Gustavo Pantuza

QUIC protocol is considered an experimental environment and a straight forward evolution of the T protocol. Applications created on top of QUIC demonstrate being more efficient than the regular HTTPS stack. Computational offload to kernel space is used as performance optimization for modern applications and brings with it a series of architectural and algorithmic challenges. This work presents eQUIC Gateway, a kernel space module that filter packets based on real time information exchanged… Mostrar más

QUIC protocol is considered an experimental environment and a straight forward evolution of the T protocol. Applications created on top of QUIC demonstrate being more efficient than the regular HTTPS stack. Computational offload to kernel space is used as performance optimization for modern applications and brings with it a series of architectural and algorithmic challenges. This work presents eQUIC Gateway, a kernel space module that filter packets based on real time information exchanged with a space QUIC application. Throughout kernel offload technique for filtering packets, eQUIC Gateway increased the packets throughput in 30.9%, reduced 65 % of the average HTTPS request duration under attack and reduced by 26.4% the U time consumed for filtering packets. The source code of eQUIC Gateway is public available on Github. Mostrar menos

A alocação de núcleos para threads de programas é um problema complexo e de grande custo computacional dentro de ambientes Unix. O escalonador Caladan é uma solução que otimiza na escala dos microssegundos a forma como threads e núcleos são escalonados. O sistema Danian proposto, melhora o algoritmo de escolha de threads para um núcleo a ser escalonado através de memoization. O custo dessa operação foi reduzido de O (n) para O (1), o tempo de U reduzido em 7%, a latência de calda reduziu em… Mostrar más

A alocação de núcleos para threads de programas é um problema complexo e de grande custo computacional dentro de ambientes Unix. O escalonador Caladan é uma solução que otimiza na escala dos microssegundos a forma como threads e núcleos são escalonados. O sistema Danian proposto, melhora o algoritmo de escolha de threads para um núcleo a ser escalonado através de memoization. O custo dessa operação foi reduzido de O (n) para O (1), o tempo de U reduzido em 7%, a latência de calda reduziu em 3% para o experimento Synthetic e 5% para o experimento Netperf. Mostrar menos

ing low latency processing in high throughput environments is a growing trend in recent years. Shenango allows servers to fastly process network packets, preserving high U efficiency through a congestion detection algorithm and using its IOKernel, a dedicated component that orchestrates the packet allocation between new cores. This system though, allocates new cores using the round-robin scheduler policy. This work presents Calisto, a new system that implements an efficient scheduler… Mostrar más

ing low latency processing in high throughput environments is a growing trend in recent years. Shenango allows servers to fastly process network packets, preserving high U efficiency through a congestion detection algorithm and using its IOKernel, a dedicated component that orchestrates the packet allocation between new cores. This system though, allocates new cores using the round-robin scheduler policy. This work presents Calisto, a new system that implements an efficient scheduler policy that reduces processing overhead and promoting packet flow affinity, reducing the latency by up to 10%. Mostrar menos

Serverless computing is a novel application deployment model that provides scaling, fine-grained billing, and platform decoupling for developers. This model has the potential for changing the landscape of technology development for s and cloud providers. In this short course, we present the Serverless Computing state of the art: concepts, limitations, use cases, research opportunities, caveats, and a view on some current platforms.

Master's thesis presented to the post graduation program in Computer Science of the Institute of exact sciences from Federal university of Minas Gerais

We propose a traffic engineering solution using Software Defined Networks. Our goal is to enforce link utilization dividing the traffic over multiple links through a simple, easy to
implement, low cost, efficient and scalable solution. The solution can be deployed on data centers or at the edge of an
Autonomous System.

Software Defined Networks (SDN) is an emergent architecture that is dynamic, flexible, manageable, low cost, consistent with the dynamics of the modern applications. This paper shows a network representation model using a graph as the control plane of an SDN controller. The graph approach provides a globally consistent view of the network in real time. Our experiments show that graphs are a reliable representation of the real network, simplifying management in Software Defined Networking.

As redes definidas por software (SDN) representam uma arquitetura emergente dinâmica, flexível, gerenciável e de baixo custo, condizente com a dinamicidade das aplicações atuais. Essa arquitetura desacopla o plano de controle do plano de dados. Redes tipicamente são representadas em forma de grafos. Em função disso, esse trabalho apresenta um modelo de representação da rede em grafos através do plano de controle de um controlador SDN. O protocolo OpenFlow é um meio para construção de soluções… Mostrar más

As redes definidas por software (SDN) representam uma arquitetura emergente dinâmica, flexível, gerenciável e de baixo custo, condizente com a dinamicidade das aplicações atuais. Essa arquitetura desacopla o plano de controle do plano de dados. Redes tipicamente são representadas em forma de grafos. Em função disso, esse trabalho apresenta um modelo de representação da rede em grafos através do plano de controle de um controlador SDN. O protocolo OpenFlow é um meio para construção de soluções em SDN. Esse trabalho é baseado no POX, um controlador SDN compatível
com dispositivos OpenFlow. Essa abordagem em grafos fornece uma visão global da rede em tempo
real e com consistência. Os experimentos demonstram que o grafo proposto representa de maneira fiel o estado da rede assim como suas mudanças em função dos eventos ocorridos ao longo do tempo dentro da rede, facilitando assim, o gerenciamento em Redes Definidas por Software. Mostrar menos

Core allocation for application threads is a problem of reasonable complexity and computational cost inside Unix systems. Caladan scheduler is a solution aiming to reduce the cost of how threads and cores are allocated in microsecond scale. The Danian system optimizes through memoization the thread picking algorithm that picks the best thread for a given core. Such improvements have direct impact on applications distributed across networks on a data center. Thread picking operation cost dropped… Mostrar más

Core allocation for application threads is a problem of reasonable complexity and computational cost inside Unix systems. Caladan scheduler is a solution aiming to reduce the cost of how threads and cores are allocated in microsecond scale. The Danian system optimizes through memoization the thread picking algorithm that picks the best thread for a given core. Such improvements have direct impact on applications distributed across networks on a data center. Thread picking operation cost dropped from O(n) to O(1), the U time reduced 7%, the tail latency reduced 3% on Caladan Synthetic experiment and 5% on the Netperf experiment. Mostrar menos