2017
Gustavo Hermínio de Araújo, Marcelo Antonio Marotta, Juliano Araujo Wickboldt, Cristiano Bonato Both, Luciano Paschoal Gaspary, Juergen Rochol, Lisandro Zambenedetti Granville
Caracterizando Estratégias de Domínio Espacial para Gerenciamento de Regras em Redes Definidas por Software Inproceedings
In: 35º Simpósio Brasileiro de Redes de Computadores e Sistemas Distribuídos, SBRC 2017, Belém, Brazil, May 15-19, 2017, pp. 788–801, SBC, 2017, ISSN: 2177-496X, (In portuguese).
Abstract Links BibTeX Tags: Resource Management Software-Defined Networking (SDN)
@inproceedings{conf/sbrc/deAraujo17,
title = {Caracterizando Estratégias de Domínio Espacial para Gerenciamento de Regras em Redes Definidas por Software},
author = {Gustavo Hermínio de Araújo and Marcelo Antonio Marotta and Juliano Araujo Wickboldt and Cristiano Bonato Both and Luciano Paschoal Gaspary and Juergen Rochol and Lisandro Zambenedetti Granville},
url = {https://sol.sbc.org.br/index.php/sbrc/article/view/2673},
issn = {2177-496X},
year = {2017},
date = {2017-01-01},
booktitle = {35º Simpósio Brasileiro de Redes de Computadores e Sistemas Distribuídos, SBRC 2017, Belém, Brazil, May 15-19, 2017},
pages = {788--801},
publisher = {SBC},
abstract = {Software-Defined Networks (SDN) based on Openflow protocol perform data forwarding through flow tables using match/action mechanisms. These tables have limited rules storage capacity, restricting network scalability and performance. Considering these restrictions, the main strategies for managing spatial domain rules, i.e., flow agregation and multiple flow tables, are promising to use the available storage space of the tables. These strategies impact packet processing in different ways, influencing the performance of the forwarding devices and the network. In spite of the impact, the comparison between flow aggregation and multiple flow tables is poorly exploited, leaving open the definition of which strategy is more appropriate for a network considering its topology and workload. In this article a quantitative characterization is proposed, defining the gains brought by these spatial domain strategies for forwarding devices and in different topologies.},
note = {In portuguese},
keywords = {Resource Management, Software-Defined Networking (SDN)},
pubstate = {published},
tppubtype = {inproceedings}
}
Software-Defined Networks (SDN) based on Openflow protocol perform data forwarding through flow tables using match/action mechanisms. These tables have limited rules storage capacity, restricting network scalability and performance. Considering these restrictions, the main strategies for managing spatial domain rules, i.e., flow agregation and multiple flow tables, are promising to use the available storage space of the tables. These strategies impact packet processing in different ways, influencing the performance of the forwarding devices and the network. In spite of the impact, the comparison between flow aggregation and multiple flow tables is poorly exploited, leaving open the definition of which strategy is more appropriate for a network considering its topology and workload. In this article a quantitative characterization is proposed, defining the gains brought by these spatial domain strategies for forwarding devices and in different topologies.
2016
Runxin Wang, Juliano Araujo Wickboldt, Rafael Pereira Esteves, Lei Shi, Brendan Jennings, Lisandro Zambenedetti Granville
Using Empirical Estimates of Effective Bandwidth in Network-Aware Placement of Virtual Machines in Datacenters Journal Article
In: IEEE Transactions Network and Service Management, 13 (2), pp. 267–280, 2016, ISSN: 1932-4537.
