SOFTWARE ARCHITECTURES ARCHITECTURAL STYLES SCALING UP PERFORMANCE

Transcription

1 SOFTWARE ARCHITECTURES ARCHITECTURAL STYLES SCALING UP PERFORMANCE Tomas Cerny, Software Engineering, FEE, CTU in Prague,

2 ARCHITECTURES SW Architectures usually complex Often we reduce the abstraction Architectural Styles Layered style Architectural Patterns Model View Controller Tomas Cerny, Software Engineering, FEE, CTU in Prague,

3 ARCHITECTURE STYLES Basic Characteristics Quality attributes Tomas Cerny, Software Engineering, FEE, CTU in Prague,

4 ARCHITECTURE STYLES Data centric Databases Call and return Part of this course Implicit invocation Events Independent components Peer to peer Virtual Machines Pipe and Filter - data flow Tomas Cerny, Software Engineering, FEE, CTU in Prague,

5 OVERVIEW Domain and context model Arch. styles Reference architecture Domain and context model Small steps towards SW Architecture Reference Architecture Software Architecture System Architecture Architectural style

6 ARCHITECTURE STYLES Data centric Databases Voice recognition Compilers Tomas Cerny, Software Engineering, FEE, CTU in Prague,

7 ARCHITECTURE STYLES Call and return OOD Procedural RPC AOP Layers Tomas Cerny, Software Engineering, FEE, CTU in Prague,

8 ARCHITECTURE STYLES Call and return OOD Procedural RPC AOP Layers Tomas Cerny, Software Engineering, FEE, CTU in Prague,

9 ARCHITECTURE STYLES Call and return OOD Procedural RPC AOP Layers Tomas Cerny, Software Engineering, FEE, CTU in Prague,

10 ARCHITECTURE STYLES Call and return OOD Procedural RPC AOP OOP OOP + AOP Layers Source code of methods Source code of methods Aspects Security Method logic Synchronization Logging Tomas Cerny, Software Engineering, FEE, CTU in Prague,

11 ARCHITECTURE STYLES Call and return OOP OOP + AOP OOD Procedural RPC Compiler" Weaver" AOP Layers o *" Executable Compiler" o *" Executable Tomas Cerny, Software Engineering, FEE, CTU in Prague,

12 ARCHITECTURE STYLES Implicit invocation Events Tomas Cerny, Software Engineering, FEE, CTU in Prague,

13 ARCHITECTURE STYLES Implicit invocation Events Tomas Cerny, Software Engineering, FEE, CTU in Prague,

14 ARCHITECTURE STYLES Independent components Tomas Cerny, Software Engineering, FEE, CTU in Prague,

15 ARCHITECTURE STYLES Virtual machines Tomas Cerny, Software Engineering, FEE, CTU in Prague,

16 ARCHITECTURE STYLES Pipes and Filters Tomas Cerny, Software Engineering, FEE, CTU in Prague,

17 ARCHITECTURE STYLES Pipes and Filters Tomas Cerny, Software Engineering, FEE, CTU in Prague,

18 ARCHITECTURE STYLES Data centric Data integration, Distribution, Control, Coordination Scalability, Low coupling, Centralization, Reuse, Modifiable, Call and return Modifiable, Reusable, Inf. hiding, Structural decomposition, Separation of concerns Implicit invocation Modifiable, Low coupling, Hard to comprehend, Independent components Integration, Scalability, Reuse, Low coupling, Distribution, Reliability Virtual Machines Simulation, Emulation, Portability!, Flexibility, Lowered Performance, Extended features Pipe and Filter Modifiable, Reuse, Easy design, Simplicity, Low Coupling, Slow, No filter cooperation, Lot of parsing Tomas Cerny, Software Engineering, FEE, CTU in Prague,

19 SCALING PERFORMANCE Usual approach is to deploy app to a web server and provide access through HTTP/S Client-server architecture Inside 3-layers and data repository Tomas Cerny, Software Engineering, FEE, CTU in Prague,

20 SCALING PERFORMANCE Usual approach is to deploy app to a web server and provide access through HTTP/S Client-server architecture Inside 3-layers and data repository Tomas Cerny, Software Engineering, FEE, CTU in Prague,

