Doug Sillars on App Optimization

Test Your Mobile Application’s Performance
with AT&T’s Application Resource Optimizer
Doug Sillars, PhD
Application Optimization Guru
Developer Advocacy, AT&T
@dougsillars

How Fast is Your App?
“Zombies appear to be
incapable of running.
The fastest have been
observed to move at a
rate of barely one step
per 1.5 seconds.”
– Max Brooks,
Zombie Survival Guide

How Important is Speed Anyway?
Traditional Web (PC) Stats:
• Yahoo! 400ms delay leads to 5-9% drop in traffic (1)
• 1 second of latency (2)
–  Customer satisfaction drops 16%
–  Conversion to sales drops 7%
• 10 Golden Principles of Successful Web Apps:
–  “Speed is always the most important feature” (3)
Sillars’ Corollary: If speed is important on the web, it
follows that speed is even more important for mobile
• 71 % of users expect mobile sites to be as fast as desktop (4)
(1) slideshare.net/stoyan/dont-make-me-wait-or-building-highperformance-web-applications
(2) http://econsultancy.com/us/blog/10195-the-need-for-online-speed-in-america
(3) 10 Golden Principles Of Successful Web Apps http://www.businessinsider.com/10-golden-principles-successful-web-apps-2010-3
(4) http://econsultancy.com/us/blog/9162-the-importance-of-speed-for-mobile-commerce

What Do Mobile Users Expect?
High Performance Apps
•  Speed
–  64% of mobile users expect pages/apps to load in <4 seconds (1)
•  Battery Life
–  Better Battery life (6.1 satisfied of 10) !!! (2)
–  72% rate phone as very good or excellent
•  Only 21% rate battery life as very good or excellent (3)
•  Reasonable data usage
–  No one wants to hit their cap
(1) http://www.keynote.com/docs/reports/Keynote-2012-Mobile-User-Survey.pdf
(2) http://www.businessnewsdaily.com/2200-smartphone-customer-satisfaction-battery-life.html
(3) Personal communication from Swiftkey study

What Do Mobile Users Get?
Our Customers Expect More!!!!
• Top Mobile frustrations “web page slow to load” (1)
• Battery life top reported problem (2)
(1) http://www.keynote.com/docs/reports/Keynote-2012-Mobile-User-Survey.pdf
(2) http://www.businessnewsdaily.com/2200-smartphone-customer-satisfaction-battery-life.html

As a Developer, What Can I Do?
• Learn How Mobile is Different
• Optimize:
–  Learn the best practices
–  Test your application
–  Fix issues before your customers discover them
• Get ahead of the game – be ready for industry standards
and grading systems
It can’t be MY developers, they’re the best
• AT&T’s Application Resource Optimizer (ARO)
–  http://Developer.att.com/ARO
–  In 2012, <2% of tested apps required no optimization

How is Mobile Different?
Internet
IP applications
Mobile applications
Wireless
State
Machine
Mobile apps often do not consider the uniqueness of the wireless network
Optimized mobile applications improve user experience with
improved battery life and perceived faster speed

Data Transfer
LTE RRC State Machine
Timer Expiration
Radio Off Radio On
IDLE
Continuous
Reception
Tail
For AT&T’s Model:
T1 = 100ms
T2 = 20ms
T3 = 10,00ms
100ms
10s

Data Transfer
LTE RRC State Machine
Timer Expiration
Radio Off Radio On
IDLE
Continuous
Reception
Tail
• LTE has higher throughput than 3G
• This does not improve application behaviour
-  The higher power actually accentuates any issues your application might have!
100ms
10s

It’s All Connected
Application traffic triggers RRC State transitions:
•  RRC State transitions control

–  Speed/User Experience

–  Device battery life

–  Device battery life
–  Network efficiency

AT&T Application Resource Optimizer
• Free diagnostic tool
• Open Source
• Test any Platform
• Emulate any carrier network
• Analysis of both Native Apps and the Mobile Web
• Winner of the 2013 GSMA Smartphone Application Challenge award at
MWC
Allows YOU as a developer to SEE exactly how data
is flowing from your app!

