Network Function Virtualization (NFV) platforms consume significant energy, introducing high operational costs in edge and data centers. This paper presents a novel framework called GreenNFV that optimizes resource usage for network function chains using deep reinforcement learning. GreenNFV optimizes resource parameters such as CPU sharing ratio, CPU frequency scaling, last-level cache (LLC) allocation, DMA buffer size, and packet batch size. GreenNFV learns the resource scheduling model from the benchmark experiments and takes Service Level Agreements (SLAs) into account to optimize resource usage models based on the different throughput and energy consumption requirements. Our evaluation shows that GreenNFV models achieve high transfer throughput and low energy consumption while satisfying various SLA constraints. Specifically, GreenNFV with Throughput SLA can achieve 4.4× higher throughput and 1.5× better energy efficiency over the baseline settings, whereas GreenNFV with Energy SLA can achieve 3× higher throughput while reducing energy consumption by 50%.

In modern data centers, enterprise cloud instances run not only foreground applications like web and databases, but also different background services (e.g., backup, virus/compliance scan, batch) to manage the cloud instances securely and improve the overall resource utilization. These background services often incur resource storms that suddenly consume a lot of shared resources on cloud instances. The resource storms significantly degrade the performance of foreground applications by interfering in the preemption of the shared resources, resulting in frequent SLA violations. However, stock OS schedulers are not designed to handle these situations, and prior works are insufficient to address such resource storms under highly dynamic cloud workloads. This article presents Orchestra, a cloud-specific framework for controlling multiple applications in the user space, aiming at meeting corresponding SLAs. Orchestra takes an online approach with lightweight monitoring and performance models for both applications on the fly. It optimizes the resource allocations to meet corresponding SLAs. We evaluate the performance of Orchestra on a production cloud with a diverse range of SLAs. Orchestra guarantees the foreground application's performance SLAs at all times. At the same time, Orchestra maintains the background's performance by minimizing its performance penalty with proper allocation of the shared resources.

Cloud Security Posture Management (CSPM) tools have been gaining popularity to automate, monitor and visualize the security posture of multi-cloud environments. The foundation to assess the risk lies on being able to analyze each vulnerability and quantify its risk. However, the number of vulnerabilities in National Vulnerability Database (NVD) has skyrocketed in recent years and surpassed 144K as of late 2020. The current standard vulnerability tracking system relies mostly on human-driven efforts. Besides, open-source libraries do not necessarily follow the standards of vulnerability reporting set by CVE and NIST, but rather use Github issues for reporting. In this paper, we propose a framework, NL2Vul, to measure score of vulnerabilities with minimal human efforts. NL2Vul makes use of deep neural networks to train on descriptions of software vulnerabilities from NVD and predicts vulnerability scores. To flexibly expand the trained NVD model for different data sources that are being used to evaluate the risk posture in CSPM, NL2Vul uses transfer learning for quick re-training. We have evaluated NL2Vul with vanilla NVD, public Github issues of open source projects, and compliance technology specification documents.

Serverless computing platforms simplify development, deployment, and automated management of modular software functions. However, existing serverless platforms typically assume an over-provisioned cloud, making them a poor fit for Edge Computing environments where resources are scarce. In this paper we propose a redesigned serverless platform that comprehensively tackles the key challenges for serverless functions in a resource constrained Edge Cloud.

Cloud applications based on the "Functions as a Service" (FaaS) paradigm have become very popular. Yet, due to their stateless nature, they must frequently interact with an external data store, which limits their performance. To mitigate this issue, we introduce OFC, a transparent, vertically and horizontally elastic in-memory caching system for FaaS platforms, distributed over the worker nodes. OFC provides these benefits cost-effectively by exploiting two common sources of resource waste: (i) most cloud tenants overprovision the memory resources reserved for their functions because their footprint is non-trivially input-dependent and (ii) FaaS providers keep function sandboxes alive for several minutes to avoid cold starts. Using machine learning models adjusted for typical function input data categories (e.g., multimedia formats), OFC estimates the actual memory resources required by each function invocation and hoards the remaining capacity to feed the cache. We build our OFC prototype based on enhancements to the OpenWhisk FaaS platform, the Swift persistent object store, and the RAM-Cloud in-memory store. Using a diverse set of workloads, we show that OFC improves by up to 82 % and 60 % respectively the execution time of single-stage and pipelined functions.

