• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
9 - Architetture Software - SOA Cloud

9 - Architetture Software - SOA Cloud






Total Views
Views on SlideShare
Embed Views



0 Embeds 0

No embeds



Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
Post Comment
Edit your comment

    9 - Architetture Software - SOA Cloud 9 - Architetture Software - SOA Cloud Presentation Transcript

    • Architectures for software servicesand cloud computing Paolo Ciancarini
    • Agenda  Rationale: the Service Economy  Software as a Service (SaaS)  Service oriented architectures  The REST architectural style  AJAX and mashups  Cloud computing architectures
    • The Service Economy  Modern economies rely upon services  This means that several companies offer support for activities in the primary (agricolture) or secondary (industry of goods) sectors, or to other companies offering services themselves  A service company is usually independent (not owned) from other companies and their services; however in several cases they should cooperate or at least be coordinated in some way
    • From Products to Services  Our society is shaped by the forces of   Specialization   Standardization   Scalability  It is now mostly “service” oriented   Transportation   Telecommunication   Retail   Healthcare   Financial services   Education and training   …
    • Attributes of physical services  A service is not owned by its user (compare with product)  Well defined, easy-to-use, standardized interface   New services can be offered by combining existing services  (almost) always available but idle until requests come  “Provision-able” (used by someone only when necessary, then reassigned to someone else)  Easily accessible and usable readily   no “integration” required   Self-contained : no visible dependencies to other services  Coarse grain  Independent of consumer context   but a service can have a context  Quantifiable quality of service   Do not compete on “What” but “How”
    • Example: mail  Customers use mail everywhere  Interfaces: PostOffice, Mail Box, stamp, postman  Apparently no relationship with other services (eg. Transportation, payments)  Quality of service: price, delivery time, lost messages
    • The global service economy Top Ten Nations by Labor Force Size (about 60% of world labor in just 10 nations) A = Agriculture, G = Goods, S = Services Nation % % % % 25 yr % 2008 WW A G S delta S 100% 90% Labor 80% 70% China 21.0 50 15 35 191 60% 50% India 17.0 60 17 23 28 40% 30% U.S. 4.8 3 27 70 21 20% 10% China 0% Indonesia 3.9 45 16 39 35 1810 1835 1860 1885 1910 1935 1960 1985 2010 Brazil 3.0 23 24 53 20 100% 90% (S) Services: Russia 2.5 12 23 65 38 80% Value from enhancing the capabilities of things (customizing, 70% distributing, etc.) and interactions between things Japan 2.4 5 25 70 40 60% 50% 40% Nigeria 2.2 70 10 20 30 30% 20% United States Banglad. 2.2 63 11 26 30 10% 0% 1800 1820 1840 1860 1880 1900 1920 1940 1960 1980 2000 Germany 1.4 3 33 64 44 >50% (S) services, >33% (S) services
    • What is meant by Service?  In economics and marketing, a service is the intangible equivalent of an economic good. Service provision is an economic activity where the buyer does not generally, except by exclusive contract, obtain exclusive ownership of the thing purchased (Wikipedia, 2010)  A service is a “provider to client” interaction that creates and captures value while sharing the risks of the interactions  Services are the application of specialized competences (skills and knowledge) through deeds, processes, and performances for the benefit of another entity or the entity itself (Vargo & Lusch, 2004a)  Services are value that can be rented (in the broad sense) by the application of some process that the renter (client) participates in. This contrasts with goods, whose value (once purchased) is owned by the customer (Lovelock & Gummesson, 2004)
    • What is a Service (1)  A facility supplying some public demand  the work performed by one that offers help, use, or benefit Merriam-Webster
    • What is a Service (2)  In economics and marketing, a service is a non-material equivalent of a good  Service composition is managed by service providers who have to offer or provision more complex services  The service-good spectrum shows that most services include some goods, and also that most goods involve some service  Example: restaurant en.wikipedia.org/wiki/Service
    • DiscussionIs software a good or a service? 11
    • Software-as-a-Service (SaaS)  1990: Web 0.9 (Tim Berners-Lee)  1995: Web 1.0 (some dynamic content, Netscape)  1997: Services (Google, e-commerce...)  1999: LoudCLoud (first infrastructure for SaaS)  2000: Web 2.0 (rich UIs, social computing)  2004: SaaS & SOA (Google Maps, Amazon S3...)  2008: Cloud Computing (pay as you go) 12
    • Software as a serviceService: the non-material equivalent of a good  Software is a service at heart, albeit an automated one, but it is sold much like a manufactured good. Customers have to pay large sums of money up front, bear much of the risk that a program may not work as promised, and cannot readily switch vendors. The Economist, 2003
