Improving resource management in virtualized data centers using application performance models

The rapid growth of virtualized data centers and cloud hosting services is making the management of physical resources such as CPU, memory, and I/O bandwidth in data center servers increasingly important. Server management now involves dealing with multiple dissimilar applications with varying Service-Level-Agreements (SLAs) and multiple resource dimensions. The multiplicity and diversity of resources and applications are rendering administrative tasks more complex and challenging. This thesis aimed to develop a framework and techniques that would help substantially reduce data center management complexity.^ We specifically addressed two crucial data center operations. First, we precisely estimated capacity requirements of client virtual machines (VMs) while renting server space in cloud environment. Second, we proposed a systematic process to efficiently allocate physical resources to hosted VMs in a data center. To realize these dual objectives, accurately capturing the effects of resource allocations on application performance is vital. The benefits of accurate application performance modeling are multifold. Cloud users can size their VMs appropriately and pay only for the resources that they need; service providers can also offer a new charging model based on the VMs performance instead of their configured sizes. As a result, clients will pay exactly for the performance they are actually experiencing; on the other hand, administrators will be able to maximize their total revenue by utilizing application performance models and SLAs. ^ This thesis made the following contributions. First, we identified resource control parameters crucial for distributing physical resources and characterizing contention for virtualized applications in a shared hosting environment. Second, we explored several modeling techniques and confirmed the suitability of two machine learning tools, Artificial Neural Network and Support Vector Machine, to accurately model the performance of virtualized applications. Moreover, we suggested and evaluated modeling optimizations necessary to improve prediction accuracy when using these modeling tools. Third, we presented an approach to optimal VM sizing by employing the performance models we created. Finally, we proposed a revenue-driven resource allocation algorithm which maximizes the SLA-generated revenue for a data center.^

Improving Resource Management in Virtualized Data Centers using Application Performance Models

The rapid growth of virtualized data centers and cloud hosting services is making the management of physical resources such as CPU, memory, and I/O bandwidth in data center servers increasingly important. Server management now involves dealing with multiple dissimilar applications with varying Service-Level-Agreements (SLAs) and multiple resource dimensions. The multiplicity and diversity of resources and applications are rendering administrative tasks more complex and challenging. This thesis aimed to develop a framework and techniques that would help substantially reduce data center management complexity. We specifically addressed two crucial data center operations. First, we precisely estimated capacity requirements of client virtual machines (VMs) while renting server space in cloud environment. Second, we proposed a systematic process to efficiently allocate physical resources to hosted VMs in a data center. To realize these dual objectives, accurately capturing the effects of resource allocations on application performance is vital. The benefits of accurate application performance modeling are multifold. Cloud users can size their VMs appropriately and pay only for the resources that they need; service providers can also offer a new charging model based on the VMs performance instead of their configured sizes. As a result, clients will pay exactly for the performance they are actually experiencing; on the other hand, administrators will be able to maximize their total revenue by utilizing application performance models and SLAs. This thesis made the following contributions. First, we identified resource control parameters crucial for distributing physical resources and characterizing contention for virtualized applications in a shared hosting environment. Second, we explored several modeling techniques and confirmed the suitability of two machine learning tools, Artificial Neural Network and Support Vector Machine, to accurately model the performance of virtualized applications. Moreover, we suggested and evaluated modeling optimizations necessary to improve prediction accuracy when using these modeling tools. Third, we presented an approach to optimal VM sizing by employing the performance models we created. Finally, we proposed a revenue-driven resource allocation algorithm which maximizes the SLA-generated revenue for a data center.

Cuba’s Role in Venezuela’s Control of the Internet and Online Social Networks

The arrival of Cuba’s Information Technology (IT) and Communications Minister Ramiro Valdés to Venezuela in the Spring of 2010 to serve as a ‘consultant’ to the Venezuelan government awakened a new reality in that country. Rampant with deep economic troubles, escalating crime, a murder rate that has doubled since Chávez took over in 1999, and an opposition movement led by university students and other activists who use the Internet as their primary weapon, Venezuela has resorted to Cuba for help. In a country where in large part traditional media outlets have been censored or are government-controlled, the Internet and its online social networks have become the place to obtain, as well as disseminate, unfiltered information. As such, Internet growth and use of its social networks has skyrocketed in Venezuela, making it one of Latin America’s highest Web users. Because of its increased use to spark political debate among Venezuelans and publish information that differs with the official government line, Chávez has embarked on an initiative to bring the Internet to the poor and others who would otherwise not have access, by establishing government-sponsored Internet Info Centers throughout the country, to disseminate information to his followers. With the help of Cuban advisors, who for years have been a part of Venezuela’s defense, education, and health care initiatives, Chávez has apparently taken to adapting Cuba’s methodology for the control of information. He has begun to take special steps toward also controlling the type of information flowing through the country’s online social networks, considering the implementation of a government-controlled single Internet access point in Venezuela. Simultaneously, in adapting to Venezuela’s Internet reality, Chávez has engaged online by creating his own Twitter account in an attempt to influence public opinion, primarily of those who browse the Web. With a rapidly growing following that may soon reach one million subscribers, Chávez claims to have set up his own online trench to wage cyber space battle.

