of LDR proposed in [5] also has the idea of distributed routing like the algorithm above It means that LDR considers only the links belonging to the candidate paths to route flows At the same time, at each intermediate node, the LDR considers only the candidate paths between the current node and the destination node The LDR routing algorithm consists of two stages: forwarding incoming flows towards their destinations and then updating local routing statistics accordingly Local routing statistics are updated with every flow forwarding attempt Like CBR, LDR maintains a predetermined set of candidate paths R between each source-destination pair where = ∪ (Rmin: the minimum hop path set and Ralt: the alternative path set) LDR checks all outgoing links that belong to the paths in the candidate path set between source and destination node The link with the least amount of blocking is selected, and then its residual bandwidth is compared against the flow QoS bandwidth request If it satisfies the flow request, this new node is then considered as the next hop on the feasible path When a new node is selected as the next hop on the feasible path, a new search cycle begins looking for the other subsequent hops on the feasible path The search cycle is repeated, as explained before, checking for the outgoing link with the lowest blocking probability This distributed searching process is repeated until the destination node is reached LDR uses the message system to control the process of checking QoS requirements from all steps When the source node receives a success message indicating a successful flow request, it maintains the selected path for the duration of the flow However, because all intermediate nodes also keep full set of paths between source node and destination node, LDR will waste a lot of memory for communication overhead, and at the same time, all the intermediate nodes must the checking path as well as the first node, so it will get a little bit congested when the load increases Some of the schemes above will be used to compare with our proposed algorithm that uses the type of hop-by-hop routing through simulations as showed in the section A NEW LOCALIZED HOP-BY-HOP ROUTING ALGORITHM 3.1 Methodology As introduced, we here describe the principle of distributed routing that will be used in our proposed algorithm Distributed routing means that routing decisions are made independently and locally at each node, towards destination node, with the route computation using corresponding topology knowledge For each incoming flow at a node, it specifies the destination node to get the next hop by consulting the routing table from itself Therefore, distributed routing is also referred to as destination-based table-driven routing and the searching for next hop is realized until the flow reaches its destination To use the distributed routing in our algorithm, we imply some theorems below: Theorem 1: In the distributed routing, shortest path found from distributed computation is loopless in case of no failure Proof: We prove it by contradiction Suppose the path A, which is the shortest path between node i and node j, contains nodes X and Y in a loop It is clear that the path A is longer than the path A’ which contains only node X (see Figure 1) 28 T M Anh, N C Trinh, “Impact of flexible mechanism … hop-by-hop routing algorithm.” Nghiên cứu khoa học công nghệ Figure Proof of theorem Nevertheless, we have supposed that path A is shortest; and no failure happens Therefore, that contradicts the supposition It means that path A cannot contain any loop This completes the proof Theorem 2: In a graph, the distance by hop count between two any nodes is limited if the path between them is loopless The shortest path is the path having shortest distance by hop count between these two nodes Proof: It is supposed that there are N nodes in that network and i, j are two nodes in that network Let path A denote the path between these two nodes From the supposition, that path is loopless Because the maximum number of nodes is N (the size of network), the maximum length of all loopless path is (N-1) Due to the fact that the path is loopless, the distance between two any nodes in the network is limited Also, the shortest path will have