    • Software-as-a-Service (SaaS)  Definition of SaaS: Internet-based deployment (for access and management) of commercially available software  Service managed from “central” locations enabling customers to access applications remotely via web  Delivery: one-to-many model (single instance, multi-tenant architecture) including architecture, pricing, partnering, and management characteristics  Centralized updating, which obviates the need for end-users to download patches and upgrades  Frequent integration into a larger network of communicating software—either as part of a mashup or a plugin to a platform as a service 14
    • Software as a service  An operating system service?   Program execution, file management  A Software Service?   Messaging Service   Log Service   Peer to peer distribution   Search engine  A Business Service?   Customer service   Bidding service
    • Software services: a multidisciplinary problem •  Industry policies Business •  Industry imperatives Issues •  Business priorities Business ModelFunctional Issues •  Functionality expressed as Services •  Map to Business Processes •  Based on standards Service Oriented Architecture Operational •  Model Driven Development Issues •  Middleware and Software •  Technology Platforms & Solutions Realized on an Infrastructure Frameworks •  Network & Hardware
    • Dimensions of Sw Engineering for ServicesApplication Middleware Distributed Model Computing Mainframe MQ Client/Server EAI, B2B DCE Web Apps BPM CORBA, J2EE Portals WS, ebXML Software Construction SalesForce.com Service Orientation SAP, Oracle… Object Orientation MDDSolutions XP Structured Computing RUP Programming Model Methodologies
    • SOC and SOA  Service Oriented Computing is a distributed programming paradigm for business applications supporting complex enterprises  Service Oriented Architectures are systems of services producers and service consumers  Mostly made of independent software systems able to be dynamically integrated
    • SOA  SOA = Service Oriented Architecture  SOA is a recent term, but in the past have been built many systems that today we could call SOAs. Example: the Web  The essence of a SOA lies in independent services interconnected by messages  On the Web, a specific SOA technology is Web Services (W3C)
    • 20
    • OMG Standards for SOA http://www.omg.org/attachments/pdf/OMG-and-the-SOA.pdf BPMN Business Process UML BPDM SBVR Task definition Interfaces defined by enterprise context Taskimplementation ODM UML Business Services RAS Interfaces defined by enterprise semantics and requirements J2EE UML ProfileComponents CORBA UML Profile CWM/Operational IMMResources KDM Servers Mainframes Servers Mainframes Data Data Data Data
    • Typical topics in SOA  Model, design, and implement a SOA  Create an agile, reusable SOA  Automate business processes by mapping to the architectural model  Orchestrate services and execute processes with the Business Process Execution Language (BPEL)  Achieve interoperability within a SOA using proven standards and best practices  Secure and govern an enterprise SOA
    • Standardizing bodies for SOAs  W3C (1994)  OASIS (1993), consortium of former GML providers, deals with applications using XML  OMG (1989)  WS-I (2002), promotes interoperability among the stack of Web Services specifications"
    • Elements of SOA Design  Business Modeling  Service Oriented Architectural Modeling and Design  Model Driven assumptions (loose coupling)  Distributed objects and MOM (Message Oriented Middleware) for component-based sw systems
    • Business modeling http://www.ibm.com/developerworks/rational/library/360.html 25
    • Service oriented modeling and architecting 26
    • Model driven SOA designwww.theenterprisearchitect.eu/archive/2009/06/03/a-framework-for-model-driven-soa 27
    • Services and SOAs  A service is a program interacting via message exchanges   On Web all messages and service descriptions are written in XML  A SOA is a set of deployed services cooperating in a given task   Adapt to new services after deployment
    • SOA Concepts Service manage OperationalRequirements bound by Expose itself using enforce State Policies Edge implement Process/Creategoverned by Message Exchange Pattern Messages Contracts describe direction & types of contain Schemas define structure of
    • SOA Reference Architecture AJAX Portlets WSRP B2B Other consumersService Consumer Integration (Enterprise Service Bus Approach) Data Architecture and Business Intelligence business processes process choreography Monitoring Infrastructure Service QoS, Security, Management, and Governance services atomic and compositeService Provider service components operational systems Packaged Custom OO Application Application Application
    • Main SOA Standards  BPEL (Business Process Execution Language): a standard for assembling sets of discrete services into an end-to-end business process  J2EE: the Java Platform, Enterprise Edition, with APIs for deploying and managing Web services  JSR 168: standard for portal and portlet interoperability  JSR 181: an API for Web services metadata annotation  SOAP (Simple Object Access Protocol): a W3C-approved standard for exchanging information among applications  UDDI (Universal Description, Discovery, and Integration): an OASIS-approved standard specification for defining Web service registries  WS-I (Web Services Interoperability): an open industry organization promoting Web services interoperability across platforms, operating systems, and languages  WS-Reliability (Web Services Reliability): a SOAP-based protocol for exchanging SOAP messages, with delivery and message-ordering guarantees  WS-Security (Web Services Security): a SOAP-based protocol that addresses data integrity, confidentiality, and authentication in Web services  WSDL (Web Service Description Language): a W3C-approved standard for using XML to define Web services  WSIF (Web Services Invocation Framework): an open source standard for specifying, in WSDL, EJB implementations for the Web server  WSRP (Web Services for Remote Portlets): an OASIS standard for integrating remote Web services into portals  XML (Extensible Markup Language): a data markup language for Web services
