Home
This Title All WIREs
WIREs RSS Feed
How to cite this WIREs title:
WIREs Comp Stat

CLOUD computing risk management for server‐farm repair rates, consumer load cycle, server‐farm repair crew count, and additional servers

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

A discrete event simulator, CLOURAM: CLOUD Risk Assessor and Manager, algorithmically estimates the risk indices in modern‐day CLOUD computing scenarios with tangible risk management targets that are favorable to the intractably tedious, theoretical Markov solutions or hand calculations overly limited in scope. The goal is to improve the operational quality of CLOUD by optimizing the number of servers for capacity addition and optimizing the final repair crew count. We too optimize the server unit repair rates, and the consumer load cycle by curbing the demand using Linear Programming (LP)based optimization with the proper objective functions and constraints. Small and large CLOUD systems are simulated with cost and benefit comparisons. The 2‐state (UP and DN) or 3‐State (UP, DN, and DER) units statistically fail and recover with Negative Exponential or Weibull densities. WIREs Comput Stat 2018, 10:e1424. doi: 10.1002/wics.1424

This article is categorized under:

  • Statistical and Graphical Methods of Data Analysis > Reliability, Survivability, and Quality Control
  • Algorithms and Computational Methods > Networks and Security
  • Statistical Models > Simulation Models
  • Algorithms and Computational Methods > Linear Programming
Sample illustration of feasible transitions from Figure , where the transition rates are defined in Table
[ Normal View | Magnified View ]
Three‐State Markov Diagram of a repairable hardware unit with UP, DOWN and DER states from Table
[ Normal View | Magnified View ]
The JAVA‐coded simulation to a simple 2‐unit 3‐State experimental CLOUD system with the tabulated data in Figure
[ Normal View | Magnified View ]
Plot of LOLP index of 13.5% with 26,237 GB for 2010 data.txt rising to 26,937 decreases to 9.5%
[ Normal View | Magnified View ]
LOLP = 13.5% with 26,237 GB for 2010 data with added capacity to 26,937 decreases to LOLP = 9.5% for the input data
[ Normal View | Magnified View ]
Plot of LOLP index of 13.5% with 443 crews for data2010.txt reducing to 93 crews increases to 32.6% for the cost data
[ Normal View | Magnified View ]
LOLP index of 13.5% with 443 crews for 2010 data.txt reduced to 93 crews (benefit↑) increases to 32.63% (cost↑)
[ Normal View | Magnified View ]
Plot of LOLP index of 5.43% with 348 crews for data 2005 reducing to 148 crews increases to 10.32% given the cost data
[ Normal View | Magnified View ]
LOLP index of 5.44% with 348 crews for the 2005 data reduced to 148 crews; LOLP increases to 10.32% for given input
[ Normal View | Magnified View ]
Large‐scale CLOUD system of 2005 data with 348 units processed for optimal load planning; however, with no feasible solution due to inappropriate choice of input parameters, resulting in a suggestion to retry with a new set of parameters
[ Normal View | Magnified View ]
Large CLOUD of 2010 data with 443 units and 500 loads processed for optimal load planning per Table , cols. 9–12
[ Normal View | Magnified View ]
Large CLOUD of 2010 data with 443 units and 500 loads prior to load planning per Table column 11 with new multiplier
[ Normal View | Magnified View ]
Large‐scale CLOUD of 2010 data with 443 units processed for optimal repair rate planning per Table , cols. 6–8
[ Normal View | Magnified View ]
The 23‐unit, 23‐cycle small CLOUD system after the load cycle optimization for an output of LOLP = 0.3197 (31.97%) improved by a relative percentage of 43.9% from a previous non‐optimized LOLP = 0.5698 (56.98%) [Correction added on 15‐February 2018 after first online publication: Figure A7 has been updated to show the correct profit value of $220.12 as in Table A.2's Profit or Loss column14 from the 23‐23 row of the Data column 1.]