the shortest distance (by hop count) between that pair of nodes That completes the proof Theorem 3: Suppose that there is a set of k loopless paths from node X to node Z which has maximum of n hops (n>1, n is an integer) (see Figure 2) Let Y be the adjacent node to X on an any path in that set Suppose that is path A We will always have a path between Y and Z (part of path A) which has maximum of (n-1) hops Figure Proof of theorem Proof: From the supposition, we can see that path A from X, Y to Z is loopless Therefore, the path from Y to Z, which is a part of path A, is loopless, too In addition, from theorem 2, we can observe that the path from X to Z is limited From the supposition, we have node Y being the 2nd node in a path between X and Z which contains Y Because the maximum distance (by hop count) of path A is n (as supposed), the part of path A from Y to Z will have: length(Y-Z) = length(X-Z)-1= length (path A)-1 ≤ n-1 (3) That completes the proof With these notions from theorems above, we can build the algorithm in the next part 3.2 A new localized Hop-by-hop routing algorithm with flexible mechanism Tạp chí Nghiên cứu KH&CN quân sự, Số 59, 02 - 2019 29 Kỹ thuật điều khiển & Điện tử As introduced in the first section, the localized routing algorithm will help to improve the overall performance of network In this part, we will propose a novel localized hop-byhop routing algorithm with Flexible Mechanism, with bandwidth as constraint, and is socalled Localized Distributed bandwidth-constraint Routing Algorithm or LDRA In our algorithm, LDRA requires every node to determine all the shortest paths to each destination at first Then, it only maintains a set K of next nodes of the source node from these shortest paths corresponding to each destination Each next node KiK is associated with a variable Ki.Bandwidth, an index Bi: total times of failing to transmit any flow of the link to Ki and a control index called pt_idx The index pt_idx is used to keep all flows to get destination At first, this index is set by the number of hops of the shortest path towards destination node (through node Ki) Let B denote the sum of flows rejected from the source node to the destination node and let T be the sum of flows destined to that destination node Then, the ratio {Bi/T} will be used to range the set K This procedure is called “Range K{Bi/T}” To choose next node, we choose the node in set K with minimum value of {Bi/T} If there are many next nodes having the same value of {Bi/T}, we choose the next criterion: the maximum value of Ki.Bandwidth or max(Ki.Bandwidth), where Ki.Bandwidth is the residual bandwidth between the current node and node Ki All the steps for choosing next node like that are called shortly min(K) When the ratio (B/T)>½, it means half of flows destined to that destination nodes is rejected, the source node will rebuild the set of nodes by extending set of paths till that destination node At the same time, the control index pt_idx will increase one for each particle of set K The procedure of rebuilding the set of nodes will be discussed in the next part Routing process: upon flow arrival, LDRA will select the next node Ki with min(K) (first particle in the set K after ranging), check the bandwidth demand of the flow and use it for choosing next node That demand is called RQ (Requested Quality) If this is the source node, the flow is added an index called fl_idx This index will traverse through network with this flow to control all the routes At the source node, the value of fl_idx is N (where N is the number of nodes in the network) for the first time when it is just added to the flow as follows: fl_idx = N (4) If this is an intermediate node, LDRA only updates new value of fl.idx as in (5) Next, LDRA will compare Ki.Bandwidth with RQ (the demand)  If Ki.Bandwidth ≥ RQ, the node Ki will be chosen Then LDRA updates the control index of this flow as follows: fl.idx = min(fl.idx,Ki.pt.idx) – (5)  If Ki.Bandwidth < RQ, the next Ki in the set K (the min(K) of the rest of K) is chosen, and the Bi of the old node increases The loop will be done until finding out the node has minimum of {Bi/T} in the set K and Ki.Bandwidth ≥ RQ If there is not any Ki (in the set K) which has Ki.Bandwidth larger than