**4.1 Simulation settings**

We performed the simulation in the MATLAB environment. Our resource discovery tree is a full n-ary tree. It means that any non leaf node should have exactly n children. According to Mastroianni et al. (2007), and Chang & Hu (2010), we also assume that the resource discovery tree for simulation has height 4.

In the first experiment tests, we compare our algorithm with a resource discovery tree with a different number of index servers. Like (Chang & Hu, 2010), in this experiment there are 200 nodes in resource discovery tree and we perform this experiment with 180 queries.

We place the resources randomly in each node and then queries are sent through tree paths and compare the number of nodes that visited in each method. In Fig. 8, the difference between the number of visited nodes with two methods are observed. In this experiment, the number of visited nodes is investigated with changing the number of index servers. For example, when the number of index servers in tree is 10, so there are 10 nodes in level 1 and 190 nodes in level 2 (19 children for each node in level 1) for a tree with height 3. Because our method in the forward path just visits one node in every level so in Figs. 8, the simulations related to our method almost show the fix values.

In the second simulation tests, we assume there are 300 queries and a tree with height 4. In Fig. 9, we compare the number of nodes that queries send in our algorithm with the resource discovery tree and show that the number of nodes visited in our proposed method is lower than the previous method.

In the third simulation tests, we show the total number of nodes that were visited in the resource discovery path and for tree updating with assume 400 queries and different number of nodes for our method and compare with the other one. In Fig. 10, it is indicated that in our algorithm, fewer nodes are visited compared with the previous one.

82 Grid Computing – Technology and Applications, Widespread Coverage and New Horizons

In this section, we compare our simulation results with the recommended algorithm in (Chang & Hu, 2010), flooding-based approach and the algorithm proposed in (Marzolla et al., 2005; Marzolla et al., 2007) (MMO). In the first simulation tests, the total number of nodes which are visited during in the resource discovery and in updating are compared with algorithm (Chang & Hu, 2010) for 400 queries. In the second experiment, the nodes which queries send during resource discovery for the flooding-based approach, MMO algorithm and resource discovery

As we know, for calculating the traffic which a method causes, we can calculate the number of links which are occupied for resource discovery or updating in the method. If the number of links visited for resource discovery or updating in a method is lower, the method has

In the next simulation tests, the traffic in our method which is caused by the increased number of tree nodes in resource discovery and updating is indicated and the results are obtained for different nodes with 300 and 1000 queries. In the last simulation, we observe that the traffic caused in our method would be lower than other methods. This test is performed supposing there are 300 queries for different nodes, and the flooding-based,

We performed the simulation in the MATLAB environment. Our resource discovery tree is a full n-ary tree. It means that any non leaf node should have exactly n children. According to Mastroianni et al. (2007), and Chang & Hu (2010), we also assume that the resource

In the first experiment tests, we compare our algorithm with a resource discovery tree with a different number of index servers. Like (Chang & Hu, 2010), in this experiment there are 200 nodes in resource discovery tree and we perform this experiment with 180 queries.

We place the resources randomly in each node and then queries are sent through tree paths and compare the number of nodes that visited in each method. In Fig. 8, the difference between the number of visited nodes with two methods are observed. In this experiment, the number of visited nodes is investigated with changing the number of index servers. For example, when the number of index servers in tree is 10, so there are 10 nodes in level 1 and 190 nodes in level 2 (19 children for each node in level 1) for a tree with height 3. Because our method in the forward path just visits one node in every level so in Figs. 8, the

In the second simulation tests, we assume there are 300 queries and a tree with height 4. In Fig. 9, we compare the number of nodes that queries send in our algorithm with the resource discovery tree and show that the number of nodes visited in our proposed method

In the third simulation tests, we show the total number of nodes that were visited in the resource discovery path and for tree updating with assume 400 queries and different number of nodes for our method and compare with the other one. In Fig. 10, it is indicated

that in our algorithm, fewer nodes are visited compared with the previous one.

**4. Simulation results** 

**4.1 Simulation settings** 

tree algorithm, are shown and compared with our method.

MMO, resource discovery tree and our methods are compared.

simulations related to our method almost show the fix values.

lower traffic and would be more efficient.

discovery tree for simulation has height 4.

is lower than the previous method.

Fig. 8. The number of nodes that queries are forwarded to for 180 queries.

Fig. 9. The number of nodes that queries are forwarded to.

A New Approach to Resource Discovery in Grid Computing 85

Fig. 12. Number of links that resource discovery queries and updates are forwarded to for

Fig. 13. Number of links that resource discovery queries and updates are forwarded to for

300 queries.

1000 queries.

Fig. 10. Total number of nodes that are visited in resource discovery and updates.

In the next simulation tests, our method is compared with flooding-based method, MMO and resource discovery tree algorithm. In the current experiment supposing that there are 300 queries, in Fig. 11, it is indicated that the average number of nodes that queries are sent to is lower than other methods in our proposed method. The test is performed in a tree with height 4.

Fig. 11. Average number of nodes that queries are forwarded to using different methods.

In the next simulation tests, the number of occupied links in our method during the resource discovery phase and update phase is observed for different nodes with 300 and 1000 queries. In Fig. 12, we show the occupied links (traffic) for resource discovery and updating for 300 (Fig. 12) and 1000 queries (Fig. 13).

