国外留学 学习生活记录（5）
10/22/2012 - 10/23/2012
I had a talk with Prof. Yu regarding to what I was doing in this few days, and she recommanded me to take talks with Hao Di. Already we had a discussion with each other last Friday concerning the calculating method of virutal network request reliability. The literatures he had read mostly assume that the probability of each server is fix and independent, while I had saw, for a instance Anne's model, it is unknow. In Anne's model, it just knows the arrival rate of hardware failure which follows Poisson distribution. The both two mehtods are all correct. They take into the same problem from different point of view.
And these two day, I was mainly reading the paper Hao Di sent to me last Friday about considering the reliability during the VN mapping. The one of these is cited by the paper what I had read last weak, and I'm reading it. It is a very interesting paper. It consideres the backup node sharing with different VNs. In this paper, the authors assume the VN n0 having enough backup nodes could share its backup node with other VNs. For each VN except n0, it also has the enough backup nodes shared from n0 to meet its reliability requirement, and they don't share their backup nodes with each other. Therefor there is no difference between having own backup nodes for them. The only difference in this share scheme is that the reliability of n0 will change. Therefor authors recalculate the reliability for n0 when it shares its backup nodes with other VNs. Using this method, the purpose of this paper it maximally improve the network capacity which could accomodate more VNs under the guarantee of reliability for each VN including n0.
I had a talk with Prof. Yu regarding to what I was doing in this few days, and she recommanded me to take talks with Hao Di. Already we had a discussion with each other last Friday concerning the calculating method of virutal network request reliability. The literatures he had read mostly assume that the probability of each server is fix and independent, while I had saw, for a instance Anne's model, it is unknow. In Anne's model, it just knows the arrival rate of hardware failure which follows Poisson distribution. The both two mehtods are all correct. They take into the same problem from different point of view.
And these two day, I was mainly reading the paper Hao Di sent to me last Friday about considering the reliability during the VN mapping. The one of these is cited by the paper what I had read last weak, and I'm reading it. It is a very interesting paper. It consideres the backup node sharing with different VNs. In this paper, the authors assume the VN n0 having enough backup nodes could share its backup node with other VNs. For each VN except n0, it also has the enough backup nodes shared from n0 to meet its reliability requirement, and they don't share their backup nodes with each other. Therefor there is no difference between having own backup nodes for them. The only difference in this share scheme is that the reliability of n0 will change. Therefor authors recalculate the reliability for n0 when it shares its backup nodes with other VNs. Using this method, the purpose of this paper it maximally improve the network capacity which could accomodate more VNs under the guarantee of reliability for each VN including n0.
评论
发表评论