Symbiotic routing in future data centers

Abstract

Building distributed applications that run in data centers is hard. The CamCube project explores the design of a shipping container sized data center with the goal of building an easier platform on which to build these applications. CamCube replaces the traditional switch-based network with a 3D torus topology, with each server directly connected to six other servers. As in other proposals, e.g. DCell and BCube, multi-hop routing in CamCube requires servers to participate in packet forwarding. To date, as in existing data centers, these approaches have all provided a single routing protocol for the applications.In this paper we explore if allowing applications to implement their own routing services is advantageous, and if we can support it efficiently. This is based on the observation that, due to the flexibility offered by the CamCube API, many applications implemented their own routing protocol in order to achieve specific application-level characteristics, such as trading off higher-latency for better path convergence. Using large-scale simulations we demonstrate the benefits and network-level impact of running multiple routing protocols. We demonstrate that applications are more efficient and do not generate additional control traffic overhead. This motivates us to design an extended routing service allowing easy implementation of application-specific routing protocols on CamCube. Finally, we demonstrate that the additional performance overhead incurred when using the extended routing service on a prototype CamCube is very low. * Work done during internship at Microsoft Research, Cambridge. Currently at

Agile Traffic Merging for DCNs

“…These hierarchical interconnection topologies are designed to maximize cross-section bandwidth and optimize the cost-effect ratio. Alternatively, some server-centric DCN architecture, such as DCell [10], BCube [9] and CamCube [2], use simple switches and push network routing to the servers, thus obtaining better scalability and fault-tolerance. In general, these proposed network topologies are intended to support increasing number of servers and provide high capacity for bandwidth-hungry services.…”

Survey on Data Center Networking Technologies

“…Abu-Libdeh et al [26] proposed CamCube, which is a platform for developing application specific routing protocols in 3D torus topology environments. The API is inspired by the Key-Based Routing (KBR) API, which is used in many structured overlay.…”

Cost, performance & flexibility in OpenFlow: Pick three

“…For example, MPLS has been used to develop a mechanism to establish and dynamically manage paths in datacenter networks [5]. Alternatively, CamCube [6] tries to address the problem of network control by removing all network devices, providing networkwide connectivity through direct server interconnection using 4 http://openmirage.org multi-port NICs. By implementing routing in servers, such designs provide fine level network programmability and enable development and deployment of clean slate network protocols.…”

“…A growing number of current system security exploits are caused by layer complexity 6 and thus this poses a direct threat to further deployment of cloud infrastructure. b) Efficiency: Support for legacy functionality in the OS generally wastes energy and CPU, resulting in increased cost without benefit for the hosted services.…”

“…The core problem can be traced to the lack of a bi-directional signalling channel between the application layer and the intermediary devices in the network. 6 E.g., http://technet.microsoft.com/en-us/security/bulletin/ms11-083 7 http://nodejs.org Such a channel could convey fine-grained information regarding path load to applications, and the per-flow requirements of applications to the network devices in return. To bridge this gap we advocate direct integration of a controller into the application.…”