The Softwarization of networks is enabled by the SDN (Software-Defined Networking), NV (Network Virtualization), and NFV (Network Function Virtualization) paradigms, and offers many advantages for network operators, service providers and data-center providers. Given the strong interest in both industry and academia in the softwarization of telecommunication networks and cloud computing infrastructures, a series of special issues was established in IEEE Transactions on Network and Service Management, which aims at the timely publication of recent innovative research results on the management of softwarized networks.

Artificial Intelligence (AI) has the potential to leverage the immense amount of operational data of clouds, services, and social and communication networks. As a concrete example, AI techniques have been adopted by telcom operators to develop virtual assistants based on advances in natural language processing (NLP) for interaction with customers and machine learning (ML) to enhance the customer experience by improving customer flow. Machine learning has also been applied to finding fraud patterns which enables operators to focus on dealing with the activity as opposed to the previous focus on detecting fraud.

Managing Network Function (NF) service chains requires careful system resource management. We propose NFVnice, a user space NF scheduling and service chain management framework to provide fair, efficient and dynamic resource scheduling capabilities on Network Function Virtualization (NFV) platforms. The NFVnice framework monitors load on a service chain at high frequency (1000Hz) and employs backpressure to shed load early in the service chain, thereby preventing wasted work. Borrowing concepts such as rate proportional scheduling from hardware packet schedulers, CPU shares are computed by accounting for heterogeneous packet processing costs of NFs, I/O, and traffic arrival characteristics. By leveraging cgroups, a user space process scheduling abstraction exposed by the operating system, NFVnice is capable of controlling when network functions should be scheduled. NFVnice improves NF performance by complementing the capabilities of the OS scheduler but without requiring changes to the OS's scheduling mechanisms. Our controlled experiments show that NFVnice provides the appropriate rate-cost proportional fair share of CPU to NFs and significantly improves NF performance (throughput and latency) by reducing wasted work across an NF chain, compared to using the default OS scheduler. NFVnice achieves this even for heterogeneous NFs with vastly different computational costs and for heterogeneous workloads.

The growing popularity of cloud computing has increased demands of application migration to the cloud, but among others, the main deterrent has been the complexity of the migration planning for the large-scale projects with 1000s of servers. The complexity is mainly derived from the plethora of different migration options and required platform/application changes for cloud fitness of applications. In addition, the automated or artificial intelligence planning has been used to generate the migration plans, but despite its effectiveness, it has not been adopted broadly in migration because the AI planning language is complex and hard to scale. In this paper, we propose Cloud Readiness Planning Tool (CRPT), a system that constitutes a Migration Type Classifier trained under a novel active learning strategy dealing with "concept drift", and an AI Planner which generates plan from automatically created domain and problem files with declarative specifications such as goal states and data in user friendly input formats. A series of experiments were conducted on a real-world migration task. The results demonstrate the Migration Type Classifier is able to effectively adapt to the changing business needs and achieve high accuracy with low labeling cost.

DevOps (software DEVelopment and information technology OPerationS) has established a culture and environment, in which building, testing, and releasing software happen more rapidly, frequently, and reliably through automated pipelines. To support this high turnover application cycle, the cluster orchestration frameworks such as Kubernetes or Docker Swarm have evolved to provide high flexibility and reliability. However, while cluster orchestrators run on the cloud infrastructure, any failure incurred from the infrastructure can directly impact the nodes of a cluster, so infrastructure failures can disrupt applications running on the cluster. In this paper, we propose proactive application placement algorithms with prediction of infrastructure failures. The proposed algorithms utilize failure-risk measurements, Failure-Index, determined from turnover rate of applications and prediction of infrastructure failures. We build stochastic models for application turnover and infrastructure failure processes, and provide various types of Failure-Index. Our placement algorithms are implemented in an orchestration framework, Kubernetes, and in the simulation model. Experimental results show that our methods reduce the amount of application disruption by 20% than the state-of-the-art algorithms in Kubernetes.