RQ, the arriving flow is cancelled, obviously In that case, the index T, B and all of Bi will increase When transmitting successfully a flow to destination node, only value of T increases We can see that: From theorem 1, all the paths received from Dijkstra algorithm to determine set of nodes at first and from the rebuilding procedure described in part 3.4 are 30 T M Anh, N C Trinh, “Impact of flexible mechanism … hop-by-hop routing algorithm.” Nghiên cứu khoa học công nghệ loopless to each destination From theorem 3, the next node will always find out a path to that destination that has maximum of (n-1) hop count Therefore, when the source node transmits data to the next node, the next node always has the available set of nodes to forward data towards destination From theorem 2, the next node always forwards data received from the source node to the destination because this process is limited with the decrement of control index It means that although the source node does not send the full path to every node on the path, the intermediary nodes always forward the data to the destination as well as the source routing algorithm model does 3.3 Procedure used to route data in LDRA To use the algorithm in reality, we build the procedure for LDRA algorithm at each node as follows: PROCEDURE LDRA(pathK); BEGIN - Range(K); - SelectNode(Ki,Bi,T); - CompareQuality(Ki.Bandwidth,Flow.Bandwidth); - Update fl.idx; - TransmitOK(OK); - If OK: UpdateNode(T); - If Not OK: UpdateNode(B,Bi,T); END 3.4 Procedure used to rebuild set of nodes To rebuild set of nodes, LDRA will at each node as follows: PROCEDURE LDRA_rebuild(setK); BEGIN - Check condition(B/T); - If larger than 1/2: +Search all loopless path with hop count = pt.idx+1; +Add next node of each path to set K; + Add new pt.idx=pt.idx+1; +Add all variables associated to new node; END 3.5 Flow chart of all steps Following the flow-chart, LDRA changes the ratio {Bi/T} of each node from the set K of nodes as well as the value of B Then, the ratio {Bi/T} is used to compare among nodes If the node has lower ratio, it means it has “better quality” After the next flow comes, LDRA will use this ratio {Bi/T} as criterion to choose node for routing, and next loop begins Therefore, after one flow is processed, the ratio {Bi/T} changes accordingly to the success/fail rate of that node and the “value” of that node changed correspondingly This influences the probability of that adjacent node to be chosen for the next flow Tạp chí Nghiên cứu KH&CN quân sự, Số 59, 02 - 2019 31 K Kỹỹ thuật điều khiển & Điện tử Figure 33 Flow chart of LDRA PERFORMANCE PERFORMANCE EVALUATION In this section, we realize many experiments to evaluate the performance of our scheme against other schemes such as CBR, LDR, and WSP All the experiments are set in the same conditions After all, we collect and analyze the results of our simulation experiments and performance metrics 4.1 Simulation m model odel We use the OMNET++ simulator as in [12], a commonly used event event-driven driven simulator To evaluate the results of experiments, we collect all of parameters of simulation as vectors, scalars and histograms to compare The setup of simulations is the same as the simulations in [4, 5] with the topology of the well well-known known ISP in the world, described as follows: network is built with 32 nodes; links of these nodes are all bidirectional with the same capacity C = 150Mbps in each direction; flows arrive to each source node according to a Poisson process with rate λ;; destination nodes are selected randomly (every node can be source node or destination node), flow duration is exponentially distributed with mean 1/; and required bandwidth for flows is uniformly distributed within [0.1 [0.1-2] Mbps The Mbps 2]Mbps network is as in Figure 32 T M Anh, N C Trinh hop-by by-hop hop routing algorithm algorithm.”” Trinh, “Impact “Impact of flexible mechanism … hop Nghiên cứu khoa học công nghệ Figure Network of 32 nodes From [10, 11], the offered network load is: bh/LC (6) where N is the number of nodes, b is the average bandwidth required by a flow, h is the average path length and L is the number of links in the network In the experiments, we set N=32, L=108, h=3.137, 1/= 60s Since the performance of