Host and network optimizations for performance enhancement and energy efficiency in data center networks

Modern data centers host hundreds of thousands of servers to achieve economies of scale. Such a huge number of servers create challenges for the data center network (DCN) to provide proportionally large bandwidth. In addition, the deployment of virtual machines (VMs) in data centers raises the requirements for efficient resource allocation and find-grained resource sharing. Further, the large number of servers and switches in the data center consume significant amounts of energy. Even though servers become more energy efficient with various energy saving techniques, DCN still accounts for 20% to 50% of the energy consumed by the entire data center. The objective of this dissertation is to enhance DCN performance as well as its energy efficiency by conducting optimizations on both host and network sides. First, as the DCN demands huge bisection bandwidth to interconnect all the servers, we propose a parallel packet switch (PPS) architecture that directly processes variable length packets without segmentation-and-reassembly (SAR). The proposed PPS achieves large bandwidth by combining switching capacities of multiple fabrics, and it further improves the switch throughput by avoiding padding bits in SAR. Second, since certain resource demands of the VM are bursty and demonstrate stochastic nature, to satisfy both deterministic and stochastic demands in VM placement, we propose the Max-Min Multidimensional Stochastic Bin Packing (M3SBP) algorithm. M3SBP calculates an equivalent deterministic value for the stochastic demands, and maximizes the minimum resource utilization ratio of each server. Third, to provide necessary traffic isolation for VMs that share the same physical network adapter, we propose the Flow-level Bandwidth Provisioning (FBP) algorithm. By reducing the flow scheduling problem to multiple stages of packet queuing problems, FBP guarantees the provisioned bandwidth and delay performance for each flow. Finally, while DCNs are typically provisioned with full bisection bandwidth, DCN traffic demonstrates fluctuating patterns, we propose a joint host-network optimization scheme to enhance the energy efficiency of DCNs during off-peak traffic hours. The proposed scheme utilizes a unified representation method that converts the VM placement problem to a routing problem and employs depth-first and best-fit search to find efficient paths for flows.

Interoperable resource brokering with policy-based provisioning and job allocation

The increasing needs for computational power in areas such as weather simulation, genomics or Internet applications have led to sharing of geographically distributed and heterogeneous resources from commercial data centers and scientific institutions. Research in the areas of utility, grid and cloud computing, together with improvements in network and hardware virtualization has resulted in methods to locate and use resources to rapidly provision virtual environments in a flexible manner, while lowering costs for consumers and providers. ^ However, there is still a lack of methodologies to enable efficient and seamless sharing of resources among institutions. In this work, we concentrate in the problem of executing parallel scientific applications across distributed resources belonging to separate organizations. Our approach can be divided in three main points. First, we define and implement an interoperable grid protocol to distribute job workloads among partners with different middleware and execution resources. Second, we research and implement different policies for virtual resource provisioning and job-to-resource allocation, taking advantage of their cooperation to improve execution cost and performance. Third, we explore the consequences of on-demand provisioning and allocation in the problem of site-selection for the execution of parallel workloads, and propose new strategies to reduce job slowdown and overall cost.^

Host and Network Optimizations for Performance Enhancement and Energy Efficiency in Data Center Networks

Modern data centers host hundreds of thousands of servers to achieve economies of scale. Such a huge number of servers create challenges for the data center network (DCN) to provide proportionally large bandwidth. In addition, the deployment of virtual machines (VMs) in data centers raises the requirements for efficient resource allocation and find-grained resource sharing. Further, the large number of servers and switches in the data center consume significant amounts of energy. Even though servers become more energy efficient with various energy saving techniques, DCN still accounts for 20% to 50% of the energy consumed by the entire data center. The objective of this dissertation is to enhance DCN performance as well as its energy efficiency by conducting optimizations on both host and network sides. First, as the DCN demands huge bisection bandwidth to interconnect all the servers, we propose a parallel packet switch (PPS) architecture that directly processes variable length packets without segmentation-and-reassembly (SAR). The proposed PPS achieves large bandwidth by combining switching capacities of multiple fabrics, and it further improves the switch throughput by avoiding padding bits in SAR. Second, since certain resource demands of the VM are bursty and demonstrate stochastic nature, to satisfy both deterministic and stochastic demands in VM placement, we propose the Max-Min Multidimensional Stochastic Bin Packing (M3SBP) algorithm. M3SBP calculates an equivalent deterministic value for the stochastic demands, and maximizes the minimum resource utilization ratio of each server. Third, to provide necessary traffic isolation for VMs that share the same physical network adapter, we propose the Flow-level Bandwidth Provisioning (FBP) algorithm. By reducing the flow scheduling problem to multiple stages of packet queuing problems, FBP guarantees the provisioned bandwidth and delay performance for each flow. Finally, while DCNs are typically provisioned with full bisection bandwidth, DCN traffic demonstrates fluctuating patterns, we propose a joint host-network optimization scheme to enhance the energy efficiency of DCNs during off-peak traffic hours. The proposed scheme utilizes a unified representation method that converts the VM placement problem to a routing problem and employs depth-first and best-fit search to find efficient paths for flows.

