Changes between Version 6 and Version 7 of WikiStart


Ignore:
Timestamp:
03/05/13 10:12:47 (12 years ago)
Author:
raul
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • WikiStart

    v6 v7  
    99 * '''service_variability''': File service_variability.py
    1010
    11 == Load and Transfer Test ==
     11== Load and Transfer Workload ==
    1212
    13 The aims of this test are:
    14  * Check if there is a relation between the bandwidth and the account free space.
    15  * Check if the provider limits the bandwidth when the account reaches an amount of GB transfered.
     13The objective of this workload was twofold: Measuring the maximum up/down transfer speed
     14of operations and detecting correlations between the transfer
     15speed and the load of an account. Intuitively, the first objective
     16was achieved by alternating upload and download operations,
     17since the provider only needed to handle one operation per
     18account at a time. We achieved the second point by acquiring
     19information about the load of an account in each API call.
     20The execution of this workload was continuously performed
     21at each node as follows: First, a node created synthetic files of
     22a size chosen at random from the aforementioned set of sizes.
     23That node uploaded files until the capacity of the account was
     24full. At this point, that node downloaded all the files also in
     25random order. After each download, the file was deleted.
    1626
    1727=== Implementation ===
    18 First of all, the script creates 4 different files of different sizes (25, 50, 100 and 150 MB). Once done, it starts to upload random files until the account is full and an error is returned from the provider.
     28First of all, the script creates 4 different files of different sizes (25, 50, 100 and 150 MB).
     29Once done, it starts to upload random files until the account is full and an error is returned from the provider.
    1930When this appends the script starts to download a random file from the account and removes it when the download has finished.
    2031
    2132This test will be running for 5 days approximately.
    2233
    23 == Service Variability ==
     34== Service Variability Workload ==
     35
     36This workload maintained in
     37every node a nearly continuous upload and download transfer
     38flow to analyze the performance variability of the service
     39over time. This workload provides an appropriate substrate
     40to elaborate a time-series analysis of these services.
     41The procedure was as follows: The upload process first
     42created files corresponding to each defined file size which
     43were labeled as “reserved”, since they were not deleted from
     44the account. By doing this we assured that the download
     45process was never interrupted, since at least the reserved files
     46were always ready for being downloaded. Then, the upload
     47process started uploading synthetic random files until the
     48account was full. When the account was full, this process
     49deleted all files with the exception of the reserved ones to
     50continue uploading files. In parallel, the download process was
     51continuously downloading random files stored in the account.
    2452
    2553=== Implementation ===
     
    3058
    3159- The download thread continuously lists all files in the account and downloads one randomly chosen. There will be always at least one file ("reserved.dat").
     60
     61=== Deployment ===
     62
     63Finally, we executed the experiments in different ways
     64depending on the chosen platform. In the case of PlanetLab,
     65we employed the same machines in each test, and therefore, we
     66needed to sequentially execute all the combinations of workloads
     67and providers. This minimized the impact of hardware
     68and network heterogeneity, since all the experiments were
     69executed in the same conditions. On the contrary, in our labs
     70we executed in parallel a certain workload for all providers
     71(i.e. assigning 10 machines per provider). This provided two
     72main advantages: The measurement process was substantially
     73faster, and fair comparison of the three services was possible
     74for the same period of time.