Previous Work

Problem Formalization

Parameters

G = (E, V) - undirected graph

E - edges

V - nodes (i)

c(v_i) - cost function to handle ith node

s_i(t) - status of the node at time t (either available or not)

Variable

x_i - if node is used in the deployed service

Model

   Min sum(x_i*c(v_i))
   s.t.
   Asys(t) >= A0, for all t
   

where Asys(t) - the service availability, A0 - min desirable level of availability

Related works

  • Colllective availability
    • Replication
    • Erasure Coding
    • Availability prediction, Markov modeling for evaluation resource availability.
  • Workflow Scheduling - main goal is to guarantee the finalization of a time-bound set of tasks

Methodology

  • A user requests a service
  • A Manager proposes an initial feasible solution
  • The Manager tries to improve the solution

SAEDES

It's simulation-based algorithm.

Used for:

  • Get list of critical resources
  • Find defining characteristics of the modified system.

Algorithm analyse historical log files for each node and as the result provide a Bernoulli random variable ⇒ which is 1 if node satisfied some requirement (like being available the last T amount of time) and 0 otherwise.

Application

  • Computing some hard problem on some available volunteers (with defined level of availability)
  • Deploying some services (2 per client) where each of them require some resources (could be 2 per each)

Results showed that number of iteration consistently improving cost of the system in general.

Changes

Explain better difference between Grid and VC

Volunteer Computing and Desktop Grid quite similar, BUT they are still not the same. Mainly because:

  • Desktop Grid has accountability
  • Desktop Grid lacks anonymity
  • Desktop Grid uses PCs within an organization

Grind computing lets organizations such as companies, research labs and others share computational resources.

Volunteer Computing, in contrast to the Grid, are effectively anonymous, are not accountable to projects and manipulated within trust worthy environment. A very good article on the Vounteer Computing presented by David P. Anderson here.

Migrated services. Read and add something about them

Migrated service can monitor its operating conditions and autonomously relocate itself to another node when the available resource capacities at the current node fall below the acceptable limits or in terms of reaching some time limits for allocation.

Readings

available_services_over_ds.txt · Last modified: 2012/06/04 15:09 by julia
 
Except where otherwise noted, content on this wiki is licensed under the following license: CC Attribution-Share Alike 3.0 Unported
Recent changes RSS feed Donate Powered by PHP Valid XHTML 1.0 Valid CSS Driven by DokuWiki