With the emergence of data deluge, the energy footprint of global data movement has surpassed 100 terawatt hours, costing more than 20 billion US dollars to the world economy. During an active data transfer, depending on the number of hops between the source and destination, the networking infrastructure consumes between 10% - 75% of the total energy, and the rest is consumed by the end systems. Even though there has been extensive research on reducing the power consumption at the networking infrastructure, the work focusing on saving energy at the end systems has been limited to the tuning of a few application-level parameters. In this paper, we introduce a novel cross-layer optimization framework which jointly considers application-level and kernel-level parameters to minimize the energy consumption without sacrificing from the transfer throughput. We present three different algorithms which can dynamically tune the CPU frequency level, number of active CPU cores, number of active transfer threads, number of parallel TCP streams, and the level of transfer command pipelining to achieve different user-set goals. Experimental results show that our proposed algorithms outperform the state-of-the-art solutions, achieving up to 80% higher throughput while consuming 48% less energy.

Network Function Virtualization seeks to run high performance middleboxes in a flexible, more configurable software environment. Even with advances such as kernel bypass and zero-copy IO, middlebox platforms still struggle to meet stringent throughput and latency requirements. To achieve line rates as network bandwidths rise, these platforms often must make tradeoffs such as inefficiently dedicating more CPU cores or weakening security and isolation properties. In this paper we explore how advances in programmable “smart NICs” can be leveraged by software middlebox platforms to improve performance, resource efficiency, and security. Our evaluation shows several use cases for smart NICs, which improve performance significantly while reducing resource consumption and providing strong isolation.

The global data movement over Internet has an estimated energy footprint of 100 terawatt hours per year, costing the world economy billions of dollars. The networking infrastructure together with source and destination nodes involved in the data transfer contribute to overall energy consumption. Although considerable amount of research has rendered power management techniques for the networking infrastructure, there has not been much prior work focusing on energy-aware data transfer solutions for minimizing the power consumed at the end-systems. In this paper, we introduce a novel application-layer solution based on historical analysis and real-time tuning called GreenDataFlow, which aims to achieve high data transfer throughput while keeping the energy consumption at the minimal levels. GreenDataFlow supports service level agreements (SLAs) which give the service providers and the consumers the ability to fine tune their goals and priorities in this optimization process. Our experimental results show that GreenDataFlow outperforms the closest competing state-of-the art solution in this area 50% for energy saving and 2.5× for the achieved end-to-end performance.

This paper presents Orchestra, a cloud-specific framework for managing both foreground applications (e.g., Web, DBMS) and background services (e.g., backup, security check, batch jobs) in the user space. Orchestra is designed to address "resource storms" caused by sudden executions of the background services on the cloud instances. The resource storms significantly degrade the performance of foreground applications by interfering in the preemption of the shared resources, resulting in frequent SLA violations and poor user experience. Orchestra takes an online approach using lightweight monitoring and creates performance models for multiple cloud applications on the fly. It then optimizes the allocations of shared resources to meet SLAs. We evaluate the performance of Orchestra on a production cloud (Amazon EC2) with a diverse range of SLA requirements. The experiment results show that Orchestra successfully guarantees the foreground application's performance to meet its SLA targets at all times. Moreover, Orchestra maintains the background's performance by minimizing its performance penalty with proper allocation of the shared resources.

