[Haizea] Developer Docs...

[J Arshad]

Hi,

I have tried the "haizea-convert-data" command after dumping the two files 
in my haizea source directory but I ran into the following error.

Could you please suggest something...

Regards

Junaid

haizea-convert-data -t per-lease -f csv /var/tmp/haizea/results.dat
lease_id,waiting_time,slowdown
Traceback (most recent call last):
   File "/home/csunix/ja/bin/haizea-convert-data", line 7, in <module>
     c.run()
   File "/home/csunix/ja/lib/python/haizea/cli/commands.py", line 299, in 
run
     print ",".join(attrs + [`lease_id`, `waitingtimes[lease_id].seconds`, 
`slowdowns[lease_id]`])
TypeError: cannot concatenate 'str' and 'list' objects



On Tue, 16 Dec 2008, Borja Sotomayor wrote:

> Hi,
>
>> Firstly, One of the files I want to change is the sample_trace.conf i.e.
>> I want to have resources with different capabilities. Currently, all the
>> nodes are assumed to have the same characteristics e.g. CPU,Mem etc. using
>> the "resources" option in the sample_trace.conf( please correct me if I am
>> wrong). However, I want to add an extra characteristic e.g. availability
>> of "MRI device" with the resources and for this, I assume that not all the
>> resources will have such device. I tried to have multiple "resources"
>> options but, of course, that does not seem to work. Could you please
>> suggest something in this regard...
>
> Adding new resource "dimensions" is not too hard, and can be done simply
> by modifying the "resources" option of the [simulation] section.
> However, I'm assuming what you need is to create a simulated resource
> pool with heterogeneous resources (some having an MRI device and others
> not, the same way some nodes could have X memory and others Y memory).
> This is currently not supported, as nodes are assumed to be homogeneous
> (actually, the underlying code supports having heterogeneous nodes;
> there is just currently no way of specifying this in the configuration
> file).
>
> This will eventually get added to Haizea, but is not something that has
> a high priority for me right now. If you are interested in implementing
> this feature, I can point you to the parts of the code that would have
> to be modified.
>
>
>> Secondly, I believe the "changepoints" logic for the Best Effort leases
>> comes into play with "suspension" being true and/or the resource is
>> migrated/resumed. Please correct me if I am wrong again...
>
> "changepoints" are used in several parts of the scheduling code, and
> represent points in the slottable where there is a change in resource
> usage (e.g., because a lease ends). So, they are relevant to a number of
> scheduling tasks, including deciding when a lease would have to be
> suspended (at the changepoint where resource availability drops below
> the point where it can meet the resource requirements of the lease).
>
>
>> Finally, can anyone please suggest ways to analyze the output of the
>> simulation.
>
> In the current version of Haizea, all simulation data is saved to a
> datafile that has to be analysed programmatically. However, I recently
> wrote a simple command that takes a datafile (or a collection of
> datafiles) and converts them to a more usable format (currently, only
> CSV is supported; in my case, I pipe that CSV to an R script that
> produces graphs based on the data). I've commited it to SVN, in case you
> want to use the command, or at least look at it to better understand how
> the datafile can be processed. The command is called
> "haizea-convert-data". You can see the code here:
>
> https://phoenixforge.cs.uchicago.edu/haizea/browser/trunk/bin?file=haizea-convert-data
>
> and in class "haizea_convert_data" at the end of this file:
>
> https://phoenixforge.cs.uchicago.edu/haizea/browser/trunk/src/haizea/cli?file=commands.py
>
> If you want to use the command, I wouldn't recommend checking out all of
> Haizea, since there have been some interface changes elsewhere in the
> code that may break work you've already done. Instead, I suggest you
> check out the two files above (and use them to replace the same files in
> your current Haizea installation)
>
> The command is used like this:
>
> 	haizea-convert-data -t <type> -f <format> <datafile(s)>
>
> <type> is either "per-experiment" or "per-lease". When using
> "per-experiment", the command will produce a CSV file with one line per
> datafile (i.e., one run of the simulator), and will include aggregated
> information (currently, just the time required to run all the best
> effort leases). When using "per-lease", then there will be one line per
> lease with information such as the waiting time of the lease and its
> slowdown.
>
> The only <format> supported right now is "csv".
>
> Anyway, I hope this help. The next release of Haizea will include this
> command and documentation on how to use it.
>
> Cheers!
>