Abstract Links BibTeX Tags: Quality of Service (QoS) Resource Management Service Level Management
@article{DBLP:journals/tnsm/WangWESJG16,
title = {Using Empirical Estimates of Effective Bandwidth in Network-Aware Placement of Virtual Machines in Datacenters},
author = {Runxin Wang and Juliano Araujo Wickboldt and Rafael Pereira Esteves and Lei Shi and Brendan Jennings and Lisandro Zambenedetti Granville},
url = {https://doi.org/10.1109/TNSM.2016.2530309},
doi = {10.1109/TNSM.2016.2530309},
issn = {1932-4537},
year = {2016},
date = {2016-01-01},
journal = {IEEE Transactions Network and Service Management},
volume = {13},
number = {2},
pages = {267--280},
abstract = {Datacenter operators are increasingly deploying virtualization platforms to improve resource usage efficiency and to simplify the management of tenant applications. Although there are significant efficiency gains to be made, predicting performance becomes a major challenge, especially given the difficulty of allocating datacenter network bandwidth to multitier applications, which generate highly variable traffic flows between their constituent software components. Static bandwidth allocation based on peak traffic rates ensures SLA compliance at the cost of significant overprovisioning, while allocation based on mean traffic rates ensures efficient usage of bandwidth at the cost of QoS violations. We describe MAPLE, a network-aware VM ensemble placement system that uses empirical estimations of the effective bandwidth required between servers to ensure that QoS violations are within targets specified in the SLAs for tenant applications. Moreover, we describe an extended version of MAPLE, termed MAPLEx, which allows the specification of anticolocation constraints relating to the placement of application VMs. Experimental results, obtained using an emulated datacenter, show that, in contrast to the Oktopus network-aware VM placement system, MAPLE can allocate computing and network resources in a manner that balances efficiency of resource utilization with performance predictability.},
keywords = {Quality of Service (QoS), Resource Management, Service Level Management},
pubstate = {published},
tppubtype = {article}
}
Datacenter operators are increasingly deploying virtualization platforms to improve resource usage efficiency and to simplify the management of tenant applications. Although there are significant efficiency gains to be made, predicting performance becomes a major challenge, especially given the difficulty of allocating datacenter network bandwidth to multitier applications, which generate highly variable traffic flows between their constituent software components. Static bandwidth allocation based on peak traffic rates ensures SLA compliance at the cost of significant overprovisioning, while allocation based on mean traffic rates ensures efficient usage of bandwidth at the cost of QoS violations. We describe MAPLE, a network-aware VM ensemble placement system that uses empirical estimations of the effective bandwidth required between servers to ensure that QoS violations are within targets specified in the SLAs for tenant applications. Moreover, we describe an extended version of MAPLE, termed MAPLEx, which allows the specification of anticolocation constraints relating to the placement of application VMs. Experimental results, obtained using an emulated datacenter, show that, in contrast to the Oktopus network-aware VM placement system, MAPLE can allocate computing and network resources in a manner that balances efficiency of resource utilization with performance predictability.
2015
Juliano Araujo Wickboldt
Flexible and integrated resource management for IaaS cloud environments based on programmability PhD Thesis
Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil, 2015.
Abstract Links BibTeX Tags: Cloud Computing Cloud Networking Infrastructure as a Service (IaaS) Resource Management Software-Defined Networking (SDN)
@phdthesis{DBLP:phd/ndltd/Wickboldt15,
title = {Flexible and integrated resource management for IaaS cloud environments based on programmability},
author = {Juliano Araujo Wickboldt},
url = {http://hdl.handle.net/10183/131894},
year = {2015},
date = {2015-01-01},
school = {Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil},
abstract = {Infrastructure as a Service (IaaS) clouds are becoming an increasingly common way to deploy modern Internet applications. Many cloud management platforms are available for users that want to build a private or public IaaS cloud (e.g., OpenStack, Eucalyptus, OpenNebula). A common design aspect of current platforms is their black-box-like controlling nature. In general, cloud management platforms ship with one or a set of resource allocation strategies hard-coded into their core. Thus, cloud administrators have few opportunities to influence how resources are actually managed (e.g., virtual machine placement or virtual link path selection). Administrators could benefit from customizations in resource management strategies, for example, to achieve environment specific objectives or to enable application-oriented resource allocation. Furthermore, resource management concerns in clouds are generally divided into computing, storage, and networking. Ideally, these three concerns should be addressed at the same level of importance by platform implementations. However, as opposed to computing and storage management, which have been extensively investigated, network management in cloud environments is rather incipient. The lack of flexibility and unbalanced support for resource management hinders the adoption of clouds as a viable execution environment for many modern Internet