21 SCALING PERFORMANCE Minimize volume Minimize volume ORM talks too much when not being careful Usually HTTP (JS/CSS) Tomas Cerny, Software Engineering, FEE, CTU in Prague,

22 DEPLOYMENT, MAINTENANCE AND REPORTS Get user experience Register your app at Google Analytics to see how it performs HTML5 allows you to see the statistics Tomas Cerny, Software Engineering, FEE, CTU in Prague,

23 CLIENT-SERVER ARCHITECTURE Properties: Centralization Easy with security Easy to locate Easy to scale Until we reach the limit Server is the bottleneck Performance influenced by the network conditions And virtual distance between client and server Server has given throughput Given by HW, our Design, Efficiency, Caching, etc. Tomas Cerny, Software Engineering, FEE, CTU in Prague,

24 CLIENT-SERVER ARCHITECTURE Client Server Tomas Cerny, Software Engineering, FEE, CTU in Prague,

25 CLIENT-SERVER ARCHITECTURE Server throughput 300 clients at once Client Server Tomas Cerny, Software Engineering, FEE, CTU in Prague,

26 CLIENT-SERVER ARCHITECTURE Through put 300 clients at once Client peak Server Load grows! Tomas Cerny, Software Engineering, FEE, CTU in Prague,

27 CLIENT-SERVER ARCHITECTURE Through put 300 clients at once Client peak Server Load grows! Tomas Cerny, Software Engineering, FEE, CTU in Prague,

28 CLIENT-SERVER ARCHITECTURE How to improve? Client peak Server Caching Performance analysis profiling Native/Custom SQL queries for reports Better Hardware, more CPU/Mem Tomas Cerny, Software Engineering, FEE, CTU in Prague,

29 CLIENT-SERVER ARCHITECTURE How to improve? Client peak Server What if it is not enough? Indirection? Tomas Cerny, Software Engineering, FEE, CTU in Prague,

30 CLIENT-SERVER ARCHITECTURE How to improve? Client peak Server Indirection? Dispatcher Tomas Cerny, Software Engineering, FEE, CTU in Prague,

31 CLIENT-DISPATCHER-SERVER ARCHITECTURE How to improve? Client peak Server Indirection? Dispatcher Tomas Cerny, Software Engineering, FEE, CTU in Prague,

32 CLIENT-DISPATCHER-SERVER ARCHITECTURE How to improve? Client peak Indirection? Dispatcher Server 1 Server 2 Tomas Cerny, Software Engineering, FEE, CTU in Prague,

33 CLIENT-DISPATCHER-SERVER ARCHITECTURE How to improve? Client peak Server 1 Indirection? Dispatcher Ser.. Server 2 Tomas Cerny, Software Engineering, FEE, CTU in Prague,

34 CLIENT-DISPATCHER-SERVER ARCHITECTURE Tomas Cerny, Software Engineering, FEE, CTU in Prague,

35 CLIENT-DISPATCHER-SERVER ARCHITECTURE Most likely we cannot expect to multiply the throughput of the single server Balancing overhead We can balance different resources Static vs. Dynamic Geo-location balancing Content-Delivery-Network (CDN) Static content (Akamai) Tomas Cerny, Software Engineering, FEE, CTU in Prague,

36 CONTENT DELIVERY NETWORK (CDN) Example Tomas Cerny, Software Engineering, FEE, CTU in Prague,

37 CONTENT DELIVERY NETWORK (CDN) Example Tomas Cerny, Software Engineering, FEE, CTU in Prague,

38 CONTENT DELIVERY NETWORK Example Tomas Cerny, Software Engineering, FEE, CTU in Prague,

39 CONTENT DELIVERY NETWORK Example Tomas Cerny, Software Engineering, FEE, CTU in Prague,

40 CONTENT DELIVERY NETWORK Example Tomas Cerny, Software Engineering, FEE, CTU in Prague,

41 CONTENT DELIVERY NETWORK Example Tomas Cerny, Software Engineering, FEE, CTU in Prague,

42 SCALING PERFORMANCE Database might be the bottleneck Database replication Tomas Cerny, Software Engineering, FEE, CTU in Prague,