According to a recent study, 30% of VMs in private cloud data centers are "comatose", in part because there is generally no strong incentive for their human owners to delete them at an appropriate time. These inactive VMs are still scheduled and executed on physical cloud resources, taking valuable access away from productive VMs. In an extreme, cloud infrastructure may deny legitimate requests for new VMs because capacity limits have been hit. It is not sufficient for cloud infrastructure to identify such inactive VMs by monitoring resource utilization (e.g., CPU utilization) - e.g., management processes (e.g. virus-scan, software update) on inactive VMs often consume high CPU and memory resources, and active VMs with lightweight jobs (e.g. text editing) show almost zero resource utilization. To properly detect and address such inactive VMs, we present iCSI: a cloud garbage VM collector to improve resource utilization and cost efficiency of enterprise data centers. iCSI includes three main components, a lightweight data collector, a VM identification model and a recommendation engine. The data collector periodically gathers primitive information from VMs. The identification model infers the purpose of a VM from the data collection and extracts the most relevant features associated with the purpose. The recommendation engine offers proper actions to end users i.e., suspending or resizing VMs. In this prototype phase, iCSI is deployed into multiple data centers in IBM and manages more than 750 production VMs. iCSI achieves 20% better accuracy (90%) in identifying active/inactive VMs compared with state-of-the-art methods. With recommendations to end users, our estimation results show that iCSI can improve internal cost efficiency with 23% and resource utilization more than 45%.

Presented BlueWall: Software Defined Network Management in Hybrid Enterprise Cloud Environments for managing firewall request when servers (or services) are created (via APIs). Demonstrated hybrid network management design and self-service capabilities. Discussed challenges arising in network management in the hybrid enterprise cloud environments.

Presented BlueShift: Self-service for orchestrating automated tasks (via APIs) and human tasks for end-to-end transformation process including, application discovery, analysis, artifact transformation and enablement of cloud value-add services. Demonstrated transformation process for PlantsByWebSphere application, with Liberty Profile runtime as a target in BlueMix (cloud-foundry based platform). Discussed challenges arising in transformation from application complexity and non-functional requirements.

Recent advance in geo-distributed systems has made distributed data processing possible, where tasks are decomposed into subtasks, deployed into multiple data centers and run in parallel. Compared to conventional approaches that process every task in a single datacenter resulting in high latency and large data aggregation, the geo-distributed cloud systems provide a highly available and more economic platform. However, distributed application (task) execution introduces extra cost and latency as data need to be exchanged between data centers. In addition, task dependency and diverse task constraints make it even more challenging to choose an appropriate task assignment strategy. In this paper, we discuss a task assignment problem in geographically distributed cloud systems. In light of growing demand from big data processing and storage, we consider data intensive tasks where a task often requires significant computing resources and its input data typically located in multiple data centers. By taking the distributed input, task dependency, heterogeneous pricing scheme, and resource constraints into account, we aim to optimize the performance when deploying tasks in geo-graphically distributed data centers. A heuristic algorithm is presented to provide an approximate solution to the proposed NP-hard problem. We perform an extensive simulation study to evaluate the performance of our solution under various settings. The simulation results demonstrate that our approach can outperform the state-of-the-art strategies, and achieve significant reduction in cost and latency.

Managing Network Function (NF) service chains requires careful system resource management. We propose NFVnice, a user space NF scheduling and service chain management framework to provide fair, efficient and dynamic resource scheduling capabilities on Network Function Virtualization (NFV) platforms. The NFVnice framework monitors load on a service chain at high frequency (1000Hz) and employs backpressure to shed load early in the service chain, thereby preventing wasted work. Borrowing concepts such as rate proportional scheduling from hardware packet schedulers, CPU shares are computed by accounting for heterogeneous packet processing costs of NFs, I/O, and traffic arrival characteristics. By leveraging cgroups, a user space process scheduling abstraction exposed by the operating system, NFVnice is capable of controlling when network functions should be scheduled. NFVnice improves NF performance by complementing the capabilities of the OS scheduler but without requiring changes to the OS's scheduling mechanisms. Our controlled experiments show that NFVnice provides the appropriate rate-cost proportional fair share of CPU to NFs and significantly improves NF performance (throughput and loss) by reducing wasted work across an NF chain, compared to using the default OS scheduler. NFVnice achieves this even for heterogeneous NFs with vastly different computational costs and for heterogeneous workloads.