What Platforms Can I Test With?
All of them!
• All Platforms: Pcap/tcpdump network trace
•  Ex: Wi-Fi Hotspot (like Connectify) and Wireshark/NetMon to collect
•  iOS: Remote Virtual Interface uses Instruments to collect pcap over 3G/LTE
• Android and Windows 8: Native Collector
•  Note: Android collector requires root

How Does ARO Work?
Transfer Trace Files Process Trace
Test Your Application
• PCAP only:
•  Network traffic collected
• Native Collector adds:
•  Video
•  User Input
•  Peripheral Usage
•  CPU usage
• Applies Radio and Power models
to data
• Graded against 15 Best Practices
• Learn how your application
ACTUALLY behaves!

Everyone Wins With ARO
•  Faster Response Times
•  Improved Battery Life
•  Efficient Data Plan Usage
•  Enriched User Experience
End User
•  App-specific Analysis
•  Key Areas to Improve
•  Cross Platform and
Network Agnostic
•  Free and Open Source
Developer
•  Increased Network
Availability
•  Signaling Load Reduction
•  Efficient Spectrum Usage
•  Improved Radio
Availability
Wireless Carrier

What’s the Industry Doing?
• Quality App Directory - App Quality Alliance (AQuA)
(http://www.appqualityalliance.org/)
• Smarter App Challenge – GSMA
(http://mobileappchallenge.com/smarterapp/)
• Mobile Content Action Team – CTIA
(http://www.ctia.org/business_resources/wic/index.cfm/AID/10333)
• Other carriers
• All are using ARO!! (http://developer.att.com/ARO)
• Numerous Customer facing apps help customers preserve battery and
reduce data usage
Get ahead of the game: Optimize NOW!

So, How Does ARO Make My App Faster?
ARO is the test kit to identify areas where performance
improvements exist
Simple, common sense development best practices
– Reducing connection times
– Caching files
– Eliminating Errors
The fixes identified by ARO will tune your application
to higher performance and speed

From Zombie to Sprinter in 3 Easy Steps
1. Close Connections
2. Cache Your Data
3. Manage Every Connection

Closing Connections? How Does That Help?
>80% of applications
do NOT close
connections when
they are finished!

Closing Connections: Example
38% more power on LTE!
(18% more power on 3G)

Closing Connections: CODE
MultiRes Sample app from Android SDK
– Modified to download images
HttpURLConnection getimagecloseconn = (HttpURLConnection)
urln.openConnection();
getimagecloseconn.setRequestProperty("connection", "close");
getimagecloseconn.connect();
String cachecontrol = getimagecloseconn.getHeaderField("Cache-Control");
InputStream isclose = getimagecloseconn.getInputStream();
bitmap = BitmapFactory.decodeStream(isclose);
getimagecloseconn.disconnect();
https://github.com/attdevsupport/ARO/tree/master/2013DevSummitTurbocharge

Caching Your Data
17% of all mobile traffic is duplicate download of
the same unaltered HTTP content (1)
“It’s just a 6 KB logo”
– 6 KB * 3 DL/session *10,000 users/day =
3.4GB/month
Reading from local cache is 75-99% faster than downloading
from the web
(1)“Web Caching on Smartphones: Ideal vs. Reality”, http://www.research.att.com/~sen/pub/Caching_mobisys12.pdf

Even if caching IS supported – it is OFF by default!!
Caching Support by Library
Android Connection Libraries iOS Libraries Android & Safari
Test Name UC HUC HC WV HRC T20 NSUR ASIHR AB SB
Basic caching ™ ™ ™ ˜ ˜ ™ ˜
Revalidation ™ ™ ™ ˜ ˜ ™ ˜ ˜ ˜ ˜
Non-caching directives ™ ™ ™ ˜ ˜ ™ ˜ » ˜ ˜
Expiration directives ™ ™ ™ ˜ ˜ ™ ˜ ˜ ˜ ˜
URL with query string ™ ™ ™ ˜ ˜ ™ ˜ ˜ ˜ ˜
Partial caching ™ ™ ™ ™ ™ ™ ™ ™ ™ ™
Redirection ™ ™ ™ ™ ™ ™ ™

