Chapter 5Synchroniza
tion
Presenter:Maria Riaz
2 Distributed Systems – Fall 2004 – Prof. SY Lee
Sequence of Presentation• Synchronization• Clock Synchronization• Logical Clocks• Global State• Election Algorithms• Mutual Exclusion• Distributed Transactions
3 Distributed Systems – Fall 2004 – Prof. SY Lee
Synchronization – Why we need it !
• Stand-alone System– Exclusive access to shared resources
• Distributes System– Exclusive access to shared resources– Ordering of Events
Each node in a distributed system has separate local clock Notion of physical time might differ among various nodes
of the system
4 Distributed Systems – Fall 2004 – Prof. SY Lee
- Clock Synchronization -• Problem due to different time values at
different nodes– When each machine has its own clock, an event that
occurred after another event may nevertheless be assigned an earlier time
5 Distributed Systems – Fall 2004 – Prof. SY Lee
Physical Clocks (1)• Can all clocks in a distributed system be synchronized ?• If we start all clocks in the system with same initial value,
will they remain synchronized for the rest of their operations ?
• Some terminology– Skew: instantaneous difference between readings– Drift: different rates of counting time
• physical variations of underlying oscillators• variance with temperature• even extremely small differences accumulate over a large number of
oscillations– Drift Rate: difference in reading bet. a clock and a nominal
“perfect clock” per unit of time measured by the reference clock• 10-6 seconds/sec for quartz crystals• 10-7 - 10-8 seconds/sec for high precision quartz crystals
• Problem – How do we synchronize them with real-world clocks– How do we synchronize the clocks with each other
6 Distributed Systems – Fall 2004 – Prof. SY Lee
Physical Clocks (2)• Some methods to measure time
– Mean solar second: measuring a large numbers of day -- taking average -- dividing by 86400
– TAI (International Atomic Time): the mean number of ticks of the cesium 133 clocks (since 1/1/1958) divided by 9,192,631,770
• Very small drift rate ~ 10-13 seconds/second– UTC: broadcast by NIST from Fort Collins,
Colorado over shortwave radio station WWV.
7 Distributed Systems – Fall 2004 – Prof. SY Lee
Clock Synchronization Algorithms (1)
• The relation between clock time and UTC when clocks tick at different rates– maximum drift rate ()– every t seconds, the worst case drift between two clocks will be at most 2t– to guarantee two clocks never differ by more than , the clocks must re-
synchronize every /2 seconds
8 Distributed Systems – Fall 2004 – Prof. SY Lee
Clock Synchronization Algorithms (2)
• Centralized Algorithms– Cristian’s Algorithm (1989)– Berkeley Algorithm (1989)
• Decentralized Algorithms– Averaging Algorithms (e.g. NTP)– Multiple External Time Sources
9 Distributed Systems – Fall 2004 – Prof. SY Lee
Cristian’s Algorithm• Assume one machine (the time server) has a
WWV receiver and all other machines are to stay synchronized with it.– Every /2 seconds, each machine sends a message
to the time server asking for the current time.– Time server responds with message containing
current time, CUTC.• Problem
– time must never run backward
10 Distributed Systems – Fall 2004 – Prof. SY Lee
Berkeley Algorithm
a) The time daemon asks all the other machines for their clock values.b) The machines answer and the time daemon computes the average.c) The time daemon tells everyone how to adjust their clock.
11 Distributed Systems – Fall 2004 – Prof. SY Lee
Averaging Algorithms• At the beginning of each interval, every machine
broadcasts the current time according to its clock• Then it starts a local timer to collect all other broadcasts
that arrive during some interval S
• The simplest algorithm is just to average the values from all other machines– A slightly more sophisticated algorithm :: Discard the m
highest and m lowest to reduce the effect of a set of faulty clocks
– Another improved algorithm :: Correct each message by adding to the received time an estimate of the propagation time from the ith source
• extra probe messages are needed to use this scheme
• One of the most widely used algorithms in the Internet is the Network Time Protocol (NTP)
12 Distributed Systems – Fall 2004 – Prof. SY Lee
- Logical Clocks -• Mostly ‘absolute’ time is not important rather ‘relative’ time
is of significance– Internal consistency and ordering of events– If two process don’t interact no need for synchronization
between them• A logical clock is a
– Monotonically increasing SW counters (COULOURIS)– Clocks on different computers that are somehow consistent
(LAMPORT)• Potential Requirements for logical clocks:
– Timestamps : C(a), C(b)– If a happens before b in the same process, C(a) < C(b).