The combination of Network Function Virtualization (NFV) and Software Defined Networking (SDN) allows flows to be flexibly steered through efficient processing pipelines. As deployment of NFV becomes more prevalent, the need to provide fine-grained customization of service chains and flow-level performance guarantees will increase, even as the diversity of Network Functions (NFs) rises. Existing NFV approaches typically route wide classes of traffic through pre-configured service chains. While this aggregation improves efficiency, it prevents flexibly steering and managing performance of flows at a fine granularity.

Network function virtualization (NFV) and Software Defined Networks (SDN) separate and abstract network functions from underlying hardware, creating a flexible virtual networking environment that reduces cost and allows policy-based decisions. One of the biggest challenges in NFV-SDN is to map the required virtual network functions (VNFs) to the underlying hardware in substrate networks in a timely manner. In this paper, we formulate the VNF placement problem via Graph Pattern Matching, with an objective function that can be easily adapted to fit various applications. Previous work only considers off-line VNF placement as it is time consuming to find an appropriate mapping path while considering all software and hardware constraints. To reduce this time, we investigate the feasibility and effectiveness of path-precomputing, where paths are calculated prior to placement. Our approach enables online VNF placement in SDNs, allowing VNF requests to be processed as they arrive. An online placement approach (OPA) is proposed to place VNF requests on substrate networks. To the best of our knowledge, this is the first work in the literature that considers the online chaining VNF placement in SDNs. In addition, we present an application of OPA over cost minimization. Simulation results demonstrate that our online approach provides competitive performance compared with off-line algorithms.

Cloud migration transforms company's data, applications and services to (or between) one or more other Cloud environments. Enterprises are increasingly migrating their IT infrastructures to Cloud, given the appeal of (pay-per-use) elastic resources. Yet, existing IT infrastructures are complex, heterogeneous and dynamic ecosystems. As a result, there is no single standardized process to seamlessly manage migration at enterprise scale, and often significant level of manual intervention is required, both in reasoning about migration and during its execution. This paper presents a system that automates the process of migration to Cloud. It embeds a Metric-FF Artificial Intelligence (AI) planning algorithm to dynamically assemble migration plans based on the properties of source and target environments, as well as available migration tooling. The paper describes the challenges in migration planning, AI domain design for migration. This work demonstrates that the system provides an effective and scalable solution to generating plans based on the source environment of 700 servers, and varying size of the migration service requests.

With the promise of providing flexible and elastic computing resources on demand, the cloud computing has been attracting enterprises and individuals to migrate workloads in the legacy environment to the public/private/hybrid clouds. However, the workload migration is often interpreted as an image migration or re-installation/data copying as the exact snapshot of the source machine, and the various cloud platforms and service models are rarely taken into consideration during migration planning. Thus, the cloud migration techniques have not provided enough options that can satisfy the various migration requirements. In this paper, we propose a model to tackle the migration challenges that transforms one resource into the same or another resource in hybrid clouds. We formulate the problem as a constraint satisfaction problem, and iteratively decompose the server components and consolidate the servers. Furthermore, we propose a compute-network mapping algorithm to match computing resources with network resources to guarantee network affinity. The ultimate goal is to recommend the optimal target cloud platform with network affinity and the minimum cost. Through the evaluation of the proposed model using real enterprise datasets (up to 2012 machines), we prove that the proposed model satisfies the goal. We show that when migrating into virtualized cloud environments, thorough resource planning can reduce 16% of current resources, 5%-10% servers can be consolidated, and more than 60% servers are possible candidates for server decomposition.

