Diferencia entre revisiones de «Portada»

De LinkinWiki
Saltar a: navegación, buscar
(Investigación en Sistemas Operativos)
Línea 1: Línea 1:
 
'''
 
'''
== Investigación en Sistemas Operativos ==
+
== Research in Operating Systems ==
  
  
La rápida evolución en las tecnologías de hardware, las redes de datos y el diseño del software de usuario contrasta con el hecho de que el diseño del kernel y los mecanismos que implementan la protección y virtualización de los recursos, en los sistemas operativos actuales, no hayan tenido avances significativos en las últimas tres décadas. Esta disparidad se ve reflejada en un overhead significativo de procesamiento, conocido como intrusión del sistema operativo.
+
The fast evolution of hardware technologies, data networks and design of user software contrasts with the fact that the design of the kernel and the mechanisms which execute the protection and virtualization of the resources, in the current operating systems, has not experienced significant progress in the last three decades. This disparity is most evident in a significant processing overhead, known as operating system intrusion.
  
Esta línea de investigación se enfoca en identificar, proponer y evaluar nuevas características relacionadas con la performance y escalabilidad del kernel Linux, teniendo en cuenta los nuevos desafíos que presentan las arquitecturas multicore y la computación centrada en las redes a los sistemas operativos de la próxima generación.'''
+
This line of investigation focuses on identifying, proposing and evaluating new characteristics related to the performance and scalability of the Linux kernel, taking into account the new challenges presented by multicore architectures and network centric computing  to next generation's operating systems.'''
  
  
'''[[Un kernel asimétrico y activo]]'''
+
'''[[An active and asymmetric kernel]]'''
  
Un kernel activo es una evolución del modelo de kernel vertical y propone dedicar CPUs o cores a realizar exclusivamente actividades del kernel. Al ser un agente activo, es posible que el kernel se comunique con el hardware y el espacio de usuario utilizando mecanismos alternativos,  menos costosos que los actuales (interrupciones por hardware y software). Nuestro trabajo se centra en el estudio de las posibilidades que ofrecen estos conceptos en el tratamiento de la entrada/salida de red. Sin embargo, se han identificado posibles aplicaciones en otros dominios (virtualización, tiempo real, etc).
+
An active kernel is an evolution of the vertical kernel model and whose proposal is to have CPUs or cores perform kernel activities only. Being an active agent, it is possible that the kernel communicates with the hardware and the user space using alternative mechanisms which are less expensive than the current ones (hardware and software interruptions). Our research focuses on the study of the possibilities offered by these concepts regarding the treatment of the input/output of the network. However, possible applications in other domains have been identified (virtualization, real time, etc).
  
  
'''Llamadas al sistema asincrónicas y basadas en memoria compartida'''
+
'''Asynchronous system calls based on shared memory'''
  
La semántica de pasaje de parámetros puede tener un impacto significativo en la eficiencia de la transferencia de datos entre el kernel y las aplicaciones. La línea de trabajo se centra en ideas alternativas orientadas a incrementar la eficiencia del kernel Linux en el manejo de entrada/salida, relacionadas principalmente con la posibilidad de realizar operaciones vectorizadas (lo que favorece las operaciones scatter-gather) y utilizando memoria compartida entre el kernel y el espacio de usuario, para evitar las copias físicas en memoria de un dominio de protección al otro. Además, para lograr mayor escalabilidad, evitando el overhead que implica el scheduling de grandes cantidades de  threads, se están evaluando APIs que poseen una semántica de entrada/salida asincrónica para operaciones de sockets y archivos.
+
The semantics of the passage of parameters can have a significant impact in the efficiency of the data transference between the kernel and the applications. The line of work focuses on alternative possibilities intended to increase the efficiency of the Linux kernel in the treatment of the input/output, mainly in relation to the possibility of performing vectorized operations (which favours scatter-gather operations) and using shared memory between the kernel and the user space, to avoid physical copies in the memory from one protection domain to the other. To achieve greater scalability, avoiding the overhead implied by the scheduling of great amounts of threads, we are evaluating APIs which have asynchronous input/output semantics for socket and file operations.
  
  
'''Nuevas técnicas de virtualización'''
+
'''New virtualization techniques'''
  
Recientemente, dos nuevas técnicas de virtualización han logrado un nivel de performance sorprendente con respecto a la performance en la ejecución nativa. La paravirtualización y la virtualización asistida por hardware, por separado o combinadas (virtualización cooperativa), se posicionan como una nueva capa de abstracción estándar que revolucionará la estructura de los centros de cómputos gracias a la consolidación del hardware, un mayor nivel de seguridad y redundancia, la reducción en el uso de energía, la migración de hosts virtuales sin interrumpir los servicios, etc. La línea de trabajo se enfoca en comparar las técnicas alternativas, evaluar la integración de esta funcionalidad en el kernel y detectar las opciones que ofrecen las arquitecturas multicore a éstas tecnologías.
+
Recently, two new techniques of virtualization have achieved an outstanding level of performance in comparison with the performance of the native execution. Para-virtualization and hardware assisted virtualization, separately or combined (cooperative virtualization), place themselves as a new layer of standard abstraction which will revolutionize the structure of computing centres thanks to the consolidation of hardware, a higher level of security and redundancy, the reduction in the use of power, the migration of virtual hosts without interrupting the services, etc. This line of investigation focuses on comparing the alternative techniques, evaluating the integration of this functionality in the kernel and detecting the options offered by multicore architectures to these technologies.

Revisión de 22:13 17 jun 2007

Research in Operating Systems

The fast evolution of hardware technologies, data networks and design of user software contrasts with the fact that the design of the kernel and the mechanisms which execute the protection and virtualization of the resources, in the current operating systems, has not experienced significant progress in the last three decades. This disparity is most evident in a significant processing overhead, known as operating system intrusion.

This line of investigation focuses on identifying, proposing and evaluating new characteristics related to the performance and scalability of the Linux kernel, taking into account the new challenges presented by multicore architectures and network centric computing to next generation's operating systems.


An active and asymmetric kernel

An active kernel is an evolution of the vertical kernel model and whose proposal is to have CPUs or cores perform kernel activities only. Being an active agent, it is possible that the kernel communicates with the hardware and the user space using alternative mechanisms which are less expensive than the current ones (hardware and software interruptions). Our research focuses on the study of the possibilities offered by these concepts regarding the treatment of the input/output of the network. However, possible applications in other domains have been identified (virtualization, real time, etc).


Asynchronous system calls based on shared memory

The semantics of the passage of parameters can have a significant impact in the efficiency of the data transference between the kernel and the applications. The line of work focuses on alternative possibilities intended to increase the efficiency of the Linux kernel in the treatment of the input/output, mainly in relation to the possibility of performing vectorized operations (which favours scatter-gather operations) and using shared memory between the kernel and the user space, to avoid physical copies in the memory from one protection domain to the other. To achieve greater scalability, avoiding the overhead implied by the scheduling of great amounts of threads, we are evaluating APIs which have asynchronous input/output semantics for socket and file operations.


New virtualization techniques

Recently, two new techniques of virtualization have achieved an outstanding level of performance in comparison with the performance of the native execution. Para-virtualization and hardware assisted virtualization, separately or combined (cooperative virtualization), place themselves as a new layer of standard abstraction which will revolutionize the structure of computing centres thanks to the consolidation of hardware, a higher level of security and redundancy, the reduction in the use of power, the migration of virtual hosts without interrupting the services, etc. This line of investigation focuses on comparing the alternative techniques, evaluating the integration of this functionality in the kernel and detecting the options offered by multicore architectures to these technologies.

Herramientas personales