Updates in highly unreliable, replicated peer-to-peer systems
read more
Citations
P-Grid: a self-organizing structured P2P system
Method and system for downloading updates
GridVine: building internet-scale semantic overlay networks
Review: A survey on content-centric technologies for the current Internet: CDN and P2P solutions
Power. Law.
References
Pastry: Scalable, Decentralized Object Location, and Routing for Large-Scale Peer-to-Peer Systems
Epidemic algorithms for replicated database maintenance
Freenet: a distributed anonymous information storage and retrieval system
The dangers of replication and a solution
Managing update conflicts in Bayou, a weakly connected replicated storage system
Related Papers (5)
Pastry: Scalable, Decentralized Object Location, and Routing for Large-Scale Peer-to-Peer Systems
Frequently Asked Questions (19)
Q2. What have the authors stated for future works in "Updates in highly unreliable, replicated peer-to-peer systems" ?
In this paper, the authors have used heuristics to find proper parameters, but they plan to explore the possibility of both feed-back and feed-forward to evolve a proper mechanism of parameter tuning using local knowledge. To verify the correctness of the analysis if some of the simplifying assumptions are relaxed, the authors plan to use simulations, which will also help us investigate whether there is bimodal2 behavior [ 4, 13 ] even in the assumed environment of very low peer presence. The authors plan to use their PGrid peer-to-peer system as a testbed for the implementation and practical tests of the algorithm. In such a scenario a relatively reliable network backbone would exist and thus would make possible further performance improvements.
Q3. What is the novel approach of their work?
The novel approach of their work is the use of push/pull in the context of replicas being offline long and frequently, and in the significant reduction of message overhead in the push phase.
Q4. What is the important metric that can be used to tune parameters?
The number of duplicate messages received by a replica also provides an essential, locally available metric that may utilise to tune parameters ,'p$ S ! and "on .
Q5. What is the main advantage of the algorithm?
Another major advantage of the algorithm is that it is totally decentralised and uses no global knowledge but exploits local knowledge instead.
Q6. What are some of the typical applications where data is added, deleted, or updated frequently?
Other typical applications where new data items are added, deleted, or updated frequently by multiple users are bulletin-board systems, shared calendars or address books, e-commerce catalogues, and project management information.
Q7. What is the purpose of the push phase?
In the push phase the algorithm uses a new mechanism, apart from traditional feedback and probabilistic methods to propagate a rumor, to avoid many duplicate messages by propagating a partial list of peers to which a particular message has already been sent.
Q8. What is the performance criterion for the pull phase of updates?
Their performance criterion for this analysis is primarily the number of messages that are generated as part of a single update, compared to the extent to which the update propagates among the online population.
Q9. What is the assumption of a very small bmd?
The assumption of a very small bmd is justified because a single push round will take a very small time (network delay for a single message), and unless there is any kind of catastrophic failure, a very small number of peers will suddenly decide to go offline.
Q10. What is the common economic paradigm for data replication in Mariposa?
Other economic paradigms to maintain distributed data replicas include [15, 16, 21] where a primary copy model is used to provide one-copy serializability.
Q11. How can the authors tune the push phase?
Tuning the push phase may not only be done through feedback mechanisms (to determine when to stop pushing), but also by a speculative (feed-forward) mechanism.
Q12. What is the way to test the consistency of the analysis?
To verify the correctness of the analysis if some of the simplifying assumptions are relaxed, the authors plan to use simulations, which will also help us investigate whether there is bimodal2 behavior [4, 13] even in the assumed environment of very low peer presence.
Q13. Why is it possible that different push rounds live in the same network?
It is indeed possible that because of variation in network latency, messages of different push rounds live in the network at the same instant of time.
Q14. What is the main contribution of this paper?
Another significant contribution of this paper is an analytical model of the gossiping algorithm unlike most of the literature which relies on simulation results.
Q15. What is the probability of information dissemination in such applications?
Most of these systems operate with a relatively high degree of imperfect knowledge, which is why probabilistic guarantee of information dissemination in such application scenarios is sufficient.
Q16. how many replicas are needed to search a database?
An intuitive explanation for such numbers is that if the authors need a 99.9% success guarantee for a search and only 10% of the replicas are online on average, then a serial search will need about 65 attempts (since ).
Q17. What is the definition of the second category of a rumor spreading algorithm?
The first category is defined by whether nodes use feedback from other nodes (for example, whether they already know the rumor or not) and thus decide on their future course, or not (generally called “blind” then).
Q18. What is the way to reduce the number of duplicate messages sent to a replica?
This strategy will only be effective for short time intervals, since over a period of time,is expected to be online according to a random distribution for all the replicas.
Q19. What is the strategy for reducing the probability of forwarding updates?
Fig. 4 indicates that the best strategy is to reduce the probability of forwarding updates with the increase in number of push rounds, which eliminates many unnecessary messages.