    • What is SOA?  Use other services as RPC servers for your app  Web 1.0: large sites organized this way internally   Yahoo!, Amazon, Google, …: external “Services” available, but complex: Doubleclick ads, Akamai  XML based Web Services msgs and descriptions  Web 2.0: consumer-facing service API’s   Services: Google API, Google Analytics, Amazon CloudFront...   Platforms: Facebook, Google Maps, ...   Mashups, e.g. housingmaps.com   User-composable services, e.g. Yahoo Pipes 32
    • Amazon.com: Web 1.0 SOA  ~50 “two-pizza” teams of web server “developer/operators” web server web server  ~10 operators   monitor the whole site   page the resolvers on alarm web server web server service A web server web server service B web server web server service C  ~1000 resolvers   10-15 per team, 1 on-call 24x7 DB DB DB   monitor own service, fix problems  Over 140 code change commits/month  Internal microcosm of service-oriented architecture (as were Yahoo, Google, others)P. Bodík et al., Advanced Tools for Operators at Amazon.com, Proc. ICAC 2005
    • SOA = RPC  Transport: HTTP(S)  Data interchange:   XML DTD (e.g., RSS)   JSON (Javascript Object Notation)  Request protocol:   SOAP (Simple Object Access Protocol)   JSON-RPC  On the horizon: WebHooks (HTTP POST callback, for “push”) 34
    • JSON-RPC  Open connection to designated port on server  Send HTTP method & request URI, with MIME type of bodyset to application/json  Then send request body:{ "version": "1.1", "method": "confirmFruitPurchase", "id": "194521489", "params": [ [ "apple", "orange", "pear" ], 1.123 ]}  Response might be something like this:{ "version": "1.1", "result": "done", "error": null, "id": "194521489"}  You have to handle substantially all errors 35
    • RSS  Request is a regular HTTP GET to a specified URL<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1"> <channel> <title>Altarena Playhouse Ticket Availability</title> <link>http://www.audience1st.com/altarena/store</link> <description>Altarena Ticket Availability</description> <item> <title>Sylvia - Friday, May 14, 8:00 PM – Buy now</title> <link>http://.../store?showdate_id=347</link> <guid permalink="false">http://.../store?ts=1271058414</guid> </item> <item> ... </item> </channel></rss> 36
    • AJAX and SOA  AJAX: client  server   client makes (async) requests to HTTP server   client-side JavaScript upcall receives reply and decides what to do   commonly, response includes XHTML/XML to update page, or JavaScript to execute   Doesn’t really make sense except in context of client  SOA: server  server or client  server   one principal makes (sync or async) requests to an HTTP server   formerly, principal was a server running some app   today, powerful JavaScript clients blur the line 37
    • Two ways to do it... “Thin” Web 2.0 browser Craigslist.org app client “Fat” browser Google Maps client+ Client portability+/– Client performance (both app download & JavaScript execution)+ Availability of utility libraries for app development– Privacy/trustworthiness of aggregator app– Caching 38
    • REST (Representational State Transfer) style  Architectural style:   Client-server, stateless, cache   description of properties that made Web 1.0 successful by constraining SOA interactions  In context of SOA for Web 2.0   HTTP is transport; HTTP methods (get, put, etc.) are the only commands   URI names are a resource   Client has resource  has enough info to request modification of the resource on server   A cookie can encode part of transferred state  If your app is RESTful, it’s easy to “SOA”-ify 39
    • What is REST?  REST is a term coined by Roy Fielding in his Ph.D dissertation to describe an architectural style of networked software systems  "Representation State Transfer is intended to evoke an image of how a well-designed Web application behaves: a network of web pages (a virtual state-machine), where the user progresses through an application by selecting links (state transitions), resulting in the next page (representing the next state of the application) being transferred to the user and rendered for use."http://www.ebuilt.com/fielding/pubs/dissertation/top.htm
    • Why is it called "Representation State Transfer"? http://www.boeing.com/aircraft/747 Client Resource Fuel requirements Maintenance ... 747.htmlThe Client references a Web resource using a URL. A representation of the resourceis returned (in this case as an HTML document).The representation (e.g., Boeing747.html) places the client application in a state.The result of the client traversing a hyperlink 747.html is another resource is accessed.The new representation places the client application into yet another state.Thus, the client application changes (transfers) state with each resource representation --> Representation State Transfer!
