A5050245571- Cloud Computing and Resource Management
- Caching and Content Delivery
- Software-Defined Networks and 5G
- Peer-to-Peer Network Technologies
- Advanced Data Storage Technologies
- Distributed and Parallel Computing Systems
- Interconnection Networks and Systems
- Distributed systems and fault tolerance
- Software System Performance and Reliability
- Network Traffic and Congestion Control
- Parallel Computing and Optimization Techniques
- Advanced Optical Network Technologies
- Network Security and Intrusion Detection
- Opportunistic and Delay-Tolerant Networks
- Internet Traffic Analysis and Secure E-voting
- Complex Network Analysis Techniques
- Optimization and Search Problems
- IoT and Edge/Fog Computing
- Green IT and Sustainability
- Wireless Networks and Protocols
- Mobile Ad Hoc Networks
- Optical Network Technologies
- Scientific Computing and Data Management
- Advanced Memory and Neural Computing
- Real-Time Systems Scheduling
Google (United States)
2015-2024
University of California, San Diego
2006-2015
University of California, Berkeley
1993-2014
Stanford University
2014
University of California System
2011-2013
UC San Diego Health System
2004-2012
University of Cyprus
2010
Universitat de Barcelona
2010
Microsoft (United States)
2009
Duke University
1999-2006
Today's data centers may contain tens of thousands computers with significant aggregate bandwidth requirements. The network architecture typically consists a tree routing and switching elements progressively more specialized expensive equipment moving up the hierarchy. Unfortunately, even when deploying highest-end IP switches/routers, resulting topologies only support 50% available at edge network, while still incurring tremendous cost. Non-uniform among center nodes complicates application...
P4 is a high-level language for programming protocol-independent packet processors. works in conjunction with SDN control protocols like OpenFlow. In its current form, OpenFlow explicitly specifies protocol headers on which it operates. This set has grown from 12 to 41 fields few years, increasing the complexity of specification while still not providing flexibility add new headers. this paper we propose as strawman proposal how should evolve future. We have three goals: (1)...
We present the design, implementation, and evaluation of B4, a private WAN connecting Google's data centers across planet. B4 has number unique characteristics: i) massive bandwidth requirements deployed to modest sites, ii) elastic traffic demand that seeks maximize average bandwidth, iii) full control over edge servers network, which enables rate limiting measurement at edge.
Today's data centers may contain tens of thousands computers with significant aggregate bandwidth requirements. The network architecture typically consists a tree routing and switching elements progressively more specialized expensive equipment moving up the hierarchy. Unfortunately, even when deploying highest-end IP switches/routers, resulting topologies only support 50% available at edge network, while still incurring tremendous cost. Non-uniform among center nodes complicates application...
Today's data centers offer tremendous aggregate bandwidth to clusters of tens thousands machines. However, because limited port densities in even the highest-end switches, center topologies typically consist multi-rooted trees with many equal-cost paths between any given pair hosts. Existing IP multipathing protocols usually rely on per-flow static hashing and can cause substantial losses due long-term collisions.In this paper, we present Hedera, a scalable, dynamic flow scheduling system...
Internet hosting centers serve multiple service sites from a common hardware base. This paper presents the design and implementation of an architecture for resource management in center operating system, with emphasis on energy as driving issue large server clusters. The goals are to provision resources co-hosted services way that automatically adapts offered load, improve efficiency clusters by dynamically resizing active set, respond power supply disruptions or thermal events degrading...
This paper considers the requirements for a scalable, easily manageable, fault-tolerant, and efficient data center network fabric. Trends in multi-core processors, end-host virtualization, commodities of scale are pointing to future single-site centers with millions virtual end points. Existing layer 2 3 protocols face some combination limitations such setting: lack scalability, difficult management, inflexible communication, or limited support machine migration. To extent, these may be...
The basic building block of ever larger data centers has shifted from a rack to modular container with hundreds or even thousands servers. Delivering scalable bandwidth among such containers is challenge. A number recent efforts promise full bisection between all servers, though significant cost, complexity, and power consumption. We present Helios, hybrid electrical/optical switch architecture that can deliver reductions in the switching elements, cabling, consumption relative recently...