• a b => C(a) < C(b)– If a and b represent the sending and receiving of a message,
respectively, C(a) < C(b).– For all distinctive events a and b, C(a) ≠ C(b).
• Two methods for assigning logical timestamps– Lamport’s Timestamps– Vector Timestamps
13 Distributed Systems – Fall 2004 – Prof. SY Lee
Lamport’s Timestamps (1)• Lamport defined a relation ”happens before”. a b ‘a
happens before b’ (1978)• Each Process has local clock LCi
– with each local event e: LCi := LCi + 1; e– with each sending of a message by process Pi: LCi := LCi
+1; send (LCi,m)– with each reception of a message “(M,LCm)” by Pj: LCj :=
MAX(LCm, LCj ); LCj := LCj +1
14 Distributed Systems – Fall 2004 – Prof. SY Lee
Lmaport’s Timestamps (2)• Can be used to implement ‘totally
ordered multicast’– A multicast operation by which all
messages are delivered in the same order to each receiver
15 Distributed Systems – Fall 2004 – Prof. SY Lee
Vector Timestamps• Each process Pi has its own vector clock Ci
– Ci : n-dimensional vector (n: number of processes)• Notation: Ci[j] : the timestamp of the last event in Pj
by which Pi has potentially been effected
Initially:all c[i] = 0
Increment C[i]:-Events-Send msg-Receive msg
16 Distributed Systems – Fall 2004 – Prof. SY Lee
- Global State -• Like a ‘distributed snapshot’ reflecting a state
in which the system might have been– represents the last event recorded for each process
• Graphically represented by a cut– Consistent : for every received message, the sender
can be identified• Cause Effect
17 Distributed Systems – Fall 2004 – Prof. SY Lee
Global State (2)
Organization of a process and channels for a distributed snapshot
• Process Q receives a marker (start) for the first time and records its local state
• Q records all incoming message• Q receives a marker (end) for its incoming
channel and finishes recording the state of the incoming channel
• final recorded state
18 Distributed Systems – Fall 2004 – Prof. SY Lee
- Election Algorithms -• Election algorithms
– algorithms for electing a coordinator (using this as a generic name for the special process)
– attempt to locate the process with the highest process number and designate it as coordinator
• Bully Algorithm• Ring Algorithm
• Goal– ensure that when an election starts, it
concludes with all processes agreeing on who the new coordinator is to be
19 Distributed Systems – Fall 2004 – Prof. SY Lee
Bully Algorithm
• A process P detects failure of coordinator and holds an election to be the coordinator– All process with ID > P response– If P receives such a response, it will step back– Processes having ID > P can hold elections
and repeat same procedure– If no response from any process with higher ID,
election holder becomes the new coordinator• Example
a) Process 4 holds electionb) Process 5 and 6 respond, telling 4 to stopc) Now 5 and 6 each hold an electiond) Process 6 tells 5 to stope) Process 6 wins and tells everyone
20 Distributed Systems – Fall 2004 – Prof. SY Lee
Ring Algorithm• Process detects failure of coordinator• Send message to neighbor with its ID• Neighbor adds its ID and pass along• When all process have added their ID, the one with highest ID
becomes the coordinator• The message is rotated once again so everyone knows
21 Distributed Systems – Fall 2004 – Prof. SY Lee
- Mutual Exclusion -• To control access to a critical
section
– Centralized Algorithm– Distributed Algorithm– Token Ring Algorithm
22 Distributed Systems – Fall 2004 – Prof. SY Lee
Centralized Algorithm
a) Process 1 asks the coordinator for permission to enter a critical region. Permission is granted
b) Process 2 then asks permission to enter the same critical region. The coordinator does not reply
c) When process 1 exits the critical region, it tells the coordinator, when then replies to 2
23 Distributed Systems – Fall 2004 – Prof. SY Lee
Distributed Algorithm
a) Two processes want to enter the same critical region at the same moment
b) Process 0 has the lowest timestamp, so it winsc) When process 0 is done, it sends an OK also, so 2 can
now enter the critical region
24 Distributed Systems – Fall 2004 – Prof. SY Lee
Token Ring Algorithma) unordered group of processes on a networkb) logical ring constructed in software
A token is passed along the ring to allow access to the critical section
25 Distributed Systems – Fall 2004 – Prof. SY Lee
Comparison
Algorithm Messages per entry/exit
