1. Distributed Computing
Umar Kalim
Dept. of Communication Systems Engineering
19/12/2007
Fall 2007
cs425

2. Agenda Synchronization Algorithms Physical Clocks Christian Berkeley
Averaging
NTP (Network Time Protocol)
Fall 2007
cs425
Ref: COMP 212, University of Liverpool

3. What can go wrong?
Updating a replicated database: Update 1 adds 100 rupees to an account, Update 2 calculates and adds 1% interest to the same account.
Clock reading might be different for different computers!
Inconsistent state
Fall 2007
cs425

4. What to do? Include the time of every update in the update message
Clock synchronization
Order events in the network
Logical clock
Transactions
Distributed Mutual Exclusion
Distributed Transactions
Fall 2007
cs425

5. Computer Clock We need to measure time accurately:
to know the time an event occurred at a computer
Algorithms for clock synchronization useful for
concurrency control based on timestamp ordering
authenticity of requests e.g. in Kerberos etc
Each computer in a DS has its own internal clock
Even if clocks on all computers in a DS are set to the same time, their clocks will eventually vary quite significantly unless corrections are applied
Fall 2007
cs425

6. Synchronization based on “Actual Time”
Note: time is really easy on a uniprocessor system.
Achieving agreement on time in a DS is not trivial.
Question: is it even possible to synchronize all the clocks in a Distributed System?
With multiple computers, “clock skew” ensures that no two machines have the same value for the “current time”. But, how do we measure time?
Fall 2007
cs425

7. How do we measure time?
Turns out that we have only been measuring time accurately with a “global” atomic clock since Jan. 1st, 1958 (the “beginning of time”).
Refer to pages of the textbook for all the details – it’s quite a story.
Bottom Line: measuring time is not as easy as one might think it should be.
Fall 2007
cs425

8. Universal Coordinated Time (UTC)
International Atomic Time is based on very accurate physical clocks (drift rate 10-13)
UTC is an international standard for time keeping
It is based on atomic time, but occasionally adjusted to astronomical time
It is broadcast from radio stations on land and satellite (e.g. GPS)
Computers with receivers can synchronize their clocks with these timing signals
Signals from land-based stations are accurate to about millisecond
Signals from GPS are accurate to about 1 microsecond
Fall 2007
cs425

9. Clock Synchronization
There exists a time server receiving signals from a UTC source
Cristian’s algorithm
There is no UTC source available
Berkley’s algorithm
Exact time does not matter!
Lamport’s algorithm
Fall 2007
cs425

10. Clock Synchronization Algorithms
Each machine has a timer that causes an interrupt every H seconds.
Results in an increment of the software clock that keeps track of the number of ticks since some agreed time.
Let the value of the clock be C.
When the UTC time is t, the value of the clock in this machine p is Cp(t).
Ideal circumstances Cp(t) = t for all p and all t
Fall 2007
cs425

11. Clock Synchronization Algorithms
Fall 2007
cs425

12. Clock Synchronization Algorithms
Problem: Real timers do not cause an interrupt exactly H times a second
For example
For a timer H=60 there should be 216,000 ticks per hour
In practice, chips have an error of 10-5
Meaning that the range will be 215,998 to 216,002 per hour
Thus, if a constant ρ exists such that 1-ρ ≤ dC/dt ≤ 1+ρ
ρ is known as the maximum drift rate
Fall 2007
cs425

13. Clock Synchronization
Thus at a time Δt two clocks drifting apart from UTC may be as much as 2ρΔt
To guarantee that clocks do not drift apart more than δ, they must be synchronized every δ/2ρ
Fall 2007
cs425

14. Clock Sync. Algorithm: Cristian's
Every computer periodically asks the “time server” for the current time
The server responds ASAP with the current time CUTC
The client sets its clock to CUTC
Fall 2007
cs425

15. Problems
Major problem: if time from time server is less than the client - resulting in time running backwards on the client! (Which cannot happen – time does not go backwards).
Introduce changes gradually
Minor problem: results from the delay introduced by the network request/response: latency
Best estimate (T1-T0)/2
If the interrupt handling time, I, is known, (T1-T0-I)/2
Use series of measurements
Fall 2007
cs425

16. Berkeley Algorithm
An algorithm for internal synchronization of a group of computers
A master polls to collect clock values from the others (slaves)
The master uses round trip times to estimate the slaves’ clock values
It takes an average
It sends the required adjustment to the slaves (better than sending the time which depends on the round trip time)
If master fails, can elect a new master to take over
Fall 2007
cs425

17. The Berkeley Algorithm
The time daemon asks all the other machines for their clock values
The machines answer
The time daemon tells everyone how to adjust their clock
Fall 2007
cs425

18. Berkeley Algorithm
Fall 2007
cs425

19. Berkeley Algorithm
Average time: 3:05
Fall 2007
cs425

20. Other Algorithms
Both Cristian’s and the Berkeley Algorithm are centralised algorithms.
Decentralised algorithms also exist, and the Internet’s Network Time Protocol (NTP) is the best known and most widely implemented.
NTP can synchronize clocks to within a 1—50 msec accuracy.
Fall 2007
cs425

21. Averaging Algorithms
Every R seconds, each machine broadcasts its current time.
The local machine collects all other broadcast time samples during some time interval, S.
The simple algorithm: the new local time is set as the average of the value received from all other machines.
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).
Fall 2007
cs425

22. Network Time Protocol
Uses a network of time servers to synchronize all processors on a network.
Time servers are connected by a synchronization subnet tree. The root is adjusted directly . Each node synchronizes its children nodes.
Fall 2007
cs425

23. NTP Contact a time server to ask the time.
What is the major issue here?
Trick is to get a good estimate of the message delays.
Fall 2007
cs425

24. NTP NTP is about making clocks correspond to the real time.
Clock servers are divided into stratums, starting with stratum-0 for the reference physical clock.
If a stratum-2 clock is running faster than a stratum-1 clock, the stratum-2 clock will adjust.
How should it adjust if it is running fast?
Time should never go backwards, but should just slow down.
Fall 2007
cs425

25. Summary & Questions?
That’s all for today!
Fall 2007
cs425