    • Motivation for REST "The motivation for developing REST was to create an architectural model for how the Web should work,such that it could serve as a framework for the Web protocol standards. REST has been applied to describe the desired Web architecture, help identify existing problems, compare alternative solutions, and ensure that protocol extensionswould not violate the core constraints that make the Web successful." - Roy Fielding
    • REST with HTTP examplesHTTP GET! HTTP PUT! HTTP POST! HTTP DELETE!Collection URI, such as http://example.com/customers/257/ordersList the members of Replace the Create a new entry in the delete thethe collection, entire collection collection. The ID created is entirecomplete with their with another usually included as part of collection"member URIs for collection" the data returned by thisfurther navigation" operation."HTTP GET! HTTP PUT! HTTP POST! HTTP DELETE!Element URI, such as http://example.com/resources/7HOU57YRetrieve a Update (or Treats the addressed Delete therepresentation of create) the member as a addressed memberthe addressed addressed collection in its own of the collection. ""member of the member of the right and creates a "collection in an collection" new subordinate of it."appropriate MIMEtype" 43
    • Example: REST vs SOAP-based  A company deploying 3 Web services to enable its customers to:   get a list of parts   get detailed information about a particular part   submit a Purchase Order (PO)  See the REST solution first, then the SOAP solution 44
    • The REST way of Implementing the Web Services HTTP GET request URL 1 Parts Response List (HTML/XML doc) HTTP response Web Server HTTP GET request URL 2 Part Response (HTML/XML doc) HTTP responseHTTP POST (HTML/XML) PO URL 3 POURL to submitted PO HTTP response
    • Implementing the Web Service using SOAP•  Service: Get detailed information about a particular part –  The client creates a SOAP document that specifies the procedure desired, along with the part-id parameter. <?xml version="1.0"?> <soap:Envelope xmlns:soap="http://schemas.xmlsoap.org/soap/envelope/"> <soap:Body> <p:getPart xmlns:p="http://www.parts-depot.com"> <part-id>00345</part-id> </p:getPart> </soap:Body> </soap:Envelope>Again, the client will HTTP POST this document to the SOAP server at: http://www.parts-depot.com/soap/servlet/messagerouterNote that this is the same URL as was used when requesting the parts list.The SOAP server peeks into this document to determine what procedure to invoke.
    • Implementing the Web Services using SOAP SOAP envelope Request HTTP POST (XML doc) URL 1 getPartsList()Response HTTP Response(XML doc) Request Web Server HTTP POST (XML doc) URL 1Response SOAP Server getPart(id) HTTP Response(XML doc) PO HTTP POST (XML doc) URL 1 submit(PO)Response HTTP Response(XML doc) Note the use of the same URL (URL 1) for all transactions. The SOAP Server parses the SOAP message to determine which method to invoke. All SOAP messages are sent using an HTTP POST.
