/[storm]/storm/doc/pegboard/storm_gisp_simulation--hemppah/peg.rst
ViewVC logotype

Diff of /storm/doc/pegboard/storm_gisp_simulation--hemppah/peg.rst

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 1.6 by hemppah, Wed Jun 4 08:38:35 2003 UTC revision 1.7 by hemppah, Wed Jun 4 11:12:52 2003 UTC
# Line 58  ISSUE: Line 58  ISSUE:
58      Why GISP? Why are we using it versus some other systems?      Why GISP? Why are we using it versus some other systems?
59      What properties does it share with others, to such a degree      What properties does it share with others, to such a degree
60      that its performance might be deduced from theirs?      that its performance might be deduced from theirs?
61    
62        RESOLVED:
63        
64        Why GISP ? Why are we using it versus some other systems?
65        
66        1) According to Benja, "it's written in Java" :)
67        2) Aasy implementation (and it's implemented)
68        3) (belief that) GISP is a Kademlia implementation (which
69         is not true - only distance function is similar)
70        
71        There is also a Python implemenation of Kademlia called
72        Kashmir_.
73        
74        What properties does it share with others, to such a degree
75        that its performance might be deduced from theirs?
76            
77      -because "it's written in Java"      GISP's routing table is based on Chord's routing
78      -Kashmir - a Python implemenation of pure Kademlia      table - O(log^2 n) messages are required to join/leave
79      -with Kademlia: XOR-metric      operations and O(log n) lookup efficiency (according to
80      -with Chord: routing table (more space for cache)      original Chord publication). GISP extends Chord's routing
81        table to have more space for cached data (peer information)    
82          
83    
84  ISSUE:  ISSUE:
85                    
86      What experiments / simulations / methods are used in the literature?      What experiments / simulations / methods are used in the literature?
87            
88      -rather static simulations      RESOLVED:
89        
90        -very early (theoretical) experiments
91         -limited amount of uncontrolled (real life), great amount of
92         controlled (simulation environment)
93        -simulations have been rather static and are often favored a
94         proposed algorithm
95      -not much real life experiments      -not much real life experiments
96      -simulations       -somewhat controlled experiments
97      -formal analyses       -very fast connections between peers
98        -some form of formal analyses ("proofs") are used with,
99         e.g., Chord and Kademlia
100        -number of peers used in simulations: from 2^7 to 2^20
101            
102  Plan  Plan
103  ====  ====
# Line 81  discusses general/theoretical aspects of Line 107  discusses general/theoretical aspects of
107  if necessary, we program (rather short) test cases which will test  if necessary, we program (rather short) test cases which will test
108  the GISP/Storm P2P properties, as discussed in this document. Finally, we will  the GISP/Storm P2P properties, as discussed in this document. Finally, we will
109  collect and analyse test cases' information and publish any interesting  collect and analyse test cases' information and publish any interesting
110  discoveries.  discoveries.
111    
112    
113  The GISP protocol  The GISP protocol
114  =================  =================
115    
116  - "GISP intends leverage those (DHT) algoritms and make a practical protocol."  According to Daishi Kato, the author of GISP, GISP ''...intends leverage
117  - *distance function* is based on XOR-metric (Kademlia uses XOR)  those (DHT) algoritms and make a practical protocol.''. GISP's distance
118  - protocol (most recent) talks about XOR-metric, publication  function is based on the XOR-metric, which is similar to Kademlia's XOR-metric
119    numerical metric  (unsigned integer of XOR). The GISP protocol specification paper describes the
120  - lacks of Kademlia's binary-tree abstraction (routing table) -the benefits  XOR-metric but the original GISP publication describes only numerical
121    of Kademlia's  lookup properties are not available  metric (which is little confusing since Chord_ uses numerical distance
122  - uses Chord_'s extended routing table expect more space is used for caching  metric).
123    (Log^2 messages): asymmetric (requires stabilization protocol like Chord)  
124    and lookup process is unidirectional (a virtual ring)  GISP extends Chord's routing table to have more peer information as a cache.
125    -Chord's routing table is rigid compared to Kademlia's (well known issue)    Thus, log^2 messages are required to join/leave operations (according to the
126  - replicates in a same way as Kademlia ?  original Chord publication). Additionally, Chord's routing table is asymmetric
127  - no concurrent RPCs during lookup -- no free of choice during  (requires stabilization protocol) and lookup process is unidirectional
128    lookups ?  (in a virtual overlay ring). Original Kademlia publication states that
129  - does GISP have the property of Kademlia: leaving peer costs nothing ?  Chord's routing table is rigid compared to Kademlia's routing table.
130     - no, since GISP's routing table is based on Chord's routing table  
131  - supports "peer strength" for peer heteregeneity  Since GISP lacks of Kademlia's binary-tree-like abstraction of the routing
132  - nothing mentioned about the overlay topology -- however, routing table  table, it is evident that the benefits of Kademlia's lookup properties (over
133    is based on Chord's routing table --> toplogy resembles Chord's virtual  other DHTs) are not available in the GISP protocol, e.g., no concurrent and
134    ring  asynchronous lookups (no free of choice) and when a peer leaves the system
135    no messages are required.
136    
137    Original GISP publication describes "peer strength" as a measure for peer
138    heteregeneity but do not describe what properties are used and how this
139    value is calculated.
140    
141    For network communication the GISP protocol uses XML-RPC.
142    
143  Research problems  Research problems
144  =================  =================
# Line 120  answers to the following questions. Line 153  answers to the following questions.
153    
154  - GISP does not use Kademlia's binary-tree abstraction - does it have influences  - GISP does not use Kademlia's binary-tree abstraction - does it have influences
155    and if it does, what are the influences ?    and if it does, what are the influences ?
156      
157    - How much better/worse pure Kademlia implementation (e.g. Kashmir) is over the GISP
158      protocol if face performance and fault-tolerance ?
159                    
160  - How well GISP is able to perform in adverse conditions, e.g., a  - How well GISP is able to perform in adverse conditions, e.g., a
161    network partition occurs ?    network partition occurs ?
# Line 203  codebase. Line 239  codebase.
239  .. _GISP: http://gisp.jxta.org  .. _GISP: http://gisp.jxta.org
240  .. _Kademlia: http://kademlia.scs.cs.nyu.edu  .. _Kademlia: http://kademlia.scs.cs.nyu.edu
241  .. _Chord: http://www.pdos.lcs.mit.edu/chord/  .. _Chord: http://www.pdos.lcs.mit.edu/chord/
242    .. _Kahsmir: http://khashmir.sourceforge.net/

Legend:
Removed from v.1.6  
changed lines
  Added in v.1.7

savannah-hackers-public@gnu.org
ViewVC Help
Powered by ViewVC 1.1.26