Storage Management in the Cloud with QoS

Application Design Document

Gal LipetzNetali AlimaChen AckermanShimi Malka

VisionThe goal of our project is to provide

performance management for enterprise disc arrays taking into account QoS specifications.

System Architecture

Data Model

Use Case: Gengerate Query

Behavioral Analysis

Algorithm Sequence Diagram

Object-Oriented Analysis

Packages Header Files – Includes all of the h files that contain the

system’s actions and activities, such as the parser, the algorithms, the correlation calculations etc.

Source Files – Includes all of the cpp files that contain the system’s actions and activities.

Header System Objects – Includes all of the system’s data objects h files, such as the extents, LUNs, etc.

Source System Objects – Includes all of the system’s data objects cpp files.

Gui - Includes all of the system’s Interface classes.

Resource Files – Includes the configuration and input files.

Main Classes Parser – Responsible for parsing the input

files. Writer – Responsible for creating relevant

data into output files. Sorter – Responsible for all sorting algorithms

used in the data placements algorithms. SystemView – All of the system’s parameters

are saved in this class.

Main Classes Cont… LunsCorrelation – Calculates the similarity between

LUNs. TimeSliceCalculations – Calculates the response

time and damage for each LUN, for the current time slice.

OptimalAlgo – The optimal data placement algorithm, used for comparisons with other algorithms.

NextAlgo – A data placement algorithm in which memory assignment is based on previous time slices.

Main Functions Algorithm: runAlgo()

Precondition : SystemView object and all DeviceData objects were created. Postcondition : The algorithm’s hit ratio, all of the devices’ utilizations, and all LUNs’ response times for each time slice were calculated.

DeviceData: calcUtilization ()Precondition : All trace data was parsed and one of the algorithms started running.Postcondition : All of the three DeviceData objects contain the new utilization for the device.

LunTimeStatistics: responseTime() Precondition : The LUN’s extents’ device placements were calculated. Postcondition : the response time for the current time slice was updated.

System States

Input Files

Data Desired

Gui

Parse Input

Get Data

Run Algorithm

Simulate

Main Menu

Solve Query

Results

QoS InputQuery

Ask Query Device Allocation

Parse Data

Device Groups

Store Data

Solve Algorithm

Result QoS Input

Algorithm Rules and Decisions

Parsed Data

Data Desired

Algorithm Chosen

User Interface Draft

Testing

Data Simulator Simulates data from the machine. Receives parameters and creates a data file with the

desired characteristics. Using the parameters we can calculate what the

system’s outcome should be and verify the correctness of our system.

Real Data System runs with real data. System completes tasks without errors. Results required are shown and relevant graphs created.

Non Functional Requirements Testing Speed

Algorithm Simulation and parsing completes in under 12 hours.

Queries give an answer after 2 minutes.

Reliability When using simulated data the system’s outcome will always

as predicted.

Platform Constraints Input files – Does not crash with incorrect input format. Output files – Output is received in the required format with

the relevant graphs.

Task List

1. Complete models, including priorities2. Add damage calculations and graphs3. Calculate System Size4. Optimize Algorithms5. Optimize Parsing and Memory Usage6. Implement Additional Algorithms7. Test all logics in the project 8. Create and test GUI 9. Draw conclusions and derive recommendations