Caching Methods (How do I do it?)
ETags
Cache Control Headers
• Each file has a Unique Tag
• Revalidated on server for each request
– High Performance Web Sites:
Rule 1 – Make Fewer HTTP Requests (1)
– Adding a connection drains battery,
adds 500-3,000 ms latency
• Important to carefully assign Max-Age times
• App will not check file on server until Max-Age
is reached
– Retrieval is strictly file processing time
(1) http://developer.yahoo.com/blogs/ydn/posts/2007/04/rule_1_make_few/

Comparing ETag vs. Cache Control
Network Usage Battery Drain Data Usage Speed
NO Cache
YES entire file
downloaded
Full network
connection made
High Slowest
ETag
Yes – ETag in cache
is validated on server
Full network
connection
Low Slow
Cache Control No N/A N/A Fast
Any Caching is better than no caching, but for mobile, Cache Control is preferred

Caching: Worth the Effort?
public void onCreate(Bundle savedInstanceState) {
super.onCreate(savedInstanceState);
setContentView(R.layout.main);
//establish a cache
try {
File httpCacheDir = new File(getCacheDir(), "http");
long httpCacheSize = 10 * 1024 * 1024; // 10 MiB
HttpResponseCache.install(httpCacheDir, httpCacheSize);
//
}
catch (IOException e) {
Log.i(TAG, "HTTP response cache installation failed:" + e);
}
Add this!
Android 4.0:
Don’t leave older devices in the cold: Consider adding reflection for
older versions of Android
http://developer.android.com/reference/android/net/http/HttpResponseCache.html

Caching Performance
File still valid in cache
Radio never turns on
100% less
power!
MultiRes sample app – modified to add caching
98% faster!

Caching Performance
File in cache has expired,
revalidate at server
If radio has gone back to
IDLE, this can take 2.5s
LTE: 14%
less power!
MultiRes sample app – modified to add caching
3G: 30%
less power!
30% faster!
89% faster!

Grouping Connections
38J of energy used!!
Imagine an app that:
1. Downloads an image every 60s
2. Downloads an Ad every 60s
3. Sends Analytics to a Server every 60s

Grouping Connections
16J of energy used!!
58% savings!
Now, look what happens if your connections are grouped together
1. Downloads an image every 60s
2. Downloads an Ad every 60s
3. Sends Analytics to a Server every 60s
16J of energy used!! 58% less energy

How Do I Group Connections
if (Tel.getDataActivity() >0){
if (Tel.getDataActivity() <4){
//ok, we are passed the minimum time to check
//and we found network activity-
//download the image here using image getter
imagegetter(counter, numberofimages);
//and show the ad
AdRequest adRequest = new AdRequest();
adRequest.addTestDevice(AdRequest.TEST_EMULATOR);
adView.loadAd(adRequest);
// Initiate a generic request to load it with an ad
adView.loadAd(new AdRequest());

Grouping Connections for Speed
Not only can you save battery, but you can speed up your
application by managing connections properly:
Threading file downloads vs. serial download
Removing redirects to files
Pre-fetching files that are used often

Threading file downloads vs. serial download

Removing redirects to files
Redirection adds ~2-3 seconds for each request
{
2 sec

Pre-fetching files that are used often

More Best Practices
Look for error codes: no 4xx 5xx
HTTP response codes should occur:

More Best Practices
Look for error codes: 404s
Error messages should be short and sweet

Ad Download every 30s
More Best Practices:
Periodic Connections
Regular 3 minute Polls for updates (20% battery/day)

Summary
• Our customers expect high performance
• App performance optimizations are easy…
Once you know the issues
• Best Practices:
–  Close Connections
–  Cache Data
–  Group Connections
• Fewer connections
• Threaded Connections
• No redirects
• 4XX HTTP response codes
• Periodic Connections

Developer.att.com/ARO
Developer.program@att.com
@dougsillars
thank you!

Doug Sillars on App Optimization

More Related Content

Similar to Doug Sillars on App Optimization (20)

More from wipjam (8)

Recently uploaded (20)

Doug Sillars on App Optimization