With the promise of providing flexible and elastic computing resources on demand, the cloud computing has been attracting enterprises and individuals to migrate workloads in the legacy environment to the public/private/hybrid clouds. Also, cloud customers want to migrate between cloud providers with different requirements such as cost, performance, and manageability. However the workload migration is often interpreted as an image migration or re-installation/data copying as the exact snapshot of the source machine. Also the various cloud platforms and service models are rarely taken into consideration during the migration analytics. Therefore, although the expectation has risen with various requirements on the target cloud platforms and environments, the cloud migration techniques have not provided enough options that can satisfy the various requirements. In this paper we propose a model to tackle the migration challenges that transform one resource into same or another resource in hybrid clouds. We formulate the problem as a constraint satisfaction problem, and iteratively decompose the server components and consolidate the servers. The ultimate goal is to recommend the optimal target cloud platform and environment with the minimum cost. Through the evaluation of the proposed model using the real enterprise dataset (up to 2012 machines), we prove that the proposed model satisfies the goal. We show that when migrating into virtualized cloud environments, the thorough resource planning can reduce 16% of current resources, about 5%-10% servers can be consolidated, and more than 60% servers are possible candidates for server decomposition.

In the last few years, changing infrastructure and business requirements are forcing enterprises to rethink their networks. Enterprises look for network infrastructures that increase network efficiency, flexibility, and cost reduction. At the same time, the emergence of Cloud and mobile in enterprise networks has introduced tremendous variability in enterprise traffic patterns at the edge. This highly mobile and dynamic traffic presents a need for dynamic capacity management and adaptive traffic steering and appeals for new infrastructures and management solutions. In this context, passive optical networks (PON) have gained attention in the last few years as a promising solution for enterprise networks, as it can offer efficiency, security, and cost reduction. However, network management in PON is not yet automated and needs humain intervention. As such, capabilities for dynamic and adaptive PON are necessary. In this paper, we present a joint solution for PON capacity management both in deployment and in operation, as to maximize peak load tolerance by dynamically allocating capacity to fit varying and migratory traffic loads. To this end, we developed the novel approaches of capacity pool based deployment and dynamic traffic steering in PON. Compared with traditional edge network design, our approach significantly reduces the need for capacity over-provisioning. Compared with generic PON networks, our approach enables dynamic traffic steering through software-defined control. We implemented our design on a production grade PON testbed, and the results demonstrate the feasibility and flexibility of our approach.

With the promise of low-cost access to flexible and elastic resources, enterprises are increasingly migrating their existing workloads into the Cloud. Yet, the heterogeneity of the workloads and existing configuration of legacy IT infrastructure make it challenging to enable a one-click, seamless migration process. There are multiple tools available for migrating servers based on their existing configurations and multiple ways of dealing with data synchronization (post migration). In this paper, we present a Cloud Migration Orchestrator (CMO), based on business process management (BPM) approach to provide a systematic framework to automate and coordinate migration activities. CMO coordinates the process of migration, starting from discovery, provisioning, network configuration, execution of migration, cutover and validation. CMO integrates multiple migration technologies, to support different migration scenarios. We present and discuss our results from a preliminary deployment of CMO to migrate 25 VMware instances and discuss how this approach improves the effectiveness of migration, and seamlessly coordinates activities required to be executed.

When planning a data center migration it is critical to discover the client's business applications and on which devices (server, storage and appliances) those applications are deployed in the infrastructure. It is also important to understand the dependencies the applications have on the infrastructure, on other applications, and in some cases on systems external to the client. Clients can only rarely provide that information in a complete and accurate manner. The usual approach then has been to obtain the information by asking the client's application and platform owners a series of questions but in most cases clients do not have the tools or skills to acquire the requested information. The lack of accurate information leads to project delays, increased cost and higher levels of risk. In this paper we present an algorithm and tools for programmatically identifying and locating business application instances in an infrastructure, based on weighted similarity metric. We discuss results from our preliminary evaluation and the correctness of the algorithm. Such automated approach to application discovery significantly helps clients to achieve their project objectives and timeline without imposing additional work on the application and platform owners.

