Architectures

1. Architectures

2. Architectural Styles (1) • Considering the logical organization of distributed systems into software components, also referred to as software architecture • Important styles of architecture for distributed systems • Layered architectures • Object-based architectures • Data-centered architectures • Event-based architectures

3. Architectural Styles (2) Figure 2-1. The (a) layered architectural style The basic idea for the layered style is simple: components are organized in a layered fashion where a component at layer L; is allowed to call components at the underlying layer Li but not the other way around. An key observation is that control generally flows from layer to layer: requests go down the hierarchy whereas the results flow upward.

4. Architectural Styles (3) Figure 2-1. (b) The object-based architectural style. Each object corresponds to what we have defined as a component, and these components are connected through a (remote) procedure call mechanism. This software architecture matches the client-server system architecture. The layered and object-based architectures still form the most important styles for large software systems

5. Reminder • IPC • Pipes, TCP/IP, shared storage • RPC • CORBA • RMI • Web Services • REST

6. Architectural Styles (4) • Data-centered architectures evolve around the idea that processes communicate through a common (passive or active) repository. • It can be argued that for distributed systems these architectures are as important as the layered and object-based architectures. • For example, a wealth of networked applications have been developed that rely on a shared distributed file system in which virtually all communication takes place through files.

7. Architectural Styles (5) In event-based architectures, processes essentially communicate through the propagation of events, which optionally also carry data. For distributed systems, event propagation has generally been associated with what are known as publish/subscribe systems. The basic idea is that processes publish events after which the middleware ensures that only those processes that subscribed to those events will receive them. The main advantage of event-based systems is that processes are loosely coupled. In principle, they need not explicitly refer to each other. Figure 2-2. (a) The event-based architectural style

8. Architectural Styles (6) Event-based architectures can be combined with data-centered architectures, yielding what is also known as shared data spaces. The essence of shared data spaces is that processes are now also decoupled in time: they need not both be active when communication takes place. Furthermore, many shared data spaces use a SQL-like interface to the shared repository in that sense that data can be accessed using a description rather than an explicit reference, as is the case with files. Figure 2-2. (b) The shared data-space architectural style.

9. ESB • An Enterprise Service Bus, ESB, is an application that gives access to other applications and services. Its main task is to be the messaging and integration backbone of an enterprise.

10. System Architectures • How many distributed systems are actually organized by considering where software components are placed. • Deciding on software components, their interaction, and their placement leads to an instance of a software architecture, also called a system architecture. • Two Types of architecture: • Centralized architecture • Decentralized architecture

11. Centralized Architectures (1) • In the basic client-server model, processes in a distributed system are divided into two (possibly overlapping) groups. • A server is a process implementing a specific service, for example, a file system service or a database service. • A client is a process that requests a service from a server by sending it a request and subsequently waiting for the server's reply. This client-server interaction, also known as request-reply behavior 11

12. Centralized Architectures (2) Figure 2-3. General interaction between a client and a server. 12

13. Centralized Architectures (3) • Communication between a client and a server can be implemented by means of a simple connectionless protocol when the underlying network is fairly reliable as in many local-area networks • When a client requests a service, it simply packages a message for the server, identifying the service it wants, along with the necessary input data. The message is then sent to the server. • Server will always wait for an incoming request, subsequently process it, and package the results in a reply message that is then sent to the client. 13

14. Centralized Architectures (4) • Advantage of connectionless protocol: • efficient • As long as messages do not get lost or corrupted, the request/reply protocol works fine. • Making the protocol resistant to occasional transmission failures is not trivial. • Solution: the client must resend the request when no reply message comes in. • Problem: the client cannot detect whether the original request message was lost, or that transmission of the reply failed. 14

15. Centralized Architectures (5) • If the reply was lost, then resending a request may result in performing the operation twice. • Examples: • If the operation was "transfer $10,000 from my bank account," then, it would be better to report an error instead. • If the operation was "tell me how much money I have left," then, it would be perfectly acceptable to resend the request. 15

16. Centralized Architectures (6) • As an alternative, many client-server systems use a reliable connection-oriented protocol. • Although this solution is not appropriate in a local-area network due to relatively low performance, it works perfectly in wide-area systems in which communication is inherently unreliable. 16

17. Centralized Architectures (7) • Example: Virtually all Internet application protocols are based on reliable TCP/IP connections. In this case, whenever a client requests a service, it first sets up a connection to the server before sending the request. The server generally uses that same connection to send the reply message, after which the connection is torn down. Trouble: setting up and tearing down a connection is relatively costly, especially when the request and reply messages are small. 17

18. Application Layering (1) Recall previously mentioned layers of architectural style • The user-interface level • The processing level • The data level

19. Application Layering (2) • Figure 2-4. The simplified organization of an Internet search engine into three different layers.

20. Multitiered Architectures (1) • The simplest organization is to have only two types of machines: • A client machine containing only the programs implementing (part of) the user-interface level • A server machine containing the rest, • the programs implementing the processing and data level

21. Multitiered Architectures (2) • Figure 2-5. Alternative client-server organizations (a)–(e).

22. Multitiered Architectures (3) • Figure 2-6. An example of a server acting as client.

23. Structured Peer-to-Peer Architectures (1) • Figure 2-7. The mapping of data items onto nodes in Chord.

24. Chord protocol • Consistent hashing function assigns each node and key an m-bit identifier using SHA-1 base hash function. • Node’s IP address is hashed. • Identifiers are ordered on a identifier circle modulo 2m called a chord ring. • succesor(k) = first node whose identifier is >= identifier of k in identifier space.