Interoperable Resource Brokering with Policy-based Provisioning and Job Allocation

The increasing needs for computational power in areas such as weather simulation, genomics or Internet applications have led to sharing of geographically distributed and heterogeneous resources from commercial data centers and scientific institutions. Research in the areas of utility, grid and cloud computing, together with improvements in network and hardware virtualization has resulted in methods to locate and use resources to rapidly provision virtual environments in a flexible manner, while lowering costs for consumers and providers. However, there is still a lack of methodologies to enable efficient and seamless sharing of resources among institutions. In this work, we concentrate in the problem of executing parallel scientific applications across distributed resources belonging to separate organizations. Our approach can be divided in three main points. First, we define and implement an interoperable grid protocol to distribute job workloads among partners with different middleware and execution resources. Second, we research and implement different policies for virtual resource provisioning and job-to-resource allocation, taking advantage of their cooperation to improve execution cost and performance. Third, we explore the consequences of on-demand provisioning and allocation in the problem of site-selection for the execution of parallel workloads, and propose new strategies to reduce job slowdown and overall cost.

Semantic wrapping for heterogeneous database management

Query processing is a commonly performed procedure and a vital and integral part of information processing. It is therefore important and necessary for information processing applications to continuously improve the accessibility of data sources as well as the ability to perform queries on those data sources. ^ It is well known that the relational database model and the Structured Query Language (SQL) are currently the most popular tools to implement and query databases. However, a certain level of expertise is needed to use SQL and to access relational databases. This study presents a semantic modeling approach that enables the average user to access and query existing relational databases without the concern of the database's structure or technicalities. This method includes an algorithm to represent relational database schemas in a more semantically rich way. The result of which is a semantic view of the relational database. The user performs queries using an adapted version of SQL, namely Semantic SQL. This method substantially reduces the size and complexity of queries. Additionally, it shortens the database application development cycle and improves maintenance and reliability by reducing the size of application programs. Furthermore, a Semantic Wrapper tool illustrating the semantic wrapping method is presented. ^ I further extend the use of this semantic wrapping method to heterogeneous database management. Relational, object-oriented databases and the Internet data sources are considered to be part of the heterogeneous database environment. Semantic schemas resulting from the algorithm presented in the method were employed to describe the structure of these data sources in a uniform way. Semantic SQL was utilized to query various data sources. As a result, this method provides users with the ability to access and perform queries on heterogeneous database systems in a more innate way. ^

Optimizing storage and memory systems for energy and performance

Electrical energy is an essential resource for the modern world. Unfortunately, its price has almost doubled in the last decade. Furthermore, energy production is also currently one of the primary sources of pollution. These concerns are becoming more important in data-centers. As more computational power is required to serve hundreds of millions of users, bigger data-centers are becoming necessary. This results in higher electrical energy consumption. Of all the energy used in data-centers, including power distribution units, lights, and cooling, computer hardware consumes as much as 80%. Consequently, there is opportunity to make data-centers more energy efficient by designing systems with lower energy footprint. Consuming less energy is critical not only in data-centers. It is also important in mobile devices where battery-based energy is a scarce resource. Reducing the energy consumption of these devices will allow them to last longer and re-charge less frequently. Saving energy in computer systems is a challenging problem. Improving a system's energy efficiency usually comes at the cost of compromises in other areas such as performance or reliability. In the case of secondary storage, for example, spinning-down the disks to save energy can incur high latencies if they are accessed while in this state. The challenge is to be able to increase the energy efficiency while keeping the system as reliable and responsive as before. This thesis tackles the problem of improving energy efficiency in existing systems while reducing the impact on performance. First, we propose a new technique to achieve fine grained energy proportionality in multi-disk systems; Second, we design and implement an energy-efficient cache system using flash memory that increases disk idleness to save energy; Finally, we identify and explore solutions for the page fetch-before-update problem in caching systems that can: (a) control better I/O traffic to secondary storage and (b) provide critical performance improvement for energy efficient systems.