    • A RESTful architecture for BPM 48
    • XML (eXtensible Markup Language)<?xml version="1.0" encoding="UTF-8"?><book year="1967"> <title>The politics of experience</title> <author> <firstname>Ronald</firstname> <lastname>Laing</lastname> </author></book>  Really a metalanguage for describing hierarchical, semistructured, schema-less data  XML Document Type Definition (DTD) specifies structural & content constraints on a particular document type 49
    • XML (eXtensible Markup Language)<?xml version="1.0" encoding="UTF-8"?><book year="1967"> Attribute Value <title>The politics of experience</title> <author> Element <firstname>Ronald</firstname> Element <lastname>Laing</lastname> </author></book>  Really a metalanguage for describing hierarchical, semistructured, schema-less data  XML Document Type Definition (DTD) specifies structural & content constraints on a particular document type 50
    • XHTML & CSS  XHTML: a document conforming to a particular DTD describing a hierarchical collection of HTML elements   Variants: Strict, loose, transitional (for compatibility with deterioriating HTML syntax 1990-95) <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">   inline (headings, tables, lists...)   embedded (images, video, Java applets, JavaScript code...)   fill-in forms—text, radio/check buttons, dropdown menus..., marshaling arguments into either URI or request body  CSS (Cascading Stylesheets) for presentation   Strict XHML forbids presentational markup   Idea: complete separation of appearance from structure 51
    • CSS Styles get visual styling applying selectors <link rel="stylesheet" href="mystyles.css"/> <div class="pageFrame" id="pageHead"> <h1> Welcome, <span id="custName">Armando</span> </h1> </div>  In mystyles.css (static asset with MIME type text/css):div.pageFrame { background-image: url(/banner.gif); }h1 { font-size: large; float: left; }#custName:hover { background-color: yellow; font-weight: bold; }  Style properties include borders, background images, and layout directives (floating, absolute positioning, min/max scaled sizes, etc.)  Changing style properties has side effect of re-rendering  e.g., change display or visibility property to show/hide elements
    • 53
    • 54
    • Dynamic content generation  Most Web 1.0 (e-commerce) sites actually run a program that generates the output  Originally: templates with embedded code “snippets”  Eventually, embedded code became “tail that wagged the dog” and moved out of the Web server  Languages/frameworks evolved to capture common tasks   Perl, PHP, Python, Ruby on Rails, ASP, ASP.NET, Java Servlet Pages, Java Beans/J2EE, ... 55
    • SaaS 3-tiers architecture  Common gateway interface (cgi): allows a Web persistent storage server to run a program storage   Server maps some URI’s to application names   When the app is run, it gets the complete HTTP request including headers app application server   “Arguments” embedded in URL with “&” syntax or sent as request body (with POST) http://www.foo.com/search?term=white %20rabbit&show=10&page=1 HTTP   App generates the entire response server   content (HTML? an image? some javascript?)   HTTP headers & response code  Plug-in modules for Web servers allow long- running CGI programs & link to language interpreters •  Various frameworks have evolved to capture this structure! 56
    • SaaS 3-tiers deployment  HTTP server (“web server”) persistent   “fat” (e.g. Apache): support virtual hosts, plugins for storage storage multiple languages, URL rewriting, reverse proxying, ....   “thin” (nginx, thin, Tomcat, ...): bare-bones machinery to support one language/framework; no frills app application   Application server server 1.  separate server process, front-ended by a “thin” HTTP server 2.  or linked to an Apache worker via FastCGI or web server HTTP plug-in: mod_perl, mod_php, mod_rails, ... server   Apache can spawn/quiesce/reap independent processes  Persistent storage   Common RDBMS (MySQL, PostgreSQL, etc.)   communicate w/app via proprietary or standardized database “connector” (ODBC, JDBC, ...)  Hence LAMP: Linux, Apache, MySQL, PHP/Perl 57
    • Frameworks  Support for more languages: Apache modules (mod_perl, mod_php, mod_rails ...)   avoid spawning new process per request   typically embed language interpreter in Apache  Support for common idioms like sessions   Cookie management   virtualize connection to database   “dispatcher” interactions with front-end HTTP server  Early “templating systems” (e.g. PHP) vs. modern “full stack frameworks” (e.g. Rails) 58
    • Example: Rails, a Ruby-based MVC Framework•  Implemented almost entirely in Ruby•  Distributed as a Ruby “gem” (collection of related libraries & tools)•  Connectors for most popular databases Relational mysql or Model, View, Controller tables Database sqlite3 Subclasses of ActiveRecord::Base models/*.rb Ruby your app interpreter views/*.html.erb controllers/*.rb Rails Rails CGI or other dispatching rendering routing apache Subclasses of Subclasses of ActionView ApplicationController firefox
