[Planetlab-devel] how to tell generic slices

Thierry Parmentelat Thierry.Parmentelat at sophia.inria.fr
Tue Nov 14 08:42:30 EST 2006


Yo there

just a couple quick questions on the current contents/coding of slices

 >>> for slice in GetSlices():
...   print slice
...
{'creator_person_id': 1, 'instantiation': u'plc-instantiated', 
'slice_attribute_ids': [1, 2, 3], 'name': u'pl_conf', 'slice_id': 1, 
'created': 1163498431L, 'url': u'http://localhost.localdomain/', 
'max_nodes': 100, 'person_ids': [], 'expires': 1164708031L, 'site_id': 
1, 'node_ids': [], 'description': u'PlanetLab Slice Creation Service 
(SCS)'}
{'creator_person_id': 1, 'instantiation': u'plc-instantiated', 
'slice_attribute_ids': [4, 5, 6], 'name': u'pl_conf_vserverslice', 
'slice_id': 2, 'created': 1163498431L, 'url': 
u'http://localhost.localdomain/', 'max_nodes': 100, 'person_ids': [], 
'expires': 1164708031L, 'site_id': 1, 'node_ids': [], 'description': 
u'Default attributes for vserver slices'}
{'creator_person_id': 3, 'instantiation': u'plc-instantiated', 
'slice_attribute_ids': [], 'name': u'one_slic1', 'slice_id': 3, 
'created': 1163499099L, 'url': u'http://planet-lab.org/slic1', 
'max_nodes': 100, 'person_ids': [], 'expires': 1164708699L, 'site_id': 
2, 'node_ids': [3, 4], 'description': u'slice one_slic1 on plc plc1 in 
federation'}

==========
so:
- how do we tell a slice is defined on all nodes (I take it this is the 
case for pl_conf) that has node_ids=[] from a one that has no node 
attached yet, which will I presume also have node_ids=[]
I have that issue that I need to kind of ignore the 2 builtin slices 
when I perform caching, at least that's the smartest idea I have about 
this, which doesn't sound that smart...
can we assume these 'generic' slices are exactly the ones with a 
creator's person_id being 1

- can I safely assume the slice_attribute_ids will make sense across 
various plc's ?
the set of possible flags clearly needs to be part of the specification 
each PLC has to be compliant with.
how about using names rather that ids, like it's done for the 
instantiation method ? just thinking aloud..


thierry



More information about the Devel mailing list