[ Normal View | Magnified View ]
The 23‐unit, 23‐load cycle CLOUD for LP optimization using JAVA input matrix to generate optimized load solution vector on the LHS as in Table A1, column 15 identical to that of Figure 's feasible solution only in a different order. On the RHS, the MIN sum of the load vector as the objective function of Equation and constraints of Equation (from row 1 to 46) to Equation (6–7) (from row 47 to 48) and Equation (row 49) e.g. Sum New Lk ≥ [{20% of the total available capacity = 2048.83} + {Sum Old Lk = 23 × 1000}] = 18810
[ Normal View | Magnified View ]
The 23‐unit, 23‐load cycle CLOUD with the LP optimization using EXCEL to generate the optimized load cycle solution vector as in Table A1's column 14 where the objective is to minimize the sum of the load values provided the binding constraints
[ Normal View | Magnified View ]
The 23‐unit, 23‐load cycle small CLOUD after the repair rate JAVA‐LP optimization for output of LOLP =0.5322(53.22%)
[ Normal View | Magnified View ]
The 23‐unit, 23‐load cycle small CLOUD before the repair rate JAVA‐LP optimization for output of LOLP = 0.5698 (56.98%)
[ Normal View | Magnified View ]
The 23‐unit 23‐cycle CLOUD with the LP optimization using EXCEL to generate the optimized repair rate solution vector as in column 4 of Table where the objective is to maximize the repair rates’ sum provided the binding constraints of Equation to
[ Normal View | Magnified View ]
(a) Illustrative CLOUD computing. (b) Core groups of server farms reaching user load centers
[ Normal View | Magnified View ]
The exact density and survival (and cumulative) density plots for Figures and are illustrated. Note the subtle exact probability density in red is out of scale
[ Normal View | Magnified View ]
Compound Poisson, otherwise Negative Binomial (provided a certain statistical assumption) distribution plot of the Loss of Load Expected (LOLE) metric with Mean = 1179, q = 13.09 is slightly skewed to the right as mean = 1179 > median = 1174
[ Normal View | Magnified View ]
Compound Poisson or otherwise Negative Binomial (provided a certain statistical assumption) distribution, the analysis of this Loss of Load Expected (LOLE) metric with M (Mean) = 1179, Median = 1174 and q = 13.09 as inspired from Figure 28's outcomes
[ Normal View | Magnified View ]
The WEIBULL output, LOLP = 13.46% with two‐state units for 2010 data in Table column 6‐WEI fairly close to 13.5%
[ Normal View | Magnified View ]
The WEIBULL output, LOLP = 5.81%, with two‐state units for 2005 data in Table column 6‐WEI fairly close to 5.45%
[ Normal View | Magnified View ]
The 50% derated output for the 2010 data per Table column 6‐DER where LOLP = 19.72% up from LOLP = 13.5%
[ Normal View | Magnified View ]
Plot of LOLP index of 7.69% with 20,950 GB for 2005 data rising to 21,350 decreases to 5.57% inspired by Figure
[ Normal View | Magnified View ]
Figure multistate input data applied to Product Planning for 2005 with Derated50% arrangement → LOLP = 5.57%, not tabulated
[ Normal View | Magnified View ]
The derated output, LOLP = 7.68%, to Figure 's input data with three‐State units for 2005 data in Table 's column‐6DER50%
[ Normal View | Magnified View ]
The three‐state input transition rates for the data 2005; note that the UP → DOWN and DOWN → UP rates are the same as before
[ Normal View | Magnified View ]
The MARKOV state space matrix steady‐state solution for the nine states regarding two‐unit system where P1 denotes both units being in the UP state. Note the theoretical P1 = 0.36 solution in favorable accord with Figure 's LSE = {1LOLP} = 0.37
[ Normal View | Magnified View ]

Browse by Topic

Algorithms and Computational Methods > Linear Programming
Algorithms and Computational Methods > Networks and Security
Statistical and Graphical Methods of Data Analysis > Reliability, Survivability, and Quality Control
Statistical Models > Simulation Models

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts