Distributed Systems Id2201

Front 1

What aspect makes distrubuted systems most different from non-distributed systems?

Back 1

Harder to handle failures.

Front 2

What is meant by access transparency?

Back 2

Local and remote resources are accessed using the same operations.

Front 3

What is meant by location transparency?

Back 3

Remote resources are accessed using location independent names.

Front 4

What is meant by concurrency transparency?

Back 4

Processes can access resources without interfering with each other.

Front 5

What is meant by failure transparency?

Back 5

Failures are concealed for the users of a resource.

Front 6

What is meant by replication transparency?

Back 6

Users of a resource access it as if it was not replicated.

Front 7

What is significant for a client-server architecture?

Back 7

The client is the active part.

Front 8

What would we call a system where one node is always reacting on requests and other nodes only communicate with this node?

Back 8

Client-server system

Front 9

What is significant for a peer to peer architecture?

Back 9

All nodes are active and can be initiator of operations.

Front 10

What do we know in a synchronous system?

Back 10

That all messages will be delivered.

Front 11

What do we call a system where the upper bound of the time for operations and message delivery are known?

Back 11

A synchronous system

Front 12

What do we call a system where the maximum times for operations and message delivery are not known?

Back 12

An asynchronous system.

Front 13

Why is it impossible to implement a perfect failure detector in an asynchronous system?

Back 13

Maximum delay for a reply is not known.

Front 14

How can we implement a perfect failure detector in an asynchronous system?

Back 14

It is not possible

Front 15

What is provided by UDP?

Back 15

A best effort delivery of messages to a process

Front 16

What does UDP give us that is not provided by IP?

Back 16

Port addressing

Front 17

What is provided by TCP

Back 17

A full-duplex stream between two processes

Front 18

What is not provided by TCP?

Back 18

A guaranteed delivery of messages.

Front 19

When is the TCP window size a limiting factor for high capacity communication

Back 19

Over a long fat communication link.

Front 20

Which address can be found in the TCP header

Back 20

The port number.

Front 21

What is a good reason for choosing UDP rather than TCP

Back 21

You have small messages that should be sent with little delay

Front 22

What is a good reason for choosing TCP rather than UDP?

Back 22

You need to know that a message is handled by the remote application?

You have large messages or a sequence of messages?

Front 23

One can determine how large the receiver window should be, in for example a TCP connection, in order to maximize the capacity of a link. How is this calculated?

Back 23

capacity of link times round trip latency of the link

Front 24

The limiting factor for capacity in a TCP connection, T bps, where the link capacity is C bps, the round trip time R sec and the receiver window W bits is:

Back 24

the lesser of C and W/R

Front 25

When estimating the maximum capacity of a TCP connection you need to know:

Back 25

the minimum link capacity and the round trip

Front 26

What is significant for synchronous communication?

Back 26

the send operation blocks and waits for the receive operation

Front 27

What is significant for asynchronous communication?

Back 27

a sender can continue without waiting for the receive operation

Front 28

What is true for asynchronous communication?

Back 28

sender does not block waiting for receiver

Front 29

Can a synchronous communication interface be used in an asynchronous way?

Back 29

yes, by performing the send operation in a separate thread

Front 30

The Erlang call gen tcp:recv(Socket, 0) will return

Back 30

the complete message as sent to the Socket

Front 31

A server using a stream socket API can communicate with several clients and separate the streams by:

Back 31

creating a new stream socket for each connecting client

Front 32

What is the purpose of the marshaling procedure?

Back 32

to encode application layer structures in an external form

Front 33

What is a gossip protocol?

Back 33

A protocol where peers randomly exchange messages to achieve, for example, unreliable multicast.

Front 34

What is meant by time uncoupling in a communication framework?

Back 34

Sender and receiver need not be active at the same time.

Front 35

What is meant by space uncoupling in a communication framework?

Back 35

A sender does not need to know the name or identifier of the receiver.

Front 36

What is meant by causal ordering in a communication framework?

Back 36

If a client is delivered two messages m1 and m2, then m1 could not have been sent by someone after having being delivered m2.

Front 37

What is meant by an idempotent operation?

Back 37

An operation that can be performed several times with the same effect as being performed once.

Front 38

How are arguments passed in Java RMI?

Back 38

remote objects as reference, all other as copies

Front 39

Which invocation semantic is provided by Java RMI?

Back 39

at most once

Front 40

If a RPC call with at least once semantics fails, we know that:

Back 40

the call might have been executed at least once

Front 41

What is given by at least once RPC semantics?

Back 41

if no failure is reported the call has been invoked at least once

Front 42

How do Erlang processes communicate?

Back 42

asynchronous message passing

Front 43

Does Erlang provide a form of location transparency?

Back 43

Yes - a process can use a process identifier without having to know the address of the node where the process lives.

Front 44

How is the destination defined in an Erlang send operation?

Back 44

B as a process identifier or a local or remote registered name

Front 45

What will prevent an Erlang process from sending a message to a process on another Erlang node?

Back 45

A the two nodes do not have the same secret cookie

Front 46

Message sending in Erlang provides the following semantics:

Back 46

best effort FIFO delivery of message

Front 47

Can we have a circular construction in Erlang?

Back 47

yes, processes can refer to each other in a circular way

Front 48

What happens with Erlang processes that are suspended on a receive statement but no longer referenced by any process?

Back 48

they will remain in the system until explicitly killed

Front 49

Can we implement a RPC system with exactly-once semantics in an asynchronous system with non-failing nodes but unreliable networks?

Back 49

yes, if the client keeps re-sending a uniquely tagged request until a reply is received and a server keeps track of all handled request in order not to duplicate a request

Front 50

What is the difference when an error is reported for an at-least once and at-most-once remote procedure call?