routing algorithms may vary across different load conditions, our simulations consider several types of different load conditions through the value of λ according to experiments of from low loads to high loads To compare with other schemes, we calculate blocking probabilities as well as the simulations in [4, 5] The results are calculated based on the most recent 100,000 flows in more than 2.5 million of emitted flows Then, the standard overall flow blocking probability is defined as: Flow Blocking Probability = |B|/|T| (7) where |T| is the total of all flows and |B| is total of blocked flows Besides, we calculate bandwidth blocking probability (BBP) which is defined as: BBP = (bandwidth of |B|)/(bandwidth of |T|) (8) Otherwise, we calculate and compare the Jain's fairness index of load balancing proposed in [9] among these algorithms From results, we can conclude the effectiveness of LDRA against other schemes 4.2 Simulation results A Flow Blocking and Bandwidth Blocking Probability With the results of experiments, we compare our scheme (LDRA) with the other schemes, as showed in figures below Figure shows the performance of LDRA against CBR, LDR and WSP in terms of flow blocking probabilities under load  varying between 0.45 and 0.75 From the figures, we can observe that under low load ( ≤ 0.5), the difference in the performance of the routing algorithms is quite small, because finding available path with sufficient bandwidth is easy and flows are almost accepted When  is higher (more than 0.5), more flows drop or fail to get destination node, therefore flow blocking probability grows rapidly as showed in figure Tạp chí Nghiên cứu KH&CN quân sự, Số 59, 02 - 2019 33 K Kỹỹ thuật điều khiển & Điện tử Figure 5 Flow blocking locking probability robability In the case of the LDRA scheme, flows are transmitted through node node by node node by the index of each node in the set K When there is a link that cannot afford for the flow, the Bi of next node corresponding to that link increases; and then the current node changes to another next node based on the rate {{B Bi/T} } Therefore, it assures the flow to be transmitted “OK” to the next node Moreover, the index of next nodes ((K Ki.pt_idx) pt_idx) is set set to avoid the congestion of flows that come to nodes simultaneously, particularly when the load increases and the links begin to become congested If congestion happens, flows will be re re directed directed to other link connected to the node that has the minimum va value lue of {{Bi/T /T} } Then, the node might diminish using of the next nodes, which have high index When the “quality” of the set of nodes goes down with the ratio ((B/T B/T)) larger than ½, that node rebuilds the set following the procedure in part 3.4 That will help LDRA to extend its set of nodes as well as improve quality of routing Therefore, the flow blocking probability of LDRA is considerably low against the one of other schemes as well Beside the comparison of flow blocking probability as in the Figure 5, w wee also compare the results of bandwidth blocking probability of our scheme with the case of WSP - one of global routing schemes Results are showed in Figure Figure 6 Bandwidth Blocking probability From the results, we can see that the scheme of LDR LDRA A has the more considerable performances The reason is the same as above due to the link changing on the path towards to destination based on the value of {{B Bi/T} /T} The BBP of WSP is higher, because WSP only forwards flows to destination in the stable path Hence, when the load is high, the congestion happens frequently, hence, it leads to the high BBP as showed in Figure 34 T M Anh, N C Trinh hop-by by-hop hop routing algorithm algorithm.”” Trinh, “Impact “Impact of flexible mechanism … hop Nghiên ccứu ứu khoa học công nghệ Therefore, we can see that with the type of distributed routing, the blocking probability goes down That involves the better quality of network than the case of WSP in some experiments that have been done B Jain's fairness index and Load Load-balancing balancing We can see that the algorithm will work better if it makes the network is more load balancing As analized in [9], network load load-balancing balancing or routing fairness is another metric that