Pequeño anticipo de mi tesina

Llevo tiempo queriendo escribir sobre mi tesina en mi blog, pero la verdad es que siempre me encuentro con que, cuando por fin tengo tiempo para escribir en el blog, lo último que me apetece es seguir pensando sobre mi trabajo. A esto hay que añadirle que la recta final de la tesina está resultando ser brutal, y ultimamente me tiro unas 12 horas al día en la oficina picando codigo, ejecutando experimentos, etc. Menos mal que cada vez queda menos :-P

Pero bueno, como muchas veces me preguntan sobre qué versa mi tesina, a falta de una explicación escrita en formato de artículo de blog, por lo menos os voy a referir a unos cuantos documentos con los que os podreis hacer una idea del tema de mi tesina (desafortunadamente, todo en inglés).

  • Mi página web académica. Esta es la página web que mantengo en la Universidad de Chicago, y donde incluyo toda mi información académica (publicaciones, presentaciones, asignaturas, etc.).
  • Transparencias de una presentación que di en el ESAC Grid Workshop ’06 hace unas semanas (“Virtual Workspaces: Dynamic Virtual Environments in the Grid”), explicando el contexto de mi tesina (los “virtual workspaces”). Mi trabajo se centra especificamente sobre lo descrito en las transparencias tituladas “Virtual Clusters”.
  • Web del grupo de Virtual Workspaces de Globus
  • Descripción de mi investigación que aparece en mi web académica:

    My research lies in the intersection of Grid Computing and virtualization. In particular, my work focuses on resource management for virtual workspaces, an abstraction for execution environments that can be dynamically deployed on a grid. Currently, the most widespread models for execution management on grids are based on the job abstraction. Virtual workspace are not meant to replace the job abstraction, but to provide a more powerful and flexible execution environment when the job abstraction imposes too many constraints on the user, and can also complement the job abstraction by providing on-demand execution environments for a user’s jobs.

    A virtual workspace focuses on providing a higher quality in two aspects: environment definition (“quality of life”) and resource allocation (“quality of service”). An environment definition allows users to fully specify the execution environment they require, instead of being constrained to using the software environment configured by the resource provider. A resource allocation is an enforceable fine-grained allocation of hardware resources (CPU%, memory, disk space, network bandwidth, …), allowing for dynamic renegotiation to reflect changing requirements and conditions, instead of using coarse-grained resource requests (such as “number of processors” and “running time”) or providing little or no resource enforcement. Although the idea of creating execution environments on demand is not new (e.g. there are multiple solutions that allow users to deploy different hard drive images on a cluster, thus changing the execution environment of the worker nodes), current workspace models and implementations fail to provide adequate quality in both aspects at the same time (e.g. deploying hard drive images takes a long time, and even so it fails to provide fine-grained resource allocation to users of the cluster).

    Virtualization technologies are a promising direction for providing higher quality workspaces, by allowing execution environments to (1) be isolated from each other, (2) have enforceable resource allocations, (3) be configured flexibly, and (4) serialize and migrate. However, using VM technologies to implement workspaces poses new challenges in resource management, security, networking, etc. My research focuses on a virtual workspace resource provisioning and management model that strives to maximize quality of life and quality of service by allowing users to specify their resource requirements in terms of virtual resources, which an RM system will map to physical resources, adequately managing and scheduling the overhead of deploying and running VMs and guaranteeing that users gets exactly the resources they requested (instead of allowing overhead to “invade” the resources they requested).

FacebookTwitterMeneameDeliciousCompartir/Guardar

0 Responses to “Pequeño anticipo de mi tesina”


  • No Comments

Leave a Reply