84 Grid Computing – Technology and Applications, Widespread Coverage and New Horizons

Fig. 10. Total number of nodes that are visited in resource discovery and updates.

height 4.

In the next simulation tests, our method is compared with flooding-based method, MMO and resource discovery tree algorithm. In the current experiment supposing that there are 300 queries, in Fig. 11, it is indicated that the average number of nodes that queries are sent to is lower than other methods in our proposed method. The test is performed in a tree with

Fig. 11. Average number of nodes that queries are forwarded to using different methods.

for 300 (Fig. 12) and 1000 queries (Fig. 13).

In the next simulation tests, the number of occupied links in our method during the resource discovery phase and update phase is observed for different nodes with 300 and 1000 queries. In Fig. 12, we show the occupied links (traffic) for resource discovery and updating

Fig. 12. Number of links that resource discovery queries and updates are forwarded to for 300 queries.

Fig. 13. Number of links that resource discovery queries and updates are forwarded to for 1000 queries.

A New Approach to Resource Discovery in Grid Computing 87

Al-Dmour, N.A. & Teahan, W.J. (2005). Peer-to-Peer protocols for resource discovery in the grid, in: *Parallel and Distributed Computing and Networks*, Innsbruck, Austria. Ali, K.; Datta, S.; Aboelaze, M. (2005). Grid resource discovery using small world overlay

Basu, S.; Banerjee, S.; Sharma, P.; Lee, S. (2005). NodeWiz: peer-topeer resource discovery for

Cai, M. & Hwang, K. (2007). Distributed aggregation algorithms with load- balancing for

Chang, R.-S. & Hu, M.-S. (2010). A resource discovery tree using bitmap for grids, *Future* 

Chien, A.; Calder, B.; Elbert, S. & Bhatia, K. (2003). Entropia: Architecture and performance of an enterprise desktop grid system, J. Parallel Distrib. Comput. 63 (5). Deng, Y.; Wang, F.; Ciura, A. (2009). Ant colony optimization inspired resource discovery in

Foster, I. & Kesselman, C. (1997). Globus: A metacomputing infrastructure toolkit, *Int. J.* 

Foster, I. & Kesselman, C. (2003). The Grid 2: Blueprint for a New Computing Infrastructure,

Germain, C.; Neri, V.; Fedak, G.; Cappello, F. (2000). XtremWeb: Building an experimental platform for global computing, in: Proc. of IEEE/ACM Grid, December. Iamnitchi, A.; Foster, I.; Nurmi, D. C. (2002). A peer-to-peer approach to resource location in

Iamnitchi, A. & Talia, D. (2005). P2P computing and interaction with Grids, *Future Gener.* 

Khanli, L.M. & Kargar, S. (2011). FRDT: Footprint Resource Discovery Tree for grids, Future

Koo, S.G.M.; Kannan, K. & Lee, C.S.G. (2006). On neighbor selection strategy in hybrid Peer-

Li, J. (2010). Grid resource discovery based on semantically linked virtual organizations,

Li, J. & Vuong, S. (2005). Grid resource discovery using semantic communities, in: *Proceedings of the 4th International Conference on Grid and Cooperative Computing*, Beijing, China Liu, H.; Luo, P. & Zeng, Z. (2006). A structured hierarchical P2P model based on a rigorous

Marzolla, M.; Mordacchini, M. & Orlando, S. (2005). Resource discovery in a dynamic

environment, in: *Proceedings of the 16th International Workshop on Database and Expert* 

grid environments, in: Proceedings of the 11th Symposium on High Performance

Berman, F. & et al. (2003). Adaptive computing on the grid using AppLeS, TPDS 14 (4). Bharambe, A.R.; Agrawal, M. & Seshan, S. (2004). Mercury: Supporting scalable multiattribute range queries, in: *Proc. of ACM SIGCOMM*, pp. 353–366. Bradley, A.; Curran, K.; Parr, G., (2006). Discovering resource in computational GRID

graphs, in: Proceedings of the 18th IEEE Canadian Conference on Electrical and

grids, in: 5th International Workshop on Global and Peer-to-Peer Computing

**6. References** 

Computer Engineering.

(GP2PC) in Conjunction with CCGrid.

*Gener. Comput. Syst.* 26 29–37.

*Comput. Syst.* 21 (3) 331–332.

Gener. Comput. Syst. 27 148–156.

*Future Gener. Comput. Syst.* 26, 361–373.

*Systems Applications*, DEXA'05, pp. 356–360.

environments. Journal of Supercomputing 35, 27–49.

scalable grid resource monitoring, in: *Proc. of IPDPS*.

Morgan Kaufmann Publishers Inc., San Francisco, CA.

Distributed Computing, Edinburgh, UK, August.

to-Peer networks, *Future Gener. Comput. Syst.* 22, 732–741.

binary tree code algorithm, *Future Gener. Comput. Syst*.

P2P Grid systems, J Supercomput 49: 4–21.

*High Perform. Comput. Appl.* 2 115–128.

Fig. 14. Number of links that are visited by resource discovery queries for 300 queries.

In the last experiment, we supposed that there are 300 queries, and we show the visited links (traffic) which are caused during resource discovery in our method and compared with flooding-based, MMO and the resource discovery tree and for different numbers of nodes. In Fig. 14, we can see the traffic caused in our method is lower than other methods