Delay before entry (in message times) Problems
Centralized 3 2 Coordinator crash
Distributed 2 ( n – 1 ) 2 ( n – 1 )Crash of any processGroup communication
Token ring 1 to 0 to n – 1 Lost token, process crash
A comparison of three mutual exclusion algorithms
26 Distributed Systems – Fall 2004 – Prof. SY Lee
- Distributed Transactions -• Basic (Flat) Transactions &
Limitations• Alternatives
– Distributed Transactions– Nested Transactions
• Problems– Concurrency Control– Synchronization
27 Distributed Systems – Fall 2004 – Prof. SY Lee
The Transaction Model (1)
Updating a master tape is fault tolerant
28 Distributed Systems – Fall 2004 – Prof. SY Lee
The Transaction Model (2)
Examples of primitives for transactions
Primitive Description
BEGIN_TRANSACTION Make the start of a transaction
END_TRANSACTION Terminate the transaction and try to commit
ABORT_TRANSACTION Kill the transaction and restore the old values
READ Read data from a file, a table, or otherwise
WRITE Write data to a file, a table, or otherwise
29 Distributed Systems – Fall 2004 – Prof. SY Lee
ACID - Four Characteristics• Atomic: to the outside world, the transaction
happens indivisibly
• Consistent: the transaction does not violate system invariants
• Isolated: concurrent transactions do not interfere with each other
• Durable: once a transaction commits, the changes are permanent
30 Distributed Systems – Fall 2004 – Prof. SY Lee
Limitations of Flat TransactionsMain limitation: do not allow partial
results to be committed or aborted
updating all of the hyperlinks to a webpage W, which moved to a new location
BEGIN_TRANSACTION reserve WP -> JFK; reserve JFK -> Nairobi; reserve Nairobi -> Malindi;END_TRANSACTION (a)
BEGIN_TRANSACTION reserve WP -> JFK; reserve JFK -> Nairobi; reserve Nairobi -> Malindi full => ABORT_TRANSACTION (b)
31 Distributed Systems – Fall 2004 – Prof. SY Lee
Classification of Transactions
a) A nested transactionb) A distributed transaction
32 Distributed Systems – Fall 2004 – Prof. SY Lee
Private Workspace
a) The file index and disk blocks for a three-block file
b) The situation after a transaction has modified block 0 and appended block 3
b) After committing
Make a copy of the original workspace and perform all operation in the copied ‘private’ space before committing
read only no need for private copy
33 Distributed Systems – Fall 2004 – Prof. SY Lee
Write-Ahead Log
a) A transactionb) – d) The log before each statement is executed
x = 0;y = 0;BEGIN_TRANSACTION; x = x + 1; y = y + 2 x = y * y;END_TRANSACTION; (a)
Log
[x = 0 / 1]
(b)
Log
[x = 0 / 1][y = 0/2]
(c)
Log
[x = 0 / 1][y = 0/2][x = 1/4]
(d)
34 Distributed Systems – Fall 2004 – Prof. SY Lee
Concurrency Control (1)
General organization of managers for handling transactions
35 Distributed Systems – Fall 2004 – Prof. SY Lee
Concurrency Control (2) General organization of managers for
handling distributed transactions
36 Distributed Systems – Fall 2004 – Prof. SY Lee
Synchronization
• Two operations are serializable if the order of operations does not change the outcome i.e., the operations do not conflict– properly schedule conflicting operations (two
read operations never conflict)
• Mechanism for synchronization– Mutual Exclusion mechanisms on shared data
(i.e locking)• Two-Phase Locking• Strict Two-Phase Locking
– Explicitly ordering operations using timestamps • Pessimistic Timestamp Ordering
37 Distributed Systems – Fall 2004 – Prof. SY Lee
Two-phase locking
• A transaction T is granted a lock if there is no conflict• The scheduler will never release a lock for data item x, until the data
manager acknowledges it has performed the operation for which the lock was set
• Once the scheduler has released a lock on behalf of a transaction T, it will never grant another lock on behalf of T
38 Distributed Systems – Fall 2004 – Prof. SY Lee
Strict two-phase locking
• In centralized 2PL: a single site is responsible for granting and releasing locks
• In primary 2PL: each data item is assigned a primary copy• In distributed 2PL: the schedulers on each machine not only take
care that locks are granted and released, but also that the operation is forwarded to the (local) data manager
39 Distributed Systems – Fall 2004 – Prof. SY Lee
Pessimistic Timestamp Ordering
Concurrency control using Timestamps
40 Distributed Systems – Fall 2004 – Prof. SY Lee
Thank you !
Questions ?