No abstract

Virtual machine technology and the ease with which VMs can be migrated within the LAN, has changed the scope of resource management from allocating resources on a single server to manipulating pools of resources within a data center. We expect WAN migration of virtual machines to likewise transform the scope of provisioning compute resources from a single data center to multiple data centers spread across the country or around the world. In this paper we present the CloudNet architecure as a cloud framework consisting of cloud computing platforms linked with a VPN based network infrastructure to provide seamless and secure connectivity between enterprise and cloud data center sites. To realize our vision of efficiently pooling geographically distributed data center resources, CloudNet provides optimized support for live WAN migration of virtual machines. Specifically, we present a set of optimizations that minimize the cost of transferring storage and virtual machine memory during migrations over low bandwidth and high latency Internet links. We evaluate our system on an operational cloud platform distributed across the continental US. During simultaneous migrations of four VMs between data centers in Texas and Illinois, CloudNet's optimizations reduce memory migration time by 65% and lower bandwidth consumption for the storage and memory transfer by 19GB, a 50% reduction.

etworks have traditionally been composed of interconnected hardware such as routers, switches, and firewalls. Employing purposebuilt hardware appliances, managed through distributed protocols, has allowed networks to achieve high performance and reliability, but it comes at the cost of limited flexibility. This tradition has mostly continued, with such purpose-specific hardware systems being deployed even for typical software functions such as proxies, firewalls, and caches, because of the desire to have a high-performance data plane. This limits the flexibility of network functions, has high cost, and makes it difficult to deploy services dynamically.Cloud data centers have increased their efficiency and flexibility by employing virtualization techniques that allow convenient (often centralized) management of dynamically created server instances. With the adoption of software defined Nnetworking (SDN) and network function virtualization (NFV), a similar revolution is happening in both wide area networks and data center networks. SDN provides a logically centralized control plane that can flexibly direct packet flows between network devices based on programmable policies [1-3]. NFV transforms networks from hardware appliances with customized application-specific integrated circuits (ASICs) into software running in VMs (VMs) on common off-the-shelf (COTS) hardware to increase flexibility and lower cost [4-6]. Together, SDN and NFV offer the potential to radically alter how networks are deployed and managed.By moving to a software-based environment, NFV makes network services easy to deploy, and allows them to be more powerful and flexible, enabling more complex topologies and feature-rich network resident functions compared to hardware-based implementations. Unfortunately, the performance limitations of commodity hardware and the overhead of server virtualization platforms have prevented high-performance network processing to fully transition away from hardware-based routers and middleboxes. In this article we describe how a carefully designed NFV architecture can overcome these virtualization-layer overheads through zero-copy packet data transfer, non-uniform memory access (NUMA)-aware scheduling, and lockless data structures. Our approach exploits advances in multi-core CPUs and modern network interface cards (NICs) to enable a truly flexible, VM-based networking platform that can process packets at line rates.SDN has already impacted how networks are deployed and managed, but current approaches do not fully exploit the benefits of an NFV-based infrastructure. SDN controllers still assume they are interacting with simple hardware devices that are incapable of making decisions on their own. The reality is that increasing deployments of middleboxes and NFV-based services means that not only will flow management become more complex, but also that data plane elements will want to make dynamic decisions about how packets are directed. The ability to dynamically steer flows through selected service functions is a key ...

No abstract

The high replication cost of Byzantine fault-tolerance (BFT) methods has been a major barrier to their widespread adoption in commercial distributed applications. We present ZZ, a new approach that reduces the replication cost of BFT services from 2f + 1 to practically f + 1. The key insight in ZZ is to use f + 1 execution replicas in the normal case and to activate additional replicas only upon failures. In data centers where multiple applications share a physical server, ZZ reduces the aggregate number of execution replicas running in the data center, improving throughput and response times. ZZ relies on virtualization-a technology already employed in modern data centers-for fast replica activation upon failures, and enables newly activated replicas to immediately begin processing requests by fetching state on-demand. A prototype implementation of ZZ using the BASE library and Xen shows that, when compared to a system with 2f + 1 replicas, our approach yields lower response times and up to 33% higher throughput in a prototype data center with four BFT web applications. We also show that ZZ can handle simultaneous failures and achieve sub-second recovery.