    • What’s new in Web 2.0?  Primitive UI => Rich UI   enable “desktop-like” interactive Web apps   enable browser as universal app platform on cell phones  “Mass customize” to consumer => Social computing   tagging (Digg), collaborative filtering (Amazon reviews), etc. => primary value from users & their social networks   write-heavy workloads (Web 1.0 was read-mostly)   lots of short writes with hard-to-capture locality (hard to shard)  Libraries => Service-oriented architecture   Integrate power of other sites with your own (e.g. mashups that exploit Google Maps; Google Checkout shopping cart/payment)   Pay-as-you-go democratization of “services are king”   Focus on your core innovation  Buy & rack => Pay-as-you-go Cloud Computing 60
    • Rich Internet Apps (RIAs)  Closing gap between desktop & Web   Highly responsive UI’s that don’t require server roundtrip per-action   More flexible drawing/rendering facilities (e.g. sprite-based animation)   Implies sophisticated client-side programmability   Local storage, so can function when disconnected  early example: Google Docs + Google Gears   include offline support, local storage, support for video, support for arbitrary drawing, ...  currently many technologies—Google Gears, Flash, MS Silverlight...   client interpreter must be embedded in browser (plugin, extension, etc.)   typically has access to low-level browser state => new security issues   N choices for framework * M browsers = N*M security headaches  proposed HTML5 may obsolete some of these 61
    • Rich UI with AJAX (Asynchronous Javascript and XML)  Web 1.0 GUI: click → page reload  Web 2.0: click → page can update in place   also timer-based interactions, drag-and-drop, animations, etc.How is this done? 1.  Document Object Model (1998, W3C) represents document as a hierarchy of elements 2.  JavaScript (1995; now ECMAScript) makes DOM available programmatically, allowing modification of page elements after page loaded 3.  XMLHttpRequest (2000) allows async HTTP transactions decoupled from page reload 4.  JavaScript libraries (jQuery, Prototype, script.aculo.us) encapsulate useful abstractions 62
    • DOM & JavaScript: Document = tree of objects  hierarchical object model representing HTML or XML doc  Exposed to JavaScript interpreter   Inspect DOM element value/attribs   Change value/attribs → redisplay or fetch new content from server  Every element can be given a unique ID  JavaScript code can walk the DOM tree or select specific nodes via provided methods<input type="text" name="phone_number" id="phone_number"/>!<script type="text/javascript">! var phone = document.getElementById(phone_number);! phone.value=555-1212;! phone.disabled=true;! document.images[0].src="http://.../some_other_image.jpg";!</script>! 63
    • JavaScript  A browser-embedded scripting language   OOP: classes, objects, first-class functions, closures   dynamic: dynamic types, code generation at runtime   JS code can be embedded inline into document... <script type="text/javascript"> <!-- # protect older browsers calculate = function() { ... } // --> </script>   ...or referenced remotely: <script src="http://evil.com/Pwn.js"/>  Current page DOM available via window, document objects   Handlers (callbacks) for UI & timer events can be attached to JS code, either inline or by function name: onClick, onMouseOver,... Changing attributes/values of DOM elements has side-effects, e.g.: <a href="#" onClick="this.innerHTML=Presto!">Click me</a> 64
    • AJAX = Asynchronous Javascript and Xml   Recipe:   attach JS handlers to events on DOM objects   in handler, inspect/modify DOM elements and optionally do asynchronous HTTP request to server   register callback to receive server response   response callback modified DOM using server-provided info   JavaScript as a target language   Google Web Toolkit (GWT): compile Java => emit JS   Rails: runtime code generation ties app abstractions to JS 65
    • A browser is an application container   “Web apps” include animation, sound, 3D graphics, disconnection, responsive GUI...   Browser =~ new OS: manage mutually-untrusting apps (sites) Source: Robert OʼCallahan (Mozilla.org),! Inside Firefox! 66
    • Social Computing  Web 1.0: add value via mass customization   select content/presentation for you based on best guesses about your interests   resource: demographic/analytic data about you  Web 2.0: add value via connecting to social network   vendor: your friends’ interests are a good indicator of your interests   user: value added to existing content == how your friends interact with it   resource: your social network  From social networking site to social network as a way of structuring applications 67
    • Social Computing  Amount of content “created” by each user small!   e.g., Digg article, rate video, play a Facebook game  but still creates lots of short random writes   consider “Like” feature on Facebook   social graphs naturally hard to partition (though would love to see a paper about this from FB)  question for Web 2.0 developers is not whether social computing is part of your app, but how  later we will discuss technical architecture of “connecting” an app to social networks 68
    • Facebook  Facebook plug-in apps FBQL! FB data! 1. Facebook.com! 2.! Your app! html 4. 3.! AJAX SO •  Facebook platform (“Facebook Connect”)! A FB data! REST! 1. html+ Your app! 2 (opt.). Facebook.com! 3. xfbml 4. èREST via JavaScript & XFBML! 69 ç HTML IFRAME w/FB content!