43 SCALING PERFORMANCE Database might be the bottleneck Datagrid Tomas Cerny, Software Engineering, FEE, CTU in Prague,

44 SCALING PERFORMANCE JBoss view on Datagrid Tomas Cerny, Software Engineering, FEE, CTU in Prague,

45 SCALING PERFORMANCE JBoss view on Datagrid Tomas Cerny, Software Engineering, FEE, CTU in Prague,

46 SCALING PERFORMANCE JBoss view on Datagrid Tomas Cerny, Software Engineering, FEE, CTU in Prague,

47 SERVICE-ORIENTED ARCHITECTURE (SOA) So far we considered that server-side app offers data, knowledge and presentation Service does not provide presentation Well accepted format Standard : JSON, SOAP, XML.. Tomas Cerny, Software Engineering, FEE, CTU in Prague,

48 SERVICE-ORIENTED ARCHITECTURE Motivation Tomas Cerny, Software Engineering, FEE, CTU in Prague,

49 SERVICE-ORIENTED ARCHITECTURE Organization Focus Mainframe Centric Internal Use Unique Data Process Focus Client Server Partial Connectivity EDI File Transfer Distributed Functions Data Centric Universal Interoperability Real-time Connectivity Tomas Cerny, Software Engineering, FEE, CTU in Prague,

50 SERVICE Loose coupling Reusable Stateless Autonomous (independent) Discoverable Abstract Composable Platform independent Tomas Cerny, Software Engineering, FEE, CTU in Prague,

51 ANATOMY OF A SERVICE Service Consumer New Service Wrapped Legacy Interface Proxy Service Interface Service Implementation Composite Service

52 SERVICES COMMUNICATE WITH MESSAGES Providing reliability and security to messages Sending messages across consumers and producers Service Orchestration Service Consumers Service Producers

53 BASIC WEB SERVICES UDDI Registry Points to description WSDL Finds Service Points to service Describes Service Web Service Client (J2EE,.NET, PL/SQL ) SOAP Invokes with XML Messages Web Service (J2EE, PL/SQL,.NET,C/C++, Legacy )

54 ENTERPRISE SERVICE BUS (ESB) It is a software architecture model used for designing and implementing the interaction and communication between mutually interacting software applications in service-oriented architecture (SOA). Model for distributed computing Variant of client server software architecture model Promotes flexibility with regards to communication & interaction between applications. Primary use in enterprise application integration (EAI) of heterogeneous and complex landscapes. 54

55 ENTERPRISE SERVICE BUS 55

56 Master of Information System Management From JBoss ESB Documentation 56

57 SOA IS AN EVOLUTIONARY STEP

58 SOA IS AN EVOLUTIONARY STEP in distributed communications EAI Project-ware SOA

59 TO ENABLE BUSINESS PROCESS OPTIMIZATION AND THE REAL TIME ENTERPRISE (RTE) BPM Expressed in terms of Services Provided/Consumed Seamless End to End Process Service to Customers Enterprise Service from Multiple Suppliers Smart Clients Stores POS Mobile 3 rd Party Agents Portal Internal Systems SOA Patterns: Single, Multi-Channel Service for consistency SOA Pattern: Standardized Service provided by multiple suppliers

60 APPLICATION CENTRIC Business scope Narrow Consumers Limited Business Processes Application Finance Application Supply Application Integration Architecture bound to EAI vendor Redundancy Manufacturing Distribution Overlapped resources Overlapped providers Business functionality is duplicated in each application that requires it. EAI leverage application silos with the drawback of data and function redundancy.

61 SERVICE CENTRIC Business scope Multiple Service Consumers Multiple Business Processes Finance Service Service Supply Service Service Service Architecture Shared Services Manufacturing Distribution Multiple Discrete Resources Multiple Service Providers SOA structures the business and its systems as a set of capabilities that are offered as Services, organized into a Service Architecture Service virtualizes how that capability is performed, and where and by whom the resources are provided, enabling multiple providers and consumers to participate together in shared business activities.

62 SERVICE CENTRIC APPROACHES Open your business to extension and evolution! Natural extension and reuse Expedia API, Paypal, Amazon API, Airfare, Heureka.. Open your system to novel needs, requirements, interaction Reuse by other vendors!