applications with strict requirements for elasticity or Quality of Service. In this thesis, a new concept of cloud management platform is introduced where resource management is made flexible by the addition of programmability to the core of the platform. Moreover, a simplified object-oriented API is introduced to enable administrators to write and run resource management programs to handle all kinds of resources from a single point. An implementation is presented as a proof of concept, including a set of drivers to deal with modern virtualization and networking technologies, such as software-defined networking with OpenFlow, Open vSwitches, and Libvirt. Two case studies are conducted to evaluate the use of resource management programs for the deployment and optimization of applications over an emulated network using Linux virtualization containers and Open vSwitches running the OpenFlow protocol. Results show the feasibility of the proposed approach and how deployment and optimization programs are able to achieve different objectives defined by the administrator.},
keywords = {Cloud Computing, Cloud Networking, Infrastructure as a Service (IaaS), Resource Management, Software-Defined Networking (SDN)},
pubstate = {published},
tppubtype = {phdthesis}
}
Infrastructure as a Service (IaaS) clouds are becoming an increasingly common way to deploy modern Internet applications. Many cloud management platforms are available for users that want to build a private or public IaaS cloud (e.g., OpenStack, Eucalyptus, OpenNebula). A common design aspect of current platforms is their black-box-like controlling nature. In general, cloud management platforms ship with one or a set of resource allocation strategies hard-coded into their core. Thus, cloud administrators have few opportunities to influence how resources are actually managed (e.g., virtual machine placement or virtual link path selection). Administrators could benefit from customizations in resource management strategies, for example, to achieve environment specific objectives or to enable application-oriented resource allocation. Furthermore, resource management concerns in clouds are generally divided into computing, storage, and networking. Ideally, these three concerns should be addressed at the same level of importance by platform implementations. However, as opposed to computing and storage management, which have been extensively investigated, network management in cloud environments is rather incipient. The lack of flexibility and unbalanced support for resource management hinders the adoption of clouds as a viable execution environment for many modern Internet applications with strict requirements for elasticity or Quality of Service. In this thesis, a new concept of cloud management platform is introduced where resource management is made flexible by the addition of programmability to the core of the platform. Moreover, a simplified object-oriented API is introduced to enable administrators to write and run resource management programs to handle all kinds of resources from a single point. An implementation is presented as a proof of concept, including a set of drivers to deal with modern virtualization and networking technologies, such as software-defined networking with OpenFlow, Open vSwitches, and Libvirt. Two case studies are conducted to evaluate the use of resource management programs for the deployment and optimization of applications over an emulated network using Linux virtualization containers and Open vSwitches running the OpenFlow protocol. Results show the feasibility of the proposed approach and how deployment and optimization programs are able to achieve different objectives defined by the administrator.
2014
Juliano Araujo Wickboldt, Rafael Pereira Esteves, Marcio Barbosa de Carvalho, Lisandro Zambenedetti Granville
Resource management in IaaS cloud platforms made flexible through programmability Journal Article
In: Elsevier Computer Networks, 68 , pp. 54–70, 2014, ISSN: 1389-1286.
Abstract Links BibTeX Tags: Cloud Computing Cloud Networking Infrastructure as a Service (IaaS) Resource Management Software-Defined Networking (SDN)
@article{DBLP:journals/cn/WickboldtECG14,
title = {Resource management in IaaS cloud platforms made flexible through programmability},
author = {Juliano Araujo Wickboldt and Rafael Pereira Esteves and Marcio Barbosa de Carvalho and Lisandro Zambenedetti Granville},
url = {https://doi.org/10.1016/j.comnet.2014.02.018},
doi = {10.1016/j.comnet.2014.02.018},
issn = {1389-1286},
year = {2014},
date = {2014-01-01},
journal = {Elsevier Computer Networks},
volume = {68},
pages = {54--70},
abstract = {Infrastructure as a Service (IaaS) clouds are becoming a customary way to deploy modern Internet applications. Many cloud management platforms are available for one who wants to build a private or public IaaS cloud (e.g., OpenStack, Eucalyptus, OpenNebula). A common design aspect of current platforms regards their black-box-like controlling nature, where cloud administrators have few opportunities to influence how resources are actually managed (e.g., virtual machine placement or virtual link path selection). We envision that administrators could benefit from customizations in resource management strategies to achieve environment specific objectives or to enable application oriented resource allocation. In this article, we introduce a new concept of cloud management platform where resource management is made flexible by the addition of programmability to the core of the platform, with a simplified object-oriented API. We present a proof of concept prototype and an evaluation of three resource management programs on an emulated network using Linux virtualization containers and Open vSwitch running the OpenFlow protocol. Results show the feasibility of our approach and how optimization programs were able to achieve different objectives defined by the administrator.},