    • Google Maps  Your app embeds Javascript-heavy client code (provided by Google)   client-side functionality: clear/draw overlays, etc.   server-side functionality: fetch new map, rescale, geocoding  Attach callbacks (handled by your app) to UI actions  Result of callback can trigger additional calls to Google Maps code, which in turn contact GMaps servers 1. Your Google 2. app Maps html+ js 4. 3. 70
    • Mashups: housingmaps.com 71
    • Scaling via Replication  The “most general” deployment scenario for a 3 tier Web application DB DB?   Many Web servers   possibly including static-asset servers   L4/L7 load balancers distribute load among them App … App  Caches and reverse proxies remember previously-computed content $ … $   whole page caching   page fragment caching, query caching WS … WS   Apache in reverse-proxy mode, or memcached process(es) addressed by Asset app server LB …LB Svr  Integration of caching with application logic the Internets varies by framework 72
    • “Scale makes availability affordable”  Goal: interchangeability (send any user request to any available server) DB DB?   each server handles 1/N load   affinity can be used to “soft-pin” users to App … App particular servers   requires good support for session state $ $ … abstraction in app framework  lose 1 server => lose 1/N capacity WS … WS   Load Balancers have logic to detect failed servers & remove from rotation Asset LB …LB Svr until they are resurrected the Internets 73
    • Asset Servers  For serving static assets (images, sound clips, CSS, etc.)  Separate Web server process, configuration optimized for fast static file serving  Web 2.0: use Amazon S3 (blob store) or CloudFront (CDN)   helps to have good asset-server abstraction in app framework 74
    • Deployment scenarios (& approximate pricing)  Buy/rack/install/configure it yourself...that’s so Web 1.0  Shared hosting ($3/month)   turnkey support for popular frameworks, hosted versions of popular building blocks (e.g. MySQL)   highly variable performance, multitenant per machine  Virtual private host ($10/month)   better isolation and security through virtualization   substantially more administration  “Framework VM” or “curated” environments (Heroku, Google AppEngine, Force.com) – pricing varies   hosted extensions: memcached, profiling, etc.   integration of 3rd party hosted services, e.g. Amazon S3 backup  Cloud Computing 75
    • Pay-as-you-go Cloud Computing  Amazon Elastic Compute Cloud (EC2)  “Compute units” $0.085/hr & up   1 CU ≈ 1.0-1.2 GHz 2007 AMD Opteron/Xeon core “Instances”! Platform! Cores! Memory! Disk! Small - $0.085 / hr" 32-bit" 1" 1.7 GB" 160 GB" Large - $0.34/ hr" 64-bit" 4" 7.5 GB" 850 GB – 2 spindles"  N XLarge - $0.68/ hr"64-bit" 8" 15.0 GB" 1690 GB – 3 spindles"  No up-front cost, no contract, noextra disk, ..." Options....extra memory, extra CPU, minimum  storage (~0.15/GB/month)  network (~0.10-0.15/GB external; 0.00 internal)  Everything virtualized, even concept of 76 independent failure 76
    • Cloud Economics: user side Cloud Computing: User side " Static provisioning for peak" - wasteful, but necessary for SLA" Capacity $Machines Demand Capacity Demand Time Time “Statically provisioned” “Virtual” data center data center in the cloud Unused resources 77
    • Cloud Economics 101Cloud Computing Provider: Could save energy " Capacity Energy Machines Demand Capacity Demand Time Time “Statically provisioned” Real data center data center in the cloud Unused resources 78
    • Risk of OverprovisioningUnderutilization results if “peak” predictions are too optimistic" Capacity Unused resources Resources Demand Time Static data center 79
    • “Risk Transfer” to Cloud  “Cost associativity” from linear pricing: 1,000 CPUs for 1 hour same price as 1 CPUs for 1,000 hours (@$0.10/hour)   Washington Post converted Hillary Clinton’s travel documents to post on WWW <1 day after released   Lab graduate students demonstrate improved Hadoop (batch job) scheduler—on 1,000 servers  Major enabler for SaaS startups   Animoto traffic doubled every 12 hours for 3 days when released as Facebook plug-in   Scaled from 50 to >3500 servers   ...then scaled back down  Goal: fix any transient problem by adding/removing nodes   Single-node performance becomes much less important 80
    • Classifying Clouds for Web 2.0  Instruction Set VM (Amazon EC2)  Managed runtime VM (Microsoft Azure)  Curated “IDE-as-a-service” (Heroku)  Platform as service (Google AppEngine, Force.com)  flexibility/portability vs. built-in functionality Higher-level, Lower-level, More managed, Less managed more value-added SW EC2 Joyent Azure Heroku, Force.com AppEngine 81
