Fundamentals Stream: Lecture 4

1
Fundamentals Stream Lecture 4

• Overview of Lecture
• Group communication
• Introducing group communication
• Focus on reliability and ordering
• Support for availability
• Active and passive replication
• Distributed transactions
• Fundamentals of transactions
• Added complexity of distribution
• Additional reading
• CDK chapter 11, section 11.4, chapter 14,
  sections 14.1-14.2
• CDK chapters 12, 13

2
Introducing Group Communication

• What is group communication?
• Enables the multicasting of a message to a group
  of processes as a single action
• Sender is unaware of the destinations for the
  message
• Why group communication?
• Support for replication/ fault tolerance
• Support the efficient dissemination of data

3
A Typical Group Service
PROCEDURE groupCreate() groupID //
creates a new group and returns the groups
UID PROCEDURE groupJoin(group groupID, port
portID) // adds the named port to the
named group PROCEDURE groupLeave(group groupID,
port portID) // removes the named port
from teh named group PROCEDURE multicast(order
orderType, group groupID, m msg, nRepliesint,
replies ARRAY of msg) // multicasts a
message to the named group with // the
specified delivery semantics, and //
optionally collects a number of replies
Crucial issue of delivery semantics (reliability,
ordering
4
Reliability Guarantees

• Unreliable multicast
• Message sent to all members and may or may not
  arrive
• Reliable multicast
• Reasonable efforts are made to ensure delivery
• Can be based on positive or negative
  acknowledgement schemes
• Atomic multicast
• All members receive message, or none do

5
Implementing Atomic Multicast
1. Originator sends a message to each member of
the group, and awaits acknowledgements 2. If
some acknowledgements are not received in a given
period of time,re-send message repeat this n
times if necessary 3. If all acknowledgements
received then report success
else remove offending members(s) from
group 4. Each member must monitor the
originator On failure -gt elect a
new 'originator' 5. Each member must also
monitor for completion -gt Await
arrival of a next message from the
same originator to indicate previous
multicast completed
6
Ordering Guarantees

• Unreliable multicast
• Messages sent from different processes may be
  delivered in different orders at different sites
• Totally ordered multicast
• Consider messages m1 and m2 sent to the group by
  (potentially) different processes
• Either m1 will be delivered before m2 or vice
  versa for all members of the group
• Causally ordered multicast
• As above, except the ordering of m1 and m2 is
  only important if a happened-before relationship
  exists between the messages

7
Implementing Total Ordering

• The sequencer approach
• Other approaches
• Distributed agreement to generate ids (as in
  Isis)
• Assign timestamps from a (global) logical or
  physical clock

1. All requests must be sent to a sequencer,
where they are given an identifier 2. The
sequencer assings consecutive increasing
identifiers as it receives requests 3. Requested
arriving at sites are held back until they are
next in sequence
8
Total Ordering vs Causal Ordering
9
Example Isis

• What is Isis?
• A framework for reliable distributed computing
  based on process groups
• Offers a range of delivery semantics
• Unordered (FBCAST), causally ordered (CBCAST),
  totally ordered (ABCAST), etc
• Also provides group-view management and state
  transfer on join
• Status
• A successful product
• Ongoing research on Horus, Ensemble, Java Groups

10
Introducing Replication

• Why replicate?
• Performance
• E.g. replication of heavily loaded web servers
• Availability
• availability 1-pn (where n no of replicas, p
  probability of individual failure)
• Requirements for a replication service
• Replication transparency ...
• in spite of network partitions, disconnections,
  etc.

11
Focus on Availability
12
A Generic Replication Architecture
13
Passive Replication

• What is passive replication (primary backup)?
• FEs communicate with a single primary RM, which
  must then communicate with secondary RMs (slaves)
• Requires election of new primary in case of
  primary failure

14
Active Replication

• What is active replication?
• FEs multicast requests to every RM, using an
  appropriate reliable group communications
  protocol, also offering ordering guarantees

15
Alternative Approach 1 Gossip Protocols

• What is a gossip protocol?
• An approach to replication designed to provide
  high availability
• Replication strategy
• Clients send queries and updates to any RM
• Updates are carried out immediately on the chosen
  RM then are propagated in a lazy fashion (using
  gossip messages)
• Queries may be blocked to maintain desired level
  of integrity

16
Alternative Approach 2 The Coda File System

• What is Coda?
• A distributed file system, based on AFS
• Designed esp. for disconnected operation
• Replication strategy
• File system consists of n volume storage groups
• Whole files also cached (and hoarded) locally
• Optimistic replication strategy in that files may
  be modified (in cache) when disconnected
• Conflicts detected for manual repair

17
Introducing Atomic Transactions

• What is a transaction?
• Ensures that a set of objects can move from one
  consistent state to another in spite of
  concurrent access and partial failure
  (concurrency and recovery transparency)
• ACID properties
• Atomicity - all or nothing
• Consistency - moving between consistent states
• Isolation - dealing with concurrency
• Durability - permanence of results

See also FOCS course on databases
18
Approaches to Concurrency Control

• Locking (pessimistic concurrency control)
• Requires use of two-phase locking to ensure
  serial equivalence
• Potential problem of deadlocks
• Optimistic concurrency control
• Assumes that conflicts will not occur
• Check status at commit time, and abort erroneous
  transactions
• Timestamp ordering
• Validate each operation as it is carried out
  depending on timestamp value

19
Recovery

• Undo logs (optimistic approach)
• Make changes to the original data, and maintain
  log of inverse operations (undo log)
• On failure, execute undo log operations to last
  checkpoint
• Intentions logs (pessimistic approach)
• Record required operations in a log
• On commit, ensure log is executed in entirety
• Shadow versions (also pessimistic)
• Make changes to shadow version and make visible
  on commit

20
Distributed Transactions

• What is a distributed transaction?
• A transaction where operations involve multiple
  servers
• Issues
• Need distributed algorithms for concurrency
  control and recovery schemes mentioned above
• Must deal with added difficulty of distributed
  deadlock
• Crucial issue of atomic commit protocols

21
The Two-phase Commit Protocol
22
The Two-phase Commit Protocol (continued)
23
Transactional Middleware TP Monitors

• Standards
• X/Open DTP, OMG Corba OTS
• Key products
• IBMs CICS and Encina (Transarc)
• BEAs Tuxedo
• Microsofts MTS (included in COM)
• SUNs Enterprise JavaBeans (see lecture 4)

24
Expected Learning Outcomes

• You should be able to appreciate the relative
  roles of group communications, replication
  strategies and atomic transactions in supporting
  more robust distributed applications
• You should understand the basic principles
  underpinning group communication, including
  techniques for reliability and ordering
• You should also understand the basic principles
  behind replication strategies including active vs
  passive replication

25
Expected Learning Outcomes (continued)

• You should appreciate the potential role of group
  communication protocols in supporting replication
  management
• You should have an understanding of the impact of
  distribution on atomic transactions and also a
  general appreciation of the various techniques
  employed

26
Exam Question of the Week

• A) Discuss the motivation for introducing
  replication in a distributed system and comment
  on how replication can enhance the three-tier
  approach to distributed systems design.
  4 marks
• B) Describe the service offered by a typical
  group communications protocol, and explain
  carefully the role of such a protocol in
  supporting replication management. Discuss in
  detail the potential semantics offered by a group
  communications protocol with emphasis on the
  options for reliability and ordering.
  10 marks
• C) Briefly outline the replication strategy
  underpinning Coda, and explain how this approach
  can be exploited to support mobile users.
  6 marks