Querying the Federation

Before we can reserve resources, it is useful to know what resources are available across the federation. This tutorial will walk you through using the Context object you created in the previous tutorial to communicate with aggregates known to geni-lib.

Finding Aggregate Locations

geni-lib contains a set of package files which have pre-built objects representing known aggregates that are ready for you to use, contained within the following Python modules:

geni.aggregate.exogeni
geni.aggregate.instageni
geni.aggregate.instageni_openflow
geni.aggregate.opengeni
geni.aggregate.protogeni
geni.aggregate.vts

While these aggregates objects will likely cover your needs, geni-lib may of course not be updated as frequently as new aggregates come online. You can find a list of the current set of aggregates on the GENI Wiki.

Getting Aggregate Information

Given that we have our previously created Context object, and a wealth of aggregate objects available to us, the GENI federation provides the ability to request two blocks of information from each aggregate - the version information (which you may have seen briefly in a previous tutorial), and a list of the advertised resources.

The result from getversion, as we saw in the previous tutorial, is reasonably concise and human readable (but also contains information about API versions and supported request formats that you may need to extract in your tools). The list of advertised resources is acquired using the listresources call, and returns a large XML document describing the available resources, which is relatively difficult to work with without a tool.

Note

We will be using GENI AM API version 2 throughout this tutorial. Some API call names will be different if you elect to interact with aggregates using AM API version 3 in the future.

  • Lets start by getting an advertisement from a single aggregate. If you built a custom context using Python code you will need to replace the code below to load your custom context:

    $ python
    >>> import geni.util
    >>> context = geni.util.loadContext()
    >>> import geni.aggregate.instageni as IGAM
    >>> ad = IGAM.Illinois.listresources(context)
    

    Now of course we have an advertisement (assuming everything went well) stored into a Python object, which is reasonably boring!

Note

If you get timeouts or failures, you may want to try a different InstaGENI aggregate (this one may be particularly busy). You can get a list of (mostly) aggregate objects by using the dir() command on the IGAM module - dir(IGAM).

  • We can simply print out the advertisement raw text to see what the aggregate sent us:

    >>> print ad.text
    <rspec xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" ......
    ...
    

    As you can see, even with this relatively small rack (5 hosts) the amount of data is significant.

  • As geni-lib has parsed this advertisement into a more functional object, we have access to data objects instead of just raw xml. For example, we can inspect the routable address space available at a site:

    >>> ad.routable_addresses.available
    167
    >>> ad.routable_addresses.capacity
    190
    
  • You may have noticed that if you just print the routable_addresses attribute, you get nothing useful:

    >>> ad.routable_addresses
    <geni.rspec.pgad.RoutableAddresses object at 0x1717f10>
    

    While we are adding online documentation for geni-lib objects, there are many objects that are undocumented. However, you can still gain some insight by using the dir() built-in to see what attributes are available:

    >>> dir(ad.routable_addresses)
    ['__class__', '__delattr__', '__dict__', '__doc__', '__format__', '__getattribute__', '__hash__',
    '__init__', '__module__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__',
    '__sizeof__', '__str__', '__subclasshook__', '__weakref__', 'available', 'capacity', 'configured']
    

    In general attributes starting with underscores are not useful to us, so we can see 3 attributes of value - available, capacity, and configured. In most cases their meanings should be obvious, so just knowing they exist even without documentation is quite helpful.

  • There are also 3 iterators that are provided with Advertisement objects - nodes, links, and shared_vlans:

    >>> for svlan in ad.shared_vlans:
    ...   print svlan
    ...
    mesoscale-openflow
    exclusive-openflow-1755
    exclusive-openflow-1756
    exclusive-openflow-1757
    ...snip...
    
  • While shared_vlans just iterates over a set of strings, node objects are much more complex and have many more attributes and nested data structures to allow you to fully inspect their state:

    >>> print dir(ad.nodes[0])
    [..., 'available', 'component_id', 'component_manager_id', 'exclusive', 'hardware_types', 'images',
    'interfaces', 'location', 'name', 'shared', 'sliver_types']
    
  • Particularly useful for the puposes of binding requests to certain nodes at a given site is the component_id:

    >>> for node in ad.nodes:
    ...     print node.component_id
    ...
    urn:publicid:IDN+instageni.illinois.edu+node+procurve2
    urn:publicid:IDN+instageni.illinois.edu+node+pc3
    urn:publicid:IDN+instageni.illinois.edu+node+pc5
    urn:publicid:IDN+instageni.illinois.edu+node+interconnect-ion
    urn:publicid:IDN+instageni.illinois.edu+node+pc1
    urn:publicid:IDN+instageni.illinois.edu+node+interconnect-campus
    urn:publicid:IDN+instageni.illinois.edu+node+pc2
    urn:publicid:IDN+instageni.illinois.edu+node+interconnect-geni-core
    urn:publicid:IDN+instageni.illinois.edu+node+pc4
    urn:publicid:IDN+instageni.illinois.edu+node+internet
    
  • Spend some time inspecting the other attributes of each node. You can get a specific node by using Python indexing on the nodes iterator:

    >>> node = ad.nodes[1]
    >>> node.component_id
    'urn:publicid:IDN+instageni.illinois.edu+node+pc3'
    

Iterating Over Aggregates

Often you will want to inspect a large number of aggregates (particularly if there are of an idential or similar type) in order to find those that have availability in the resources that you require. The aggregate modules in geni-lib provide some convenience methods for assisting in this task:

>>> import geni.aggregate.instageni as IGAM
>>> for am in IGAM.aggregates():
...     print am.name
...
ig-cenic
ig-cwru
ig-clemson
ig-cornell
ig-ohmetrodc
ig-gatech
ig-gpo
ig-illinois
...snip...

Using this iterator you can act on each aggregate in a given module with the same snippet of code.

  • Lets try getting (and saving) the getversion output from each InstaGENI site:

    >>> import json
    >>> for am in IGAM.aggregates():
    ...     print am.name
    ...     verdata = am.getversion(context)
    ...     ver_file = open("%s-version.json" % (am.name), "w+")
    ...     json.dump(verdata, ver_file)
    ...
    ig-cenic
    ig-cwru
    ig-clemson
    ...snip...
    

    This will write out a file for every aggregate (barring any exceptions) to the current directory.

Note

verdata in the above case is a Python dict object, so we need to pick a way to write it (in a human readable form) to a file. In the above example we pick serializing to JSON (which is reasonably readable), but you could also use the pprint module to format it nicely to a file as a nice string.

Exercises

We can now combine all of the above pieces, plus some Python knowledge, into some useful scripts.

  1. Move the getversion code fragment above into a standalone script, and improve it to continue to the next aggregate if any exceptions are thrown by the current aggregate (unreachable, busy, etc.).

  2. Write a script that prints out the number of availble routable IPs for each InstaGENI aggregate.