Partial Overlapping Chunk Based Mutiple-Path Transmission

Watching online high definition video,consuming a lot of bandwidth,becomes a common network behavior.It is a possible way by aggregating the bandwidths of multiple paths to meet the high throughput demand of video data,so as to further improve user watching experience.In the existing research work on multipath transmission,a classical method uses all multiple paths to redundantly transmit a data packet,for assuring timely arrival of the data.However,it wastes too many network resources.Another classical method,for assuring timely and orderly arrival of data packet,transmits different data along different paths,where data length is proportionally sized according to the predicted bandwidth of each path.However,since bandwidth follows Gaussian distribution,compared with the head of a transmitted data packet,the tail has a lower possibility of arriving on time.As long as any piece of data does not arrive on time,the playback quality will be affected.In this paper,we propose a new idea of multipath transmission,to further improve the probability of data arrival on time.According to this idea,the whole download process within a client consists of many continuous time cycles.In each cycle,the client will download a certain length of data.Assume Gaussian distribution of bandwidth of a path.Since the tail of data has a lower probability to be downloaded than the head,we need to adjust the probability by redundantly transmitting some data of the tail,so as to optimize the timely arrival probability of the whole data in a cycle.The main contributions of this paper are as follows:(1)We establish a partial overlapping chunk based multipath transmission model.With this model,we clearly describe the idea about how to optimize the timely arrival probability of the whole data in a cycle.(2)We propose a new multipath transmission method named PCT(Partial Overlapping Chunk Based Mutiple-Path Transmission).To realize the above optimization objective,in each cycle,PCT sorts the available paths according to their transmission capacities(means and variances of bandwidth),assign them with different data segments,which have the same end positions but have different start positions.We formulate a optimization function to mathematically compute these optimal starting positions.(3)We prove the performance of PCT by theoretical analysis and a serial of simulations.Simulation results demonstrate that,our method can improve timely arrival probability by up to 19.3%,compared to the traditional scheme without chunk overlapping,at the cost of less than 14.4%of redundancy.Furthermore,the result of real network based simulation shows that,PCT has significant advantage over the traditional scheme in small aggregation bandwidth situation.