25. Chord Protocol • Assumes communication in underlying network is both symmetric and transitive. • Assigns keys to nodes with consistent hashing • Hash function balances the load • When Nth node joins or leaves only O(1/N) fraction of keys moved.

26. Chord protocol m = 6 10 nodes

27. Theorem • For any set of N nodes and K keys, with high probability: • Each node is responsible for at most (1+e)K/N keys. • When an (N+1)st node joins or leaves the network, responsibility for O(K/N) keys changes hands. e = O(log N)

28. Simple Key Location Scheme N1 N8 K45 N48 N14 N42 N38 N21 N32

29. Scalable Lookup Scheme N1 Finger Table for N8 N56 N8 N51 finger 6 finger 1,2,3 N48 N14 finger 5 N42 finger 4 N38 N21 N32 finger [k] = first node that succeeds (n+2k-1)mod2m

30. Scalable Lookup Scheme // ask node n to find the successor of id n.find_successor(id) if (id belongs to (n, successor]) return successor; else n0 = closest preceding node(id); return n0.find_successor(id); // search the local table for the highest predecessor of id n.closest_preceding_node(id) for i = m downto 1 if (finger[i] belongs to (n, id)) return finger[i]; return n;

31. Lookup Using Finger Table N1 lookup(54) N56 N8 N51 N48 N14 N42 N38 N21 N32

32. Scalable Lookup Scheme • Each node forwards query at least halfway along distance remaining to the target • Theorem: With high probability, the number of nodes that must be contacted to find a successor in a N-node network is O(log N)

33. Dynamic Operations and Failures Need to deal with: • Node Joins and Stabilization • Impact of Node Joins on Lookups • Failure and Replication • Voluntary Node Departures

34. Node Joins and Stabilization • Node’s successor pointer should be up to date • For correctly executing lookups • Each node periodically runs a “Stabilization” Protocol • Updates finger tables and successor pointers

35. Node Joins and Stabilization • Contains 6 functions: • create() • join() • stabilize() • notify() • fix_fingers() • check_predecessor()

36. Create() • Creates a new Chord ring n.create() predecessor = nil; successor = n;

37. Join() • Asks m to find the immediate successorof n. • Doesn’t make rest of the networkaware of n. n.join(m) predecessor = nil; successor = m.find_successor(n);

38. Stabilize() • Called periodically to learn about new nodes • Asks n’s immediate successor about successor’s predecessor p • Checks whether p should be n’s successor instead • Also notifies n’s successor about n’s existence, so that successor may change its predecessor to n, if necessary n.stabilize() x = successor.predecessor; if (x  (n, successor)) successor = x; successor.notify(n);

39. Notify() • m thinks it might be n’s predecessor n.notify(m) if (predecessor is nil or m  (predecessor, n)) predecessor = m;

40. Fix_fingers() • Periodically called to make sure that fingertable entries are correct • New nodes initializetheir finger tables • Existing nodes incorporatenew nodes into their finger tables n.fix_fingers() next = next + 1 ; if (next > m) next = 1 ; finger[next] = find_successor(n + 2next-1);

41. Check_predecessor() • Periodically called to check whether predecessor has failed • If yes, it clears the predecessor pointer, which can then be modified by notify() n.check_predecessor() if (predecessor has failed) predecessor = nil;

42. Theorem 3 • If any sequence of join operations is executed interleaved with stabilizations, then at some time after the last join the successor pointers will form a cycle on all nodes in the network

43. Stabilization Protocol • Guarantees to add nodes in a fashion to preserve reach ability • By itself won’t correct a Chord system that has split into multiple disjoint cycles, or a single cycle that loops multiple times around the identifier space

44. Impact of Node Joins on Lookups • Correctness • If finger table entries are reasonably current • Lookup finds the correct successor in O(log N) steps • If successor pointers are correct but finger tables are incorrect • Correct lookup but slower • If incorrect successor pointers • Lookup may fail

45. Impact of Node Joins on Lookups • Performance • If stabilization is complete • Lookup can be done in O(log N) time • If stabilization is not complete • Existing nodes finger tables may not reflect the new nodes • Doesn’t significantly affect lookup speed • Newly joined nodes can affect the lookup speed, if the new nodes ID’s are in between target and target’s predecessor • Lookup will have to be forwarded through the intervening nodes, one at a time

46. Theorem 4 • If we take a stable network with N nodes with correct finger pointers, and another set of up to N nodes joins the network, and all successor pointers (but perhaps not all finger pointers) are correct, then lookups will still take O(log N) time with high probability

47. Failure and Replication • Correctness of the protocol relies on the fact of knowing correct successor • To improve robustness • Each node maintains a successor list of ‘r’ nodes • This can be handled using modified version of stabilize procedure • Also helps higher-layer software to replicate data

48. Theorem 5 • If we use successor list of length r = O(log N) in a network that is initially stable, and then every node fails with probability ½, then with high probability find_successor returns the closest living successor to the query key

49. Theorem 6 • In a network that is initially stable, if every node fails with probability ½, then the expected time to execute find_successor is O(log N)

50. Voluntary Node Departures • Can be treated as node failures • Two possible enhancements • Leaving node may transfers all its keys to its successor • Leaving node may notify its predecessor and successor about each other so that they can update their links