can indicate the efficiency of network resource utilization The index prop osed by [9] proposed is called Jain's Fairness airness Index or Jain’s Index and is defined as follows: Jain’s Index = ∑ ∑ (9)) where L is the number of links, xi is the load on the link The Jain’s Index will be if the load is fairly distributed If not, the index will diminish down until the minimum value of 1/ 1/L L where the load concentrates on only one link Therefore, the index of Jain shows the fairness of offered load distributed across all network li links nks From the Figure 7, we can perceive that WSP has the lowest value of index ndex and the index of our scheme gets the most favorable value against other schemes It is due to the fact that it distributes the flows more flexibly with the flexible set of nodes as described in the previous section Hence, it helps to improve the network load balancing Figure 77 Jain’s Index for comparison In concluding, with the type of hop hop by by-hop hop routing and the control index above, we can perceive that the proposed algorithm has better performances than other schemes in the experiments of simulation 4.3 Complexity and overhead 4.3 The case of WSP uses the algorithm Dijkstra, like almost global QoS routing algorithms, takes at least O(NlogN+E) time, where N is the size of the network measured in the number of nodes, and E is the number of links (edges), see [2] At the same time, the localized schemes use the way of routing that selects path from the set of candidate paths R like [4-5], [4 5], [13 [13 14] 14] In the CBR algorithm aass [4], the path selection is an invocation of a weighted weighted-round round-robin round robin like path selector (wrrps), whose worst case time complexity is O(|R|), similarly LDR requires better than that, for it uses only the minhop paths as explained above (with ||R|| here is the number of candidate paths) At the same time, LDRA also has worst worst-case case time complexity is O O(|K K|), |), in which |K |K|| is the number of next nodes of each source node This ||K|| is quite smaller than the ||R R|| in other localized schemes above Tạp ạp chí Nghi Nghiên ên cứu cứu KH&CN quân uân sự, sự, Số 599, 022 - 2019 20 35 Kỹ thuật điều khiển & Điện tử In addition, these localized schemes require updating information, which takes a constant time O(1) Hence, with communication overhead, LDRA or other localized schemes requires very little over and above computing the blocking probability based on acceptance or rejection of a node, while at the same time, global algorithms require a huge amount of overhead to keep the link state information updated In conclusion, the computation of our case is much smaller than the one of traditional global algorithm cases CONCLUSION AND ONGOING WORK In this paper, we study and propose the notions concerned to distributed routing, which can be used for localized routing algorithms Besides, we propose the new localized hopby-hop routing algorithm (LDRA) using the flexible set of nodes to route flows Our proposed algorithm uses bandwidth as constraint as other schemes (LDR, CBR …) With other QoS parameters like delay, packet loss etc, we can convert to bandwidth as [15] in some contexts We have realized a lot of experiments to compare the performances among our scheme and other schemes like CBR, LDR, WSP etc These experiments have already showed better performances of LDRA, which uses the distributed routing type against the others As part of future work, we will investigate the effect of the topology of network on the operation of the localized routing algorithm The algorithms work more effectively when the network has the more balancing topology With the proposition of distributed routing algorithm on that better topology network, the load will be distributed more equally to all links in network, and at the same time, it brings better results for network routing REFERENCES [1] C Pornavalai, G Chakraborty, N Shiratori,QoS based routing algorithm in integrated services packet networks, in: Proceedings of the IEEE ICNP, 1997 [2] S Chen and K Nahrstedt,“An Overview of Quality-of-Service Routing for the Next Generation High-Speed Networks: Problems and Solutions,” IEEE Network, special issue on transmission and distribution