Delay-sensitive service request scheduling for cloud computing

Cloud computing realizes the long-held dream of converting computing capability into a type of utility. It has the potential to fundamentally change the landscape of the IT industry and our way of life. However, as cloud computing expanding substantially in both scale and scope, ensuring its sustainable growth is a critical problem. Service providers have long been suffering from high operational costs. Especially the costs associated with the skyrocketing power consumption of large data centers. In the meantime, while efficient power/energy utilization is indispensable for the sustainable growth of cloud computing, service providers must also satisfy a user's quality of service (QoS) requirements. This problem becomes even more challenging considering the increasingly stringent power/energy and QoS constraints, as well as other factors such as the highly dynamic, heterogeneous, and distributed nature of the computing infrastructures, etc. ^ In this dissertation, we study the problem of delay-sensitive cloud service scheduling for the sustainable development of cloud computing. We first focus our research on the development of scheduling methods for delay-sensitive cloud services on a single server with the goal of maximizing a service provider's profit. We then extend our study to scheduling cloud services in distributed environments. In particular, we develop a queue-based model and derive efficient request dispatching and processing decisions in a multi-electricity-market environment to improve the profits for service providers. We next study a problem of multi-tier service scheduling. By carefully assigning sub deadlines to the service tiers, our approach can significantly improve resource usage efficiencies with statistically guaranteed QoS. Finally, we study the power conscious resource provision problem for service requests with different QoS requirements. By properly sharing computing resources among different requests, our method statistically guarantees all QoS requirements with a minimized number of powered-on servers and thus the power consumptions. The significance of our research is that it is one part of the integrated effort from both industry and academia to ensure the sustainable growth of cloud computing as it continues to evolve and change our society profoundly.^

Delay-Sensitive Service Request Scheduling for Cloud Computing

Cloud computing realizes the long-held dream of converting computing capability into a type of utility. It has the potential to fundamentally change the landscape of the IT industry and our way of life. However, as cloud computing expanding substantially in both scale and scope, ensuring its sustainable growth is a critical problem. Service providers have long been suffering from high operational costs. Especially the costs associated with the skyrocketing power consumption of large data centers. In the meantime, while efficient power/energy utilization is indispensable for the sustainable growth of cloud computing, service providers must also satisfy a user's quality of service (QoS) requirements. This problem becomes even more challenging considering the increasingly stringent power/energy and QoS constraints, as well as other factors such as the highly dynamic, heterogeneous, and distributed nature of the computing infrastructures, etc. In this dissertation, we study the problem of delay-sensitive cloud service scheduling for the sustainable development of cloud computing. We first focus our research on the development of scheduling methods for delay-sensitive cloud services on a single server with the goal of maximizing a service provider's profit. We then extend our study to scheduling cloud services in distributed environments. In particular, we develop a queue-based model and derive efficient request dispatching and processing decisions in a multi-electricity-market environment to improve the profits for service providers. We next study a problem of multi-tier service scheduling. By carefully assigning sub deadlines to the service tiers, our approach can significantly improve resource usage efficiencies with statistically guaranteed QoS. Finally, we study the power conscious resource provision problem for service requests with different QoS requirements. By properly sharing computing resources among different requests, our method statistically guarantees all QoS requirements with a minimized number of powered-on servers and thus the power consumptions. The significance of our research is that it is one part of the integrated effort from both industry and academia to ensure the sustainable growth of cloud computing as it continues to evolve and change our society profoundly.

Semantic distributed database management with applications to the Internet dissemination of environmental data

The research presented in this dissertation is comprised of several parts which jointly attain the goal of Semantic Distributed Database Management with Applications to Internet Dissemination of Environmental Data. ^ Part of the research into more effective and efficient data management has been pursued through enhancements to the Semantic Binary Object-Oriented database (Sem-ODB) such as more effective load balancing techniques for the database engine, and the use of Sem-ODB as a tool for integrating structured and unstructured heterogeneous data sources. Another part of the research in data management has pursued methods for optimizing queries in distributed databases through the intelligent use of network bandwidth; this has applications in networks that provide varying levels of Quality of Service or throughput. ^ The application of the Semantic Binary database model as a tool for relational database modeling has also been pursued. This has resulted in database applications that are used by researchers at the Everglades National Park to store environmental data and to remotely-sensed imagery. ^ The areas of research described above have contributed to the creation TerraFly, which provides for the dissemination of geospatial data via the Internet. TerraFly research presented herein ranges from the development of TerraFly's back-end database and interfaces, through the features that are presented to the public (such as the ability to provide autopilot scripts and on-demand data about a point), to applications of TerraFly in the areas of hazard mitigation, recreation, and aviation. ^