    • Deployment  “Deployment-as-a-service” increasingly common   monthly pay-as-you-go curated environment (Heroku)   hourly pay-as-you-go cloud computing (EC2)   hybrid: overflow from fixed capacity to elastic capacity   Remember administration costs when comparing!  Good framework can help at deployment time   Separate abstractions for different types of state: session state, asset server, caching, database   ORM – natural fit for social computing, and abstracts away from SQL (vs Web 1.0 PHP, e.g.)   REST – encourages you to make your app RESTful from start, so that “SOA”-ifying it is trivial  Scaling structured storage: open challenge 82
    • Cloud Computing  CC makes it possible to obtain results on 1000’s of servers   Es. BOOM (declarative cloud programming) boom.cs.berkeley.edu  Eucalyptus www.eucalyptus.com makes hybrid cloud computing reasonably practical   Run small experiments locally, then “scale up” to cloud for results  Why aren’t you using cloud computing yet? 83
    • Example: Facebook  Facebook has 2 datacenters, 1 per coast   reads spread across both   writes only to W. Coast; periodically (~10 minutes) replicated to E. Coast   >2000 MySQL servers, >25TB RAM for memcached  Challenge: inconsistency due to stale data   I change status message => Friends on East Coast datacenter do not see my change for 10 min   What if E.Coast person changes his own status?? 84
    • Facebook cloud serversSource: “How Facebook Works”,Technology Review, Jul/Aug 2008 85
    • SCADS: Scalable, Consistency-Adjustable Data Storage   Most popular websites follow the same pattern   Outgrow initial prototype (on MySQL) due to scale   Build large, complicated ad-hoc systems to deal with scaling limitations as they arise   Want Scale Independence as new users join:   No changes to application   Cost per user & request latency don’t increase   Key Innovations 1.  Performance-{safe,insightful} query language 2.  Declarative performance/consistency tradeoffs 3.  Automatic scale up & down using machine learningM. Armbrust et al., SCADS: Scalable Consistency-Adjustable Data Storage for Interactive Applications. Proc. CIDR 2009! M. Armbrust et al., PIQL: A Performance-Insightful Query Language. Proc. SOCC 2010. !
    • Understanding workload spikes  death of Michael Jackson: 23% of tweets, 15% of Wikipedia traffic...how to understand/synthesize spikes?   characterized 5 real spikes   large differences in many characteristics: steepness, magnitude, # data hotspots, …   spike synthesis tool generates realistic workload with few parameters using statistically sound model  Cloud computing offers alternative to overprovisioning? model   Explora/on  of  performance  [ACDC  ’09]   policy   Policy  simulator  [HotCloud  ’09]   workload   ac/ons   statefu Resource  alloca/on  for  stateful   l   systems   system     P. Bodik et al., Characterizing, modeling, and generating workload spikes for stateful services, SOCC 2010!P. Bodik et al., Statistical machine learning makes automatic control practical for Internet datacenters, HotCloud 09! P. Bodik et al., Automatic exploration of datacenter performance regimes, ACDC 09 ! ! !
    • Browsers are more and more fat  Performance, security and GUI choke point   Cookie management, JavaScript, XSS attacks   CSS layout, side-effects of DOM manipulation   AJAX: document can change anytime after loading   RIA frameworks/plugins: Flash, HTML 5, Silverlight   Device variation: new twist due to power limits   See http://www.google.com/googlebooks/chrome for a great fun-to-read description in comic-book form  Bring “desktop-quality” browsing to mobile tel   Enabled by 4G networks, better output devices   Parsing, rendering, scripting are all potential bottlenecks 88 
    • Parallelizing the Browser (R.Bodík et al.)  New algorithms for CSS, layout, rendering, up to 80X speedup on popular web sites  Constraint-based browser programming language unifies layout, rendering, scripting Slashdot (CSS Selectors) 50 45 40 2ms Speedup 35 30 25 20 15 10 5 0 84ms 1 2 3 4 5 6 7 8 Hardware Contexts 89
    • Try it yourself1.  Download & install RoR (rubyonrails.org)   A good intro book: Simply Rails 2 by Patrick Lenz2.  Signup for a free Heroku.com trial account3.  Install various useful “Ruby gems” (libraries): gem install heroku cucumber-rails rspec-rails4.  Create app skeleton: rails mynewapp ; git commit –m “new app” mynewapp heroku create mynewapp5.  Start writing user stories and track your progress (PivotalTracker.com)6.  When ready to deploy on Heroku: git push heroku masterMore resources: http://radlab.cs.berkeley.edu > Courses 90
    • Summary  New ecosystem: changes software process, deployment process, interdependencies of DADO (everything-as-a-service)  Web 2.0 tools have reached a new plateau of productivity  Gap closing between “web app” user experience vs. desktop app  New interesting research problems 91
    • Self test  Which are the consequences of defining software a service insted of a good?  What are the main original features of the REST style?  Which software architectural issues are typical of social software?  What are the economic basis of cloud computing? 92
    • References  www.armandofox.com/geek/teaching/  radlab.cs.berkeley.edu  msdn.microsoft.com/en-us/library/aa479069.aspx 93
    • Questions?