of digital video, vol 12, pp 64-79, 1998 and references therein [3] R Guerin, S Kamat, A Orda, T Przygienda and D Williams, “QoS Routing Mechanisms and OSPF Extensions,” Work in Progress, Internet Draft, March 1997 [4] S Alabbad, M E Woodward “Localized Credit Based Routing: Performance Evaluation Using Simulation", Proceedings of IEEE 39th Annual Simulation Symposium, Huntsville, Al USA April 2-6, 2006 [5] F M Aldosari and M E Woodward, “Localized approach to distributed QoS routing with a bandwidth guarantee,” in Proceedings of IEEE 3rd International Conference on Communication Software and Networks (ICCSN), Xi’an, 2011, pp 271-276 [6] S K Kweon and K G Shin, “Distributed QoS routing with bounded flooding for real-time applications,” Technical report, University of Michigan, 1999 [7] P V Mieghem, H D Neve and F Kuipers, “Hop-by-hop quality of service routing,” Computer Networks , v.37, p.407-423, Nov 2001 [8] Dijkstra Algorithm https://www.cs.auckland.ac.nz/software/AlgAnim/dijkstra.html [9] R Jain, W Hawe, D Chiu, “A Quantitative measure of fairness and discrimination for resource allocation in Shared Computer Systems,” DEC-TR-301, September 26, 1984, http://www.cse.wustl.edu/~jain/papers/ftp/fairness.pdf 01/5/2016 [10] A Shaikh, J Rexford, K.G Shin, “Load-Sensitive Routing of Long-Lived IP Flows”, ACM SIGCOMM 1999 36 T M Anh, N C Trinh, “Impact of flexible mechanism … hop-by-hop routing algorithm.” Nghiên cứu khoa học công nghệ [11] A.Shaikh, J.Rexford, K.G.Shin.Efficient Precomputation of Quality-of-Service Routes Proceedings of IEEE NOSSDAV 98, July, 1998 [12] A Varga,The OMNET++ Discrete Event Simulation System, the European Simulation Multiconference, Prague, Czech Republic, 2001 [13] T A Al Ghamdi, M E.Bandwidth as a dominant metric in localized QoS algorithm", in Proceedings of the 17th Telecommunications forum TELFOR 2009, pp 145-148, 2009 [14] T A Al Ghamdi, M E Woodward, "Novel algorithms for QoS localized routing in communication networks", in Proceedings of First Asian Himalayas International Conference on Internet, Kathmandu, Nepal, pp 1-7, 2009 [15] K Kar, M Kodialam, T V Lakshman,Minimum Interference Routing of Bandwidth Guaranteed Tunnels with MPLS Traffic Engineering Applications, IEEE Journal on Selected Areas in Communications , Vol.18, No 12, December 2000pp 2566-2579 TÓM TẮT ẢNH HƯỞNG CỦA CƠ CHẾ LINH HOẠT VỚI THUẬT TỐN ĐỊNH TUYẾN PHÂN TÁN DÙNG THƠNG TIN NỘI BỘ Ngày nay, yêu cầu chất lượng mạng lưới ngày trở nên phổ biến phức tạp để đáp ứng nhu cầu thị trường viễn thông Để đảm bảo cho chất lượng mạng, kiểu thuật tốn định tuyến đảm bảo QoS sử dụng thơng tin nội nút để định tuyến thông tin nghiên cứu nhiều nay, giải pháp đầy hứa hẹn cho kiểu thuật tốn đảm bảo QoS truyền thống Với thơng tin thu thập nội nút để đưa định chọn đường, thuật tốn định tuyến dùng thơng tin nội khắc phục nhiều nhược điểm kiểu định tuyến theo truyền thống như: thơng tin tồn cục khơng xác, thơng tin khơng cập nhật Bài báo nghiên cứu đề xuất thuật tốn định tuyến dùng thơng tin nội sử dụng kiểu định tuyến phân tán để định tuyến thông tin Bài báo so sánh hoạt động thuật toán với thuật toán định tuyến khác với mơi trường thí nghiệm Qua đó, báo cho thấy hiệu tốt thuật toán định tuyến đảm bảo QoS sử dụng thông tin nội ứng dụng kiểu định tuyến phân tán chế linh hoạt thuật tốn Từ khóa: Giải thuật; Phân tán; Nội bộ; QoS; Định tuyến Nhận ngày 14 tháng 11 năm 2018 Hoàn thiện ngày 20 tháng 12 năm 2018 Chấp nhận đăng ngày 19 tháng 02 năm 2019 Author affiliations: Posts and Telecommunications Institute of Technology, Hanoi City, Vietnam *Corresponding author: anhtm.dng@vnpt.vn Tạp chí Nghiên cứu KH&CN quân sự, Số 59, 02 - 2019 37 ... C Trinh hop- by by -hop hop routing algorithm algorithm.”” Trinh, Impact Impact of flexible mechanism … hop Nghiên cứu khoa học công nghệ Figure Network of 32 nodes From [10, 11], the offered... hop- by by -hop hop routing algorithm algorithm.”” Trinh, Impact Impact of flexible mechanism … hop Nghiên ccứu ứu khoa học công nghệ Therefore, we can see that with the type of distributed routing, ... performance of routing algorithms may vary across different load conditions, our simulations consider several types of different load conditions through the value of λ according to experiments of from