NetVM brings virtualization to the Network by enabling high bandwidth network functions to operate at near line speed, while taking advantage of the flexibility and customization of low cost commodity servers. NetVM allows customizable data plane processing capabilities such as firewalls, proxies, and routers to be embedded within virtual machines, complementing the control plane capabilities of Software Defined Networking. NetVM makes it easy to dynamically scale, deploy, and reprogram network functions. This provides far greater flexibility than existing purpose-built, sometimes proprietary hardware, while still allowing complex policies and full packet inspection to determine subsequent processing. It does so with dramatically higher throughput than existing software router platforms. NetVM is built on top of the KVM platform and Intel DPDK library. We detail many of the challenges we have solved such as adding support for high-speed inter-VM communication through shared huge pages and enhancing the CPU scheduler to prevent overheads caused by inter-core communication and context switching. NetVM allows true zero-copy delivery of data to VMs both for packet processing and messaging among VMs within a trust boundary. Our evaluation shows how NetVM can compose complex network functionality from multiple pipelined VMs and still obtain throughputs up to 10 Gbps, an improvement of more than 250% compared to existing techniques that use SR-IOV for virtualized networking.

Virtualization technology and the ease with which virtual machines (VMs) can be migrated within the LAN have 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 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 architecture consisting of cloud computing platforms linked with a virtual private network (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 19 GB, a 50% reduction.

Application and OS-level caches are crucial for hiding I/O latency and improving application performance. However, caches are designed to greedily consume memory, which can cause memory-hogging problems in a virtualized data centers since the hypervisor cannot tell for what a virtual machine uses its memory. A group of virtual machines may contain a wide range of caches: database query pools, memcached key-value stores, disk caches, etc., each of which would like as much memory as possible. The relative importance of these caches can vary significantly, yet system administrators currently have no easy way to dynamically manage the resources assigned to a range of virtual machine data caches in a unified way. To improve this situation, we have developed UniCache, a system that provides a hypervisor managed volatile data store that can cache data either in hypervisor controlled main memory (hot data) or on Flash based storage (cold data). We propose a two-level cache management system that uses a combination of recency information, object size, and a prediction of the cost to recover an object to guide its eviction algorithm. We have built a prototype of UniCache using Xen, and have evaluated its effectiveness in a shared environment where multiple virtual machines compete for storage resources.

Cloud data centers are difficult to manage because providers have no knowledge of what applications are being run by customers or how they interact. As a consequence, current clouds provide minimal automated management functionality, passing the problem on to users who have access to even fewer tools since they lack insight into the underlying infrastructure. Ideally, the cloud platform, not the customer, should be managing data center resources in order to both use them efficiently and provide strong application-level performance and reliability guarantees. To do this, we believe that clouds must become "distibuted-aware" so that they can deduce the overall structure and dependencies within a client's distributed applications and use that knowledge to better guide management services. Towards this end we are developing a light-weight topology detection system that maps distributed applications and a service classification algorithm that can determine not only overall application types, but individual VM roles as well.

Data center servers are typically overprovisioned, leaving spare memory and CPU capacity idle to handle unpredictable workload bursts by the virtual machines running on them. While this allows for fast hotspot mitigation, it is also wasteful. Unfortunately, making use of spare capacity without impacting active applications is particularly difficult for memory since it typically must be allocated in coarse chunks over long timescales. In this work we propose re- purposing the poorly utilized memory in a data center to store a volatile data store that is managed by the hypervisor. We present two uses for our Mortar framework: as a cache for prefetching disk blocks, and as an application-level distributed cache that follows the memcached protocol. Both prototypes use the framework to ask the hypervisor to store useful, but recoverable data within its free memory pool. This allows the hypervisor to control eviction policies and prioritize access to the cache. We demonstrate the benefits of our prototypes using realistic web applications and disk benchmarks, as well as memory traces gathered from live servers in our university's IT department. By expanding and contracting the data store size based on the free memory available, Mortar improves average response time of a web application by up to 35% compared to a fixed size memcached deployment, and improves overall video streaming performance by 45% through prefetching.

