Distributed Computing,Time and Global States

1. TIME AND GLOBAL STATES
Deepak John
Department Of Computer Applications,SJCET-Pala

2. IntroductionIntroduction
 Time is an Important and interesting issue in distributes systems.
One we can measure accurately. Can use as a metric.y
 Consistency of distributed data, transactions, authenticity checks
(ticket lifetimes),duplication detection, distributed debugging and
b d t ti tgarbage detection, etc.
 A device that count oscillations occurring in a crystal at a definite
frequencyq y
 Hardware time: Hi(t)
 The counts of oscillation since an original point
S ft ti C ( ) H ( )+ Software time: Ci(t) = Hi(t)+
 Timestamp of an event

3. Clock skew and clock drift
 l k kClock skew
Instantaneous difference between readings of any two clocks
Clock drift
Variations in how clocks count time (oscillations in a crystal), which
cause divergence between clocks
Network

4. Clock synchronization using a time server: Cristian’s
method
mr
mt
p Time server,S
Periodically, each machine sends a message to the time server
asking for the current timeg
Machine responds with CUTC as fast as it can

6. The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted. Restart your computer, and then open the file again. If the red x still appears, you may have to delete the image and then insert it again.

7. 1 The master polls the slaves’ clocks
The Berkeley algorithms
1. The master polls the slaves’ clocks
2. The master estimates the slaves’ clocks by round-trip time
 Similar to Christian’s algorithmg
3. The master averages the slaves’ clock values
 Cancel out the individual clock’s tendencies to run fast or
lslow
4. The master sends back to the slaves the amount that the slaves’
clocks should adjust by
 Positive or negative value
 Avoid further uncertainty due to the message transmission
timetime
 Slave adjust its clock

8. NTP: The Network Time Protocol- Mills
 To provide a service enabling clients across the internet to be
synchronized accurately to UTC
 NTP employs techniques for filtering of timing data and it
discriminates between the quality of timing data from differentdiscriminates between the quality of timing data from different
servers.
 To provide a reliable service that can survive lengthy losses of
connectivity:
 There are redundant servers and redundant paths between servers;
server can reconfigure in case of failuresserver can reconfigure in case of failures.
 scale to large number of clients and servers
 To provide protection against interference with the time service To provide protection against interference with the time service
whether accidental or malicious

9.  Hierarchical structure for the time servers
 NTP servers synchronize with each other in one of three modes: in
the order of accuracy.
1. Multicast mode
 Intend for use on a high speed LAN Intend for use on a high speed LAN
 Assuming a small delay
 Low accuracy but efficient
2. Procedure-call mode
 Similar to Christian’s
 Higher accuracy than multicast Higher accuracy than multicast
3. Symmetric mode
 The highest accuracyg y
 Reconfigure when servers become unreachable

10. An example synchronization subnet in an NTP implementation
1
Primary servers are
connected to a time
source,
2 2
Secondary servers
are synchronized
with primary servers.
3 3 3
p y
Note: Arrows denote synchronization control, numbers denote strata.

11. Messages exchanged between a pair of NTP peers
All messages use UDP and so deliver unreliably.
Each message carries three time stamps: sent time, recd time, time of
this message.
Ti-1Ti-2Server B
Time
g
m m'
TiTi- 3Server A
Time

12. Logical Time and logical ClockLogical Time and logical Clock
 Logical time is an alternative; it gives ordering of events, rather than
exact timestamp of occurrence.
 Happen-before relation (→) is a partial order on events that reflects a
flow of information between them.
Th l f ‘H B f ’ l ti Three rules for ‘Happen Before’ relation.
The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted. Restart your computer, and then open the file again. If the red x still appears, you may have to delete the image and then insert it again.

13. Logical clocks (LC)Logical clocks (LC)
 Logical clock is a monotonically increasing software counter that is
updated according to the happened-before relationship between
tevents.
 Each process pi has a logical clock Li, which can be used to apply
logical timestamps to events.logical timestamps to events.
 Rules or Lamports algorithm for logical clocks:The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted. Restart your computer, and then open the file again. If the red x still appears, you may have to delete the image and then insert it again.

14. Events occurring at three processes
The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted. Restart your computer, and then open the file again. If the red x still appears, you may have to delete the image and then insert it again.

15. Lamport timestamps
 e  e`  L(e) < L(e`)
 L(e) < L(e`)  e  e` or e||e` L(e) L(e )  e  e or e||e
The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted. Restart your computer, and then open the file again. If the red x still appears, you may have to delete the image and then insert it again.

16. Vector Clock
 A vector clock for a system of N processes is an array of N integersy p y g
 Each process keeps its own vector clock Vi, which it uses to
timestamp local event
 Processes piggyback vector timestamps on the messages they send Processes piggyback vector timestamps on the messages they send
to one another

17. Global StateGlobal State
 The global state of a distributed computation is the set of local states
of all individual processes involved in the computation plus the state
of the communication channels.
 The set of local states for a global set called the “cut”. A cut
represents the last event that has been recorded for each of severalrepresents the last event that has been recorded for each of several
processes.
 All recorded message receipts have a corresponding recorded send
event

18. A cut C is consistent if for each event it contains all the events that
happened-before(HB) that event.
An inconsistent cut would have a receipt of a message but no
corresponding send event.
A consistent global state is one that corresponds to a consistent cut
p1
e 1
0
e 1
1
e 1
2
e 1
3
A consistent global state is one that corresponds to a consistent cut.
m1 m2
p2
Physical
p2 time
Inconsistent cut
e 2
0
e 2
1
e 2
2
Consistent cut
Inconsistent cut

19.  Distributed garbage collection
Requirements of global states
 Based on reference counting
 Should include the state of communication channels.
 referring object may be in a message in transit referring object may be in a message in transit.
The image cannot be displayed. Your computer may not have enough memory to open the image, or the image may have been corrupted. Restart your computer, and then open the file again. If the red x still appears, you may have to delete the image and then insert it again.

20.  Distributed deadlock detection
 Look for “waits-for”
relationship
 Distributed termination
detection.detection.
 consider a system where each
process is active or passive.
h k b we check processes one-by-
one; has the system
terminated if all are passive?p
 p2 is passive but there is a
message for p1 in transit
Distributed debuggingDistributed debugging
Need collect values of distributed variables at the same time

21. Chandy and Lamport’s Snapshot Algorithm
 Assume the distributed system can be represented as a collection of Assume the distributed system can be represented as a collection of
processes connected to each other through uni-directional point-to-
point communication channels.
 the algorithm may be initiated by any process. Consider two
processes P and Q
 P records its own local state P records its own local state
 It sends a marker along each of its outgoing channels, indicating
that the receiver should participate in recording the global state.
 When process Q receives the marker through an incoming
channel C, its action depends on whether or not it has already
saved its local statesaved its local state

22. Marker receiving rule for process pi
On pi’s receipt of a marker message over channel c:
if ( h t t d d it t t ) itif (pi has not yet recorded its state) it
records its process state now;
records the state of c as the empty set;
turns on recording of messages arriving over other incoming
channels;
else
pi records the state of c as the set of messages it has received over c
since it saved its state.
end ifend if
Marker sending rule for process pi
After pi has recorded its state, for each outgoing channel c:
p sends one marker message over cpi sends one marker message over c
(before it sends any other message over c).