Source author record

Wenzhi Cui

Wenzhi Cui appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Catalog footprint

What is connected

2works
1topics
1close collaborators

Actions

Connect this record

Log in to claim

Research graph

See the researcher in context

Open full explorer

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

2 published item(s)

preprint2014arXiv

Dual-structure Data Center Multicast Using Software Defined Networking

Data center applications use multicast as an effective method to reduce bandwidth cost. However, traditional multicast protocols designed for IP networks are usually bottlenecked by the limited state capacity on switches. In this paper, we propose a scalable multicast solution on fat tree networks based on the observation that data center multicast traffic has strong heterogeneity. We propose to remove the multicast management logic from switches and use the SDN controller to manage multicast groups. The proposed Dual-structure Multicast (DuSM) determines elephant and mice groups according to their traffic amounts and treats them separately. For each elephant group, the controller installs multicast state to maintain multiple shared trees and the group traffic will be balanced evenly among the trees to avoid congestion. For mice groups, the controller applies state-free mutlicast that trades bandwidth capacity for state capacity, such as multicast-to-unicast translation. Our experiments using real multicast traffic data show that the number of groups DuSM supports can be 300% of that of IP multicast. DuSM also achieves traffic balance among links.

preprint2013arXiv

DiFS: Distributed Flow Scheduling for Data Center Networks

Data center networks leverage multiple parallel paths connecting end host pairs to offer high bisection bandwidth for cluster computing applications. However, state of the art distributed multi-pathing protocols such as Equal Cost Multipath (ECMP) use static flow-to-link assignment, which is load-oblivious. They may cause bandwidth loss due to \emph{flow collisions} on a same link. Recently proposed centralized scheduling algorithm or host-based multi-pathing may suffer from scalability problems. In this paper, we present Distributed Flow Scheduling (DiFS) for data center networks, which is a switch-only distributed solution. DiFS allows switches cooperate to avoid over-utilized links and find available paths without centralized control. DiFS is scalable and can react quickly to dynamic traffic, because it is independently executed on switches and requires no synchronization. Extensive experiments show that the aggregate bisection bandwidth of DiFS using various traffic patterns is much better than that of ECMP, and is similar to or higher than that of a recent proposed centralized scheduling algorithm.