keywords = {Cloud Computing, Cloud Networking, Infrastructure as a Service (IaaS), Resource Management, Software-Defined Networking (SDN)},
pubstate = {published},
tppubtype = {article}
}
Infrastructure as a Service (IaaS) clouds are becoming a customary way to deploy modern Internet applications. Many cloud management platforms are available for one who wants to build a private or public IaaS cloud (e.g., OpenStack, Eucalyptus, OpenNebula). A common design aspect of current platforms regards their black-box-like controlling nature, where cloud administrators have few opportunities to influence how resources are actually managed (e.g., virtual machine placement or virtual link path selection). We envision that administrators could benefit from customizations in resource management strategies to achieve environment specific objectives or to enable application oriented resource allocation. In this article, we introduce a new concept of cloud management platform where resource management is made flexible by the addition of programmability to the core of the platform, with a simplified object-oriented API. We present a proof of concept prototype and an evaluation of three resource management programs on an emulated network using Linux virtualization containers and Open vSwitch running the OpenFlow protocol. Results show the feasibility of our approach and how optimization programs were able to achieve different objectives defined by the administrator.
Runxin Wang, Rafael Pereira Esteves, Lei Shi, Juliano Araujo Wickboldt, Brendan Jennings, Lisandro Zambenedetti Granville
Network-aware placement of virtual machine ensembles using effective bandwidth estimation Inproceedings
In: 10th International Conference on Network and Service Management, CNSM 2014 and Workshop, Rio de Janeiro, Brazil, November 17-21, 2014, pp. 100–108, IEEE Computer Society, 2014, ISSN: 2165-9605.
Abstract Links BibTeX Tags: Quality of Service (QoS) Resource Management Service Level Management
@inproceedings{DBLP:conf/cnsm/WangESWJG14,
title = {Network-aware placement of virtual machine ensembles using effective bandwidth estimation},
author = {Runxin Wang and Rafael Pereira Esteves and Lei Shi and Juliano Araujo Wickboldt and Brendan Jennings and Lisandro Zambenedetti Granville},
url = {https://doi.org/10.1109/CNSM.2014.7014146},
doi = {10.1109/CNSM.2014.7014146},
issn = {2165-9605},
year = {2014},
date = {2014-01-01},
booktitle = {10th International Conference on Network and Service Management, CNSM 2014 and Workshop, Rio de Janeiro, Brazil, November 17-21, 2014},
pages = {100--108},
publisher = {IEEE Computer Society},
abstract = {Modern datacenters rely heavily on virtualization technologies to offer customized computing and network resources on demand to a large number of tenant applications. However, efficiency in resource utilization delivered by virtualization technologies that exploit statistical multiplexing of resources across applications means that predictability in performance remains a challenge. Allocation of network bandwidth is particularly difficult, given the variability of traffic flows between the components of multi-tier applications. Static bandwidth allocation based on peak traffic rates ensures SLA compliance at the cost of significant overprovisioning, while allocation based on mean traffic rates ensures efficient usage of bandwidth at the cost of QoS violations. We describe MAPLE, a network-aware VM ensemble placement scheme that uses empirical estimations of the effective bandwidth required between servers to ensure that QoS violations are within targets specified in the SLA for the tenant application. Experimental results obtained using traffic traces collected from an emulated datacenter show that, in contrast to the Oktopus network-aware VM placement system, MAPLE is able to allocate computing and network resources in a manner that balances efficiency of resource utilization with performance predictability.},
keywords = {Quality of Service (QoS), Resource Management, Service Level Management},
pubstate = {published},
tppubtype = {inproceedings}
}
Modern datacenters rely heavily on virtualization technologies to offer customized computing and network resources on demand to a large number of tenant applications. However, efficiency in resource utilization delivered by virtualization technologies that exploit statistical multiplexing of resources across applications means that predictability in performance remains a challenge. Allocation of network bandwidth is particularly difficult, given the variability of traffic flows between the components of multi-tier applications. Static bandwidth allocation based on peak traffic rates ensures SLA compliance at the cost of significant overprovisioning, while allocation based on mean traffic rates ensures efficient usage of bandwidth at the cost of QoS violations. We describe MAPLE, a network-aware VM ensemble placement scheme that uses empirical estimations of the effective bandwidth required between servers to ensure that QoS violations are within targets specified in the SLA for the tenant application. Experimental results obtained using traffic traces collected from an emulated datacenter show that, in contrast to the Oktopus network-aware VM placement system, MAPLE is able to allocate computing and network resources in a manner that balances efficiency of resource utilization with performance predictability.
