Skip to content

The gossip protocol has been implemented for group communication and aggregate computation. Here through a simulator, convergence of the algorithms is determined. The algorithms used here are GOSSIP ALGORITHM and PUSH-SUM PROTOCOL. Along with the algorithms, topology also plays a crucial role and the topologies used here are: 1) FULL NETWORK 2) …

Notifications You must be signed in to change notification settings

meghasaik/GOSSIP-PROTOCOL-

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 

Repository files navigation

GOSSIP-PROTOCOL-

The gossip protocol has been implemented for group communication and aggregate computation. Here through a simulator, convergence of the algorithms is determined. The algorithms used here are GOSSIP ALGORITHM and PUSH-SUM PROTOCOL. Along with the algorithms, topology also plays a crucial role and the topologies used here are: 1) FULL NETWORK 2) LINE 3)3-D TORUS 4) HONEYCOMB 5) HONEYCOMB WITH RANDOM NEIGHBOURS 6)RANDOM 2-D GRID.

PROJECT REPORT

The implementation has been done in both Gossip as well as Push-sum algorithms, which covers the following topologies: Line, Fully Connected, Random-2D, 3D Torus, Honeycomb and Random Honeycomb. Below are the findings: ##TABLE:

                                                          ##GOSSIP (Time in milliseconds)

                                      
                                          Inputs	LINE	  FULL	  Rand2D	  3DTorus	  Honeycomb	  RandomHoneycomb
                                            50	  5359	  2031	  1734	      1735	    2219	        1875
                                            100	  11203	  2204	  2328	      1687	    4594	        3110
                                            500	  58422	  2859	  2156	      2484	    3563	        2672
                                            1000	81218	  16110	  2844	      4344	    3843	        2813
                                            2500	173719	507625	5578	      2422	    7047	        3937
                                            5000	389937		-     4250	      3422	    6547	        3172

##TABLE: ##PUSH-SUM (Time in milliseconds)

                                           Inputs	    LINE	  FULL	Rand2D	3DTorus	  Honeycomb	  RandomHoneycomb
                                             50	      2234	  610	   881	    703	        875	         1047
                                            100	      3687	  1094	 1110	    672	       4110	         1546
                                            500	      2937	  2097	  937	   2094	       8000	         2921
                                            1000	    3690	  6562	 1406	   3297	      11313	         2728
                                            2500			 -       -     1578	   5375	      14297	         4438
                                            5000			 -       -     2015	  15422	      15578	         4953

#OBSERVATIONS- • All the topologies implemented in this project, work well. • The least-time consuming algorithm for push-sum turns out to be random2D grid. It converges faster than others for higher number of nodes. The highest number of nodes that converged without any failure were: 5000. • The random honeycomb algorithm used in this project has four nodes instead of three where three nodes are adjacent, and one node is a random one. Because of randomization, this topology works quite better than the other topologies used. It works better than the fully connected topology when higher number of nodes are taken for both gossip and push-sum algorithm. • For the push-sum algorithm, the fully connected topology does not work beyond 2500 nodes when given. The maximum number of nodes that were working was 1200 nodes. • The most-time consuming algorithm for both gossip as well as the push-sum was the line topology. It was the slowest in convergence when compared with other topologies. The reason was in the process of its implementation. • For the push-sum algorithm, the line topology when compared to other topologies performed badly, as the convergence failure occurred from the 800th node itself. Thus, it could provide convergence for 800 nodes. • In gossip algorithm, when taking higher number of nodes, the fully connected topology performs the worst and provides the highest convergence time when compared to the other topologies.

About

The gossip protocol has been implemented for group communication and aggregate computation. Here through a simulator, convergence of the algorithms is determined. The algorithms used here are GOSSIP ALGORITHM and PUSH-SUM PROTOCOL. Along with the algorithms, topology also plays a crucial role and the topologies used here are: 1) FULL NETWORK 2) …

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages