دانلود رایگان مقاله طرح دو مرحله ای چندمسیره با عملکرد بالا برای شبکه دیتا سنتر

Abstract

Multipath forwarding has been recently proposed to improve utilization in data centers leveraged by its redundant network design. However, most multipath proposals require significant modifications to the tenants’ network stack and therefore are only feasible in private clouds. In this paper, we propose the Two-Phase Multipath (TPM) forwarding scheme for public clouds. The proposal improves tenants’ network throughput, whereas keeping unmodified network stack on tenants. Our scheme is composed of a smart offline configuration phase that discover optimal disjoint paths, and a fast online path selection phase that improves flow throughput at run time. A logically centralized manager uses a genetic algorithm to generate and install sets of paths, summarized into trees, during multipath configuration, and a local controller performs the multipath selection based on network usage. We analyze TPM for different workloads and topologies under several scenarios of usage database locations and update policies. The results show that our proposal yields up to 77% throughput gains over previously proposed approaches.

7. Conclusion

The main goal of the proposed two-phase multipath (tpm) scheme is to improve the network performance of cloud computing data centers with no modification to the tenants’ network stack requiring no modifications on the data-center infrastructure. We formulate the multipath forwarding problem as two different problems: generating disjoint multiple trees, and selecting the tree that contains the least used links to forward a new flow. We improve the network performance summarizing a set of paths on VLAN trees, taking advantage of the VLAN mechanism that is commonly available in commercial off-the-shelf switches. tpm divides the forwarding into two phases: an offline multipath configuration phase that calculates available multipaths and install them in the virtual switches, and a fast online multipath selection phase to distribute flows among the available paths. The multipath configuration is based on a genetic algorithm proposed to find disjoint VLAN trees connecting all ToR switches. Our formulation of the genetic algorithm uses an innovative representation of trees as a 2-dimenson vector, which enable a fast convergence of the algorithm. Our initial solution, before optimization, creates already randomized disjoint tree, which fasts the genetic algorithm into reaching the optimal solution. The fast online multipath selection phase uses heuristics based on network usage to select the path for a new flow. The path selection heuristic may use either a local or a global database to keep track of link usage. We demonstrate through simulations that the proposed tpm scheme has better performance than the conventional equal cost multipath (ecmp) scheme in high workload scenarios. The experiments show that our formulation of genetic algorithm is able to find a set of completely disjoint trees, 100% disjoint trees, in typical datacenter topology within a few iterations. The simulation results show that the tpm achieves up to 77% and 27% gains when compared to stp and ecmp, respectively.