Juliano Araujo Wickboldt, Lisandro Zambenedetti Granville
Resource Management in IaaS Cloud Platforms made Flexible through Programmability Inproceedings
In: 15th Workshop da Rede Nacional de Ensino e Pesquisa, WRNP 2014, Florianópolis, Brazil, May 5-6, 2014, RNP, 2014, (Invited talk, In portuguese).
Abstract Links BibTeX Tags: Cloud Computing Cloud Networking Infrastructure as a Service (IaaS) Resource Management Software-Defined Networking (SDN)
@inproceedings{conf/wrnp/Wickboldt14,
title = {Resource Management in IaaS Cloud Platforms made Flexible through Programmability},
author = {Juliano Araujo Wickboldt and Lisandro Zambenedetti Granville},
url = {https://video.rnp.br/portal/video.action?idItem=21706},
year = {2014},
date = {2014-01-01},
booktitle = {15th Workshop da Rede Nacional de Ensino e Pesquisa, WRNP 2014, Florianópolis, Brazil, May 5-6, 2014},
publisher = {RNP},
abstract = {Infrastructure as a Service (IaaS) clouds are becoming a customary way to deploy modern Internet applications. Many cloud management platforms are available for one who wants to build a private or public IaaS cloud (e.g., OpenStack, Eucalyptus, OpenNebula). A common design aspect of current platforms regards their black-box-like controlling nature, where cloud administrators have few opportunities to influence how resources are actually managed (e.g., virtual machine placement or virtual link path selection). We envision that administrators could benefit from customizations in resource management strategies to achieve environment specific objectives or to enable application oriented resource allocation. Therefore, we present Aurora Cloud Manager platform, which introduces a new level of flexibility in resource management by the addition of programmability to the core of the platform, with a simplified object-oriented API. The platform is mostly developed in Python and brings together several third party tools, libraries, and systems in order to materialize the proposed concepts. Lower level libraries, such as Libvirt, OpenFlow, and Open vSwitch, provide abstractions for node virtualization and networking operations. The Aurora platform also ships with a set of high-level and customizable resource allocation algorithms for both deployment and optimization of cloud applications.},
note = {Invited talk, In portuguese},
keywords = {Cloud Computing, Cloud Networking, Infrastructure as a Service (IaaS), Resource Management, Software-Defined Networking (SDN)},
pubstate = {published},
tppubtype = {inproceedings}
}
Infrastructure as a Service (IaaS) clouds are becoming a customary way to deploy modern Internet applications. Many cloud management platforms are available for one who wants to build a private or public IaaS cloud (e.g., OpenStack, Eucalyptus, OpenNebula). A common design aspect of current platforms regards their black-box-like controlling nature, where cloud administrators have few opportunities to influence how resources are actually managed (e.g., virtual machine placement or virtual link path selection). We envision that administrators could benefit from customizations in resource management strategies to achieve environment specific objectives or to enable application oriented resource allocation. Therefore, we present Aurora Cloud Manager platform, which introduces a new level of flexibility in resource management by the addition of programmability to the core of the platform, with a simplified object-oriented API. The platform is mostly developed in Python and brings together several third party tools, libraries, and systems in order to materialize the proposed concepts. Lower level libraries, such as Libvirt, OpenFlow, and Open vSwitch, provide abstractions for node virtualization and networking operations. The Aurora platform also ships with a set of high-level and customizable resource allocation algorithms for both deployment and optimization of cloud applications.