Data center servers are typically overprovisioned, leaving spare memory and CPU capacity idle to handle unpredictable workload bursts by the virtual machines running on them [1, 2, 3]. While this allows for fast hotspot mitigation, it is also wasteful. Unfortunately, making use of spare capacity without impacting active applications is particularly difficult for memory since it typically must be allocated in coarse chunks over long timescales [4, 5, 6, 7]. In this work we propose repurposing the poorly utilized memory in a data center to store a volatile data store that is managed by the hypervisor. We present two uses for our Mortar framework: as a cache for prefetching disk blocks [8, 9, 10], and as an application-level distributed cache that follows the memcached protocol [11, 12]. Both prototypes use the framework to ask the hypervisor to store useful, but recoverable data within its free memory pool. This allows the hypervisor to control eviction policies and prioritize access to the cache.

Virtual Desktop Infrastructures (VDIs) are gaining popularity in cloud computing by allowing companies to deploy their office environments in a virtualized setting instead of relying on physical desktop machines. Consolidating many users into a VDI environment can significantly lower IT management expenses and enables new features such as “available-anywhere” desktops. However, barriers to broad adoption include the slow performance of virtualized I/O, CPU scheduling interference problems, and shared-cache contention. In this paper, we propose a new soft real-time scheduling algorithm that employs flexible priority designations (via utility functions) and automated scheduler class detection (via hypervisor monitoring of user behavior) to provide a higher quality user experience. We have implemented our scheduler within the Xen virtualization platform, and demonstrate that the overheads incurred from co-locating large numbers of virtual machines can be reduced from 66% with existing schedulers to under 2% in our system. We evaluate the benefits and overheads of using a smaller scheduling time quantum in a VDI setting, and show that the average overhead time per scheduler call is on the same order as the existing SEDF and Credit schedulers.

This paper considers the problem of maximizing the system throughput in IEEE 802.16 broadband access networks with mesh topology, and the following results are presented. We consider a linear chain network and discuss its applicability for providing cost effective solutions in sparsely populated areas, such as interstate highways and rural communities. We provide an optimal scheduling algorithm and establish an analytical result on the length of the schedule for linear chain networks. We also consider the problem of routing and packet scheduling in general topology, and show its NP-completeness. Based on our optimal algorithm for linear networks, we propose algorithms that find routes and schedules of packet transmissions in general mesh topologies. The performance of our proposed algorithms is analyzed using the NS-2 simulator. The results show that the proposed algorithms perform significantly better than other existing algorithms.

To support high data rates, CDMA lx Ev- DO utilizes TDMA technology on the downlink allowing a single receiver per time slot. Accordingly, a scheduling algorithm is necessary to determine which user receives data in a given time slot. Typically, schedulers consider efficiency (increasing network throughput) or fairness (fair allocation of resources among users) as the basis for scheduling. To realize a reasonable tradeoff between efficiency and fairness, opportunistic schedulers have been proposed that take advantage of instantaneous improvements in users radio conditions by serving them at higher rates. However, given the increasing demand to support user QoS requirements, a QoS-aware scheduler may need to serve users even at times of inadequate radio conditions. In this paper, we propose a QoS-aware packet scheduling algorithm that takes into account policy rules that govern the relationships between different user QoS classes. Our scheduler uses marginal utility functions defined to embody the given rules when selecting users to be served on the downlink. Our scheduling algorithm is implemented in the OPNET module of the CDMA lx Ev-DO system that we have developed, and its performance is compared to other scheduling algorithms. The simulation results show that our scheduler performs well ensuring the policy rules are followed, meeting users' QoS requirements, and providing fairness among users within the same QoS class.