(2) Abstract

Mobile application development faces the same performance challenges as application development in general, which means that performance considerations have to be an integral part of any real-world application design. However, on mobile devices, the challenges related to performance issues are quite a bit bigger, because the runtime platforms may be limited in their functionality, the platforms are inherently limited in their resources (mostly processing power and memory), and the devices often have connectivity that may be substantially slower than best-case scenarios, or may even be intermittent.

(3) Mobile Application Design

Design depends largely on constraints.

Performance Bottlenecks

Outline (Performance Bottlenecks)

1. Performance Bottlenecks [4]
2. Client-Side Performance Analysis [6]

Performance Bottlenecks Erik Wilde UC Berkeley School of Information : Constraints of Mobile Applications

(5) Client-Side Constraints

• Limitations of the Mobile Platform
  • general limitations of the mobile platform ( HTML5 has no camera API )
  • specific limitations of one implementation of that platform ( Firefox has no SQL support )
  • memory is a problem on all mobile platforms (no support for virtual memory)
  • limited processing power is a problem on all mobile platforms
  • smartphones are increasingly used as general-purpose computers
  • layering creates inefficiencies (JavaScript on iPhone 100 times slower)
  • layering takes away possibilities for optimization (UI and execution)
  • some platforms oppose/prevent layering out of general principle

  Performance Bottlenecks Erik Wilde UC Berkeley School of Information : Constraints of Mobile Applications

  (6) Client Devices

  • iPhone hardware comes in two different memory configurations
    • first generation iPhones have 128MB of RAM
    • 3G and 3GS iPhones have 256MB of RAM
    • the Droid [Hardware/Platform Opinions] has 256MB of RAM
    • for many devices getting hardware specs is almost impossible
    • Android applications can only be installed in the onboard memory (Droid: 512MB)
    • applications requiring a lot of local data should store it separately

    Performance Bottlenecks Erik Wilde UC Berkeley School of Information : Constraints of Mobile Applications

    (7) Network Constraints

    • Network connectivity is potentially slow and unreliable
      • applications should be designed to work well with a slow network
      • applications should be designed to behave reasonably with no network
      • considering offline behavior in use cases should always be included
      • not all browsers support offline applications
      • browsers may start removing cached applications more aggressively
      • HTML5 AppCache [Offline-Capable Applications; Caching Application Code (1)] is only a caching hint for a browser

      Performance Bottlenecks Erik Wilde UC Berkeley School of Information : Constraints of Mobile Applications

      (8) Native vs. Offline Web

      • Native applications are under control of the user
        • users decide what to install, update, and uninstall
        • the only networking part is how to get the installation package
        • services decide to provide a manifest and become off-line capable
        • browsers implement some strategy for caching those applications
        • updates have to be initiated and implemented by the service
        • giving users better control over what to cache
        • caching data probably should be handled different from caching code
        • allowing applications to expose widget-like behavior

        Client-Side Performance Analysis

        Outline (Client-Side Performance Analysis)

        1. Performance Bottlenecks [4]
        2. Client-Side Performance Analysis [6]

        Client-Side Performance Analysis Erik Wilde UC Berkeley School of Information : Constraints of Mobile Applications

        (10) In-App Analytics

        • Analytics are useful for a variety of purposes
          • they allow to track user behavior and identify possible improvements
          • they allow to track application behavior and identify possible problems
          • they are important for all kinds of business-oriented analyses
          • Web pages embed code hosted by Google Analytics
          • scripting collects information which is collected by Google Analytics
          • Web masters can access Web-based tools for accessing the collected data

          Client-Side Performance Analysis Erik Wilde UC Berkeley School of Information : Constraints of Mobile Applications

          (11) Google Analytics

          

          Client-Side Performance Analysis Erik Wilde UC Berkeley School of Information : Constraints of Mobile Applications

          (12) Pinch Media

          

          Client-Side Performance Analysis Erik Wilde UC Berkeley School of Information : Constraints of Mobile Applications

          (13) Browser-Based Analysis

          • Web-based applications run in standardized implementations
            • users use the same resources regardless of the implementation
            • using browser-based debugging yields cross-platform results
            • is the scripting code distributed across many JavaScript files?
            • would it be possible to minify the JavaScript code?
            • is the CSS code distributed across many CSS files?
            • would it be possible to minify the CSS code?
            • can the images be compressed more effectively?
            • are components served with useful HTTP metadata?
            • are components served in a compressed format?
            • does the site use cookies?

            Client-Side Performance Analysis Erik Wilde UC Berkeley School of Information : Constraints of Mobile Applications

            (14) Browser Tools

            

            Client-Side Performance Analysis Erik Wilde UC Berkeley School of Information : Constraints of Mobile Applications

            (15) Server-Side Analysis

            • Less intrusive but more limited than In-App Analytics [In-App Analytics (1)]
              • all requests go to the server-side (unless there is a CDN)
              • server-side analysis is able to capture client requests
              • in the simplest scenario log files are written by the server
              • one entry per request with a configurable set of fields in the entry
              • can be very convenient for immediate analysis and query support
              • can be very risky because database writing tends to be slow
              • entries are added to simple text-based files
              • log files are rotated on a regular basis
              • completed log files are processed for traffic analysis