Back 50

in the at-least-once case, the remote call will not have been executed at all

Front 51

What is a soft link in a file system?

Back 51

a path that is resolved to another path

Front 52

What is a hard link in a file system?

Back 52

A a mapping of a name to a file identifier

Front 53

Can two Unix processes simultaneous write to different positions in a single file?

Back 53

Yes, the two processes will have their own file table entries

Front 54

What is the purpose of the Unix lseek operation?

Back 54

set the read/write pointer of an opened file

Front 55

How is a NFS client-side cache entry validated?

Back 55

if the validity was checked less than t seconds ago or if the server modification time is equal to the client modification time

Front 56

How is authentication control handled in Sun NFS?

Back 56

authentication is provided by RPC in each operation

Front 57

How does a NFS server know at what position to read and write to?

Back 57

each read and write operation holds the position

Front 58

How is AFS client side caching implemented?

Back 58

the server promise to notify the client if a file is modified by another client

Front 59

What is the advantage of using AFS over NFS?

Back 59

the client need not periodically check validity

Front 60

How are inconsistencies of the resolver cache handled in the DNS architecture?

Back 60

A each entry has a time-to-live

Front 61

What is the advantage of using recursive navigation for DNS queries?

Back 61

server does not need to hold a request state

Front 62

How are inconsistent cached entries removed from a DNS resolver?

Back 62

all entries have an expiration time set when cached

Front 63

What is the advantage of a flat name space?

Back 63

it is a simple solution

Front 64

How can we make two computer clocks perfectly synchronized?

Back 64

we can not

Front 65

What accuracy can be provided using Christian’s algorithm?

Back 65

±(Tround ÷ 2 − minimum latency)

Front 66

What is the purpose of the Berkeley algorithm?

Back 66

to perform internal synchronization

Front 67

What does the reply from a NTP server contain?

Back 67

send and receive time of request and send time of reply

Front 68

When is it better to use the Berkeley algorithm rather than NTP for clock synchronization?

Back 68

when implementing internal synchronization

Front 69

At time 117 you receive a NTP reply with the following information: request sent at 82, received at 111, reply sent at 120. How should you adjust your time?

Back 69

advance 16 steps

Front 70

What is true if A happened before B?

Back 70

A must have occurred in real-time before B

Front 71

What is true for events A and B?

Back 71

if A caused B then A happened before B

Front 72

What can we know if we use Lamport clocks?

Back 72

if L(a) < L(b) then a could have caused b

Front 73

What can we know if we use Lamport clocks?

Back 73

if, and only if, L(a) < L(b) then a happened before b ???

Front 74

Which is the most natural representation of a vector clock?

Back 74

a record with one element per process

Front 75

What is the difference between a Lamport clock and a vector clock?

Back 75

only the vector clock gives a complete description of the ”happened before order”

Front 76

When is it problematic to use vectors clocks?

Back 76

when we have a dynamic set of processes

Front 77

An alternative way of implementing a vector clock would be to keep a set of the highest counters seen from each process (including own), send it along with any message, update own counter and merging the own set and received set when a message is received. This would have the following advantage:

Back 77

new processes can easily be added

Front 78

If events in a given set are to be ordered in a total order that respect the happened before order what is the advantage of using vector clocks?

Back 78

only vector clocks will give us the happened before order

Front 79

Is it possible to produce a total order of a set of events stamped with a Lamport time that does not violate any causal relationship?

Back 79

no, causal relationships will need real-time clock time stamps

Front 80

What is the definition of a consistent cut?

Back 80

if e is in the cut and f happened-before e then f is in the cut

Front 81

What is the definition of a stable global state predicate?

Back 81

if a system enters a state where the predicate holds true it will remain true in all future states

Front 82

What is the benefit of a reliable multicast?

Back 82

messages are guaranteed to be delivered to all correct processes

Front 83

Can we implement a reliable multicast using only basic multicast?

Back 83

yes, by re-sending each received message to all other nodes

Front 84

We sometimes want to include the behavior of faulty nodes in our requirements and talk about uniform agreement for multicast. What do we mean by uniform agreement?

Back 84

If any node, including faulty, deliver a message then all correct nodes deliver the message.

Front 85

1p. What is the difference between uniform agreement and nonuniform agreement for multicast?

Back 85

Uniform agreement includes the behaviour of non-correct node.

Front 86

What is the definition of total order multicast?

Back 86

messages are delivered in the same sequence

Front 87

What would you call a multicast service that would never deliver a message m2 before another message m1, if m2 was sent as a response to m1?

Back 87

total order

Front 88

What does Lamport clocks give us when implementing distributed mutual-exclusion?

Back 88

request are granted in happened before order

Front 89

What are the requirements for using the Bully algorithm?

Back 89

we must have reliable failure detectors

Front 90

What is two-phase locking in a transaction?

Back 90

not taking any locks once a lock has been released

Front 91

What does it mean that a transaction meets the isolation property?

Back 91

intermediate results must not be visible to other transactions

Front 92

In a distributed transaction one often use two-phase commit, why is this better than one-phase commit?

Back 92

We ensure that if one transaction commits then all transactions will commit.

Front 93

In a view-synchronous group membership protocol, a process that enters the group, and is included in the delivered view, will be guaranteed to be delivered:

Back 93

all messages starting from the view where it is included in the group

Front 94

What is sent by the primary replica manager to the backup replica manager in a passive replicated system?

Back 94

A unique identifier, the state change and the response.

Front 95

A routing strategy, in a distributed hash table of n nodes, would typically have an asymptotic complexity of:

Back 95

O(log(n))

Front 96

What is the purpose of hashing in a DHT?

Back 96

to generate uniformly distributed keys