Scribd
Upload
134 views

Distributed System Assignment 3

This document contains a student's responses to 4 questions about distributed systems. 1. It explains why a multicast-based mutual exclusion algorithm would be inefficient if one process frequently needs access to the critical section before others. It proposes changing process states to improve performance in this scenario. 2. It lists the main guarantees expected of conventional servers as consistent object states and continuous availability. 3. It discusses anonymity in peer-to-peer systems and ways attacks could compromise it, such as through network monitoring. It suggests generating dummy messages to improve resistance. 4. It argues that leader election is possible in a synchronous ring where all but one node have the same ID, by having the unique node declare itself

Uploaded by

Vaibhav Singh
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
134 views

Distributed System Assignment 3

This document contains a student's responses to 4 questions about distributed systems. 1. It explains why a multicast-based mutual exclusion algorithm would be inefficient if one process frequently needs access to the critical section before others. It proposes changing process states to improve performance in this scenario. 2. It lists the main guarantees expected of conventional servers as consistent object states and continuous availability. 3. It discusses anonymity in peer-to-peer systems and ways attacks could compromise it, such as through network monitoring. It suggests generating dummy messages to improve resistance. 4. It argues that leader election is possible in a synchronous ring where all but one node have the same ID, by having the unique node declare itself

Uploaded by

Vaibhav Singh
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

DistributedSystem

Assignment3
-Submittedby:
VaibhavSingh(14B00033)

1. Inacertainsystem,eachprocesstypicallyusesacriticalsectionmanytimes
beforeanotherprocessrequiresit.ExplainwhyRicartandAgrawalas
multicast-basedmutualexclusionalgorithmisinefficientforthiscase.Isit
possibletoimprovetheperformanceofthisalgorithminthiscasebychanging
theprocessstatesforthisparticularscenario?Explain.
A: This algorithm which is multicast based algorithm multicasts requests and
requires reply from all other processes before entering a critical section but when a
single process needs multiple access to the critical section before any other request,
thenitcanbeexpensiveatthecostoftimeinefficiency.
One such solution can be we can change the state from HELD to TEMP rather than
RELEASE when process is done with the critical section. If it needs access again, the
statecanbechangedtoHELDfromTEMPwithoutsendingthemulticastmessage.

2. Whatarethemainguaranteesthatusersexpectconventionalservers(e.g.web
serversorfileservers)tooffer?
A:Themainguaranteesare:
1. Tomaintaintheconsistentstatesoftheobjectsserverstore
2. Continuousavailabilityofservice

3. Itisoftenarguedthatpeer-to-peersystemscanofferanonymityfor
(i)clientsaccessingresources
(ii)thehostsprovidingaccesstoresources.
Discusseachofthesepropositions.Suggestawayinwhichtheresistanceto
attacksonanonymitymightbeimproved.
A: TCP/IP messages contains the IP addresses of the source and the destination
nodes. When an application-level multi-hop routing is used, only the previous and next
nodeintheroutecanbediscoveredwhenpacketsareinterceptedorloggedsomewhere
in the network. GUID doesnt provide information about the location of the node.But if
some hacker/attacker can gain knowledge of the contents of some routing tablesthen
the property iscompromised. Anattackerwithunauthorizedaccessatseveralpointsin
the network could send the probe messages to specific GUIDs and observe the
resultant traffic. This willrevealthelotofinformationaboutthelocationoftheGUID.So
the proposition that clients and resource hosts can remain anonymous is only true for
weak attackers with limited access to the network. This resistance to attacksmightbe
improved by generating several outgoing messages for each incoming request at an
intermediate node, allbutoneofthemessageswouldbetreatedasadummymessage
anddestroyedatthenextnode.Thiswouldincuradditionalcostinnetworktraffic.

4. Isleaderelectionpossibleinasynchronousringinwhichallbutoneprocessor
havethesameidentifier?Justifyyourchoice(whetheryourchoiceistrueor
false).
A: Yes, it is possible, since one of the processors has an id which is different from
every other processors id. We can propose different algorithms to choose the process
withdifferentidastheleader.Weassumenofailureshappen.\
Onepossiblealgorithm:
1. Tostartelection
2. Send(Election,Myid)totheleftandrightprocessorsinthering.
3. Whenreceivingmessage(election,id)frombothleftandrightneighbours.
4. ifmyIDisnotequaltoanyofmytwoneighborsidthen
5. Iamselectedastheleader
6. send(elected,myID)tomyrightneighbor
7. endif
8. (elected,id) is forwarded clockwise in the ring until it comes back to leader (to
informeveryprocessorwhichprocessoristheleader)

You might also like