We present the design, implementation, and evaluation of B4, a private WAN connecting Google's data centers across planet. B4 has number unique characteristics: i) massive bandwidth requirements deployed to modest sites, ii) elastic traffic demand that seeks maximize average bandwidth, iii) full control over edge servers network, which enables rate limiting measurement at edge. These characteristics led Software Defined Networking architecture using OpenFlow relatively simple switches built...
In recent years, overlay networks have become an effective alternative to IP multicast for efficient point multipoint communication across the Internet. Typically, nodes self-organize with goal of forming tree, one that meets performance targets without placing undue burden on underlying network. this paper, we target high-bandwidth data distribution from a single source large number receivers. Applications include large-file transfers and real-time multimedia streaming. For these...
We present our approach for overcoming the cost, operational complexity, and limited scale endemic to datacenter networks a decade ago. Three themes unify five generations of detailed in this paper. First, multi-stage Clos topologies built from commodity switch silicon can support cost-effective deployment building-scale networks. Second, much general, but complex, decentralized network routing management protocols supporting arbitrary scenarios were overkill single-operator, pre-planned...
This paper presents ModelNet, a scalable Internet emulation environment that enables researchers to deploy unmodified software prototypes in configurable Internet-like and subject them faults varying network conditions. Edge nodes running user-specified OS application are configured route their packets through set of ModelNet core nodes, which cooperate the traffic bandwidth, congestion constraints, latency, loss profile target topology.This describes evaluates architecture its...
Datacenter transports aim to deliver low latency messaging together with high throughput. We show that simple packet delay, measured as round-trip times at hosts, is an effective congestion signal without the need for switch feedback. First, we advances in NIC hardware have made RTT measurement possible microsecond accuracy, and these RTTs are sufficient estimate queueing. Then describe how TIMELY can adjust transmission rates using gradients keep while delivering bandwidth. implement our...
Energy consumption has recently been widely recognized as a major challenge of computer systems design. This paper explores how to support energy first-class operating system resource. Energy, because its global nature, presents challenges beyond those conventional resource management. To meet these we propose the Currentcy Model that unifies accounting over diverse hardware components and enables fair allocation available among applications. Our particular goal is extend battery lifetime by...
This paper considers the requirements for a scalable, easily manageable, fault-tolerant, and efficient data center network fabric. Trends in multi-core processors, end-host virtualization, commodities of scale are pointing to future single-site centers with millions virtual end points. Existing layer 2 3 protocols face some combination limitations such setting: lack scalability, difficult management, inflexible communication, or limited support machine migration. To extent, these may be...
The primary motivation for enterprises to adopt virtualization technologies is create a more agile and dynamic IT infrastructure -- with server consolidation, high resource utilization, the ability quickly add adjust capacity on demand while lowering total cost of ownership responding effectively changing business conditions. However, effective management virtualized environments introduces new unique requirements, such as dynamically resizing migrating virtual machines (VMs) in response...
The basic building block of ever larger data centers has shifted from a rack to modular container with hundreds or even thousands servers. Delivering scalable bandwidth among such containers is challenge. A number recent efforts promise full bisection between all servers, though significant cost, complexity, and power consumption. We present Helios, hybrid electrical/optical switch architecture that can deliver reductions in the switching elements, cabling, consumption relative recently...
Recent proposals have employed optical circuit switching (OCS) to reduce the cost of data center networks. However, relatively slow times (10--100 ms) assumed by these approaches, and accompanying latencies their control planes, has limited its use only largest networks with highly aggregated constrained workloads. As faster switch technologies become available, designing a plane capable supporting them becomes key challenge.
Modern data centers are massive, and support a range of distributed applications across potentially hundreds server racks. As their utilization bandwidth needs continue to grow, traditional methods augmenting have proven complex costly in time resources. Recent measurements show that center traffic is often limited by congestion loss caused short bursts. Thus an attractive alternative adding physical augment wired links with wireless the 60 GHz band.