/[gzz]/gzz/Documentation/misc/hemppah-progradu/masterthesis.tex
ViewVC logotype

Diff of /gzz/Documentation/misc/hemppah-progradu/masterthesis.tex

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

revision 1.89 by hemppah, Thu Feb 27 09:38:48 2003 UTC revision 1.90 by hemppah, Thu Feb 27 10:41:47 2003 UTC
# Line 284  networks\footnote{In power-law networks Line 284  networks\footnote{In power-law networks
284  links and major of peers have low nuber of neighbor links.} and they have found that by  links and major of peers have low nuber of neighbor links.} and they have found that by
285  instructing peers forwarding queries to select high degree peers the data lookup's  instructing peers forwarding queries to select high degree peers the data lookup's
286  performance increases signficantly. As a result, some of the most recent loosely  performance increases signficantly. As a result, some of the most recent loosely
287  structured Peer-to-Peer system have adopted this method with some modifications  structured Peer-to-Peer systems have adopted this method with some modifications
288  \cite{gnutella2url}, \cite{shareazaurl}, \cite{fasttrackurl}, \cite{morpheusurl},  \cite{gnutella2url}, \cite{shareazaurl}, \cite{fasttrackurl}, \cite{morpheusurl},
289  \cite{kazaaurl}. Figures \ref{fig:gnutella_overlay_supernodes} and \ref{fig:gnutella_overlay_cluster}  \cite{kazaaurl}, \cite{jxtaurl}, \cite{jxtaoverview}, \cite{botros01jxtasearch},
290  illustrated two possible variations of power-law overlay networks. However, it's  \cite{ganesan02yappers}.
291  not clear whether this algorithm is scalable or not, as majority of the query  Figures \ref{fig:gnutella_overlay_supernodes} and \ref{fig:gnutella_overlay_cluster}
292  request are sent only to the high degree peers, making them stress the overhead  illustrated two possible variations of power-law overlay networks. All the systems
293  of nearly entire system.  share the property of that high degree peers maintain index of all other peers
294    they know about. However, it's not clear whether this algorithm is scalable or not,
295    as majority of the query request are sent only to the high degree peers, making
296    them stress the overhead of nearly entire system.
297    
298  \begin{figure}  \begin{figure}
299  \centering  \centering
# Line 307  of nearly entire system. Line 309  of nearly entire system.
309  \label{fig:gnutella_overlay_cluster}  \label{fig:gnutella_overlay_cluster}
310  \end{figure}    \end{figure}  
311    
312    Additionally, there has been other improvements also. In iterative deepening
313    \cite{yang02improvingsearch}, multiple breadt-first searches are initiated
314    with successively larger TTL depth limits, until either the query is satisfied,
315    or the maximumum depth $D$ has been reached. To perform a data lookup, query
316    originator starts a flood with small TTL value. If the search is not succesful,
317    the query originator increases the TTL value and performs another flood. This
318    process is repeated until the desired data is found or maximumum depth $D$
319    has been reached. Expanding ring, proposed by Shenker et al., \cite{lv02searchreplication},
320    is similar to iterative deepening techique. With these techniques, search
321    may not be fast when desired data item requires many consecutive flooding rounds.
322    
323    Directed breadt-first search \cite{yang02improvingsearch} optimizes the original
324    breadt-first searche in way that peer selects neighbors with many quality results
325    may be reached, thereby maintaining the the quality of costs and decreasing the amount
326    of messages sent to network. Alpine Peer-to-Peer system \cite{alpineurl} uses
327    somewhat similar method when performing data lookups.
328    
329    Local indices \cite{yang02improvingsearch} in one variation of active caching.
330    In this scheme, each peer maintains an index over the data of all nodes within
331    $h$ hops of itself, where $h$ is a system-wide variable, called radius of the
332    index\footnote{In normal BFS case, the value of $h$ is 0, as peer only has index
333    over its local content.}. Mutual index caching architecture, as proposed in
334    \cite{osokine02distnetworks}, is one variation of local indices techique.
335    
336    In random walk approach \cite{lv02searchreplication}, peer forwards a query to
337    randomly selected neighbor. The basic random walk approach decreases the
338    overhead generated by messages. On the other hand, basic random walk approach
339    has poor response time. As suggested in \cite{lv02searchreplication},
340    random walk approach can be done more effective by introducing
341    multiple ''walkers''. Freenet \cite{clarke00freenet} Peer-to-Peer system uses
342    random walk searches in query lookups. Indeed, Freenet's query resembles
343    depth-first traversal and peers' routing tables are dynamically built
344    using caching. This is an outcome of Freenet's main design priciples,
345    i.e., anonymity.
346    
347    Previously presented improvements are only partial solutions. Obviously, more
348    research is required to make loosely structured approach's data lookup more
349    scalable and effective.
350    
351    principles
352    
353    
354  power-law disribution  power-law disribution
# Line 337  to a single neighbor (determined by ID) Line 377  to a single neighbor (determined by ID)
377  before forwarding the query to another neighbor (if query not ok), or forwarding results back  before forwarding the query to another neighbor (if query not ok), or forwarding results back
378  to the query source (if query ok)  to the query source (if query ok)
379    
 BFS  
 -Search results are fast, because BFS sends queries to every possible nodes  
 -Wastes resources, because BFS sends queries to every possible nodes  
   
   
 DFS  
 -Poor response time, beecause each node processes the query sequentially...  
 -...and thereby minimazing cost  
   
   
   
380    
381    
 \cite{yang02comparinghybrid}  
   
382  \subsection{Formal definition}  \subsection{Formal definition}
383    
384  -let S be the aggregate of all services s in system (data, service, computing power)  -let S be the aggregate of all services s in system (data, service, computing power)
# Line 378  Notice: pros and cons are not presented Line 405  Notice: pros and cons are not presented
405  -with every lookup query, a node determines how proficient a given node is to another node's objectives  -with every lookup query, a node determines how proficient a given node is to another node's objectives
406    
407    
408  Freenet \cite{clarke00freenet}  
409  Improve Freenet performance with small worlds \cite{zhang02using}  Improve Freenet performance with small worlds \cite{zhang02using}
410  Milgram's small world experiment \cite{milgram67smallworld}  
 Small worlds \cite{adamic99small}  
411  \cite{ramanathan02goodpeers}  \cite{ramanathan02goodpeers}
412  \cite{kleinberg99small}  \cite{kleinberg99small}
413  \cite{watts00dynamics}  \cite{watts00dynamics}
414  \cite{nips02-Kleinberg}  \cite{nips02-Kleinberg}
 \cite{ganesan02yappers}  
 \cite{gnutellaurl}  
415    
416  \cite{jxtaurl}  
417  \cite{jxtaoverview}  
418  \cite{botros01jxtasearch}  
419  \cite{kato02gisp}  \cite{kato02gisp}
420  \cite{alpineurl}  
421  \cite{joseph02neurogrid}  \cite{joseph02neurogrid}
422    
423    

Legend:
Removed from v.1.89  
changed lines
  Added in v.1.90

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