UPD Graduate Talk 2013

Transcript

1. Save22 @ UPD 2013 Ronald Cheung CTO Tuesday, February 25, 14
2. What is Save22 • Save22 is a Singaporean startup company developing a web/mobile platform that helps Southeast Asian shoppers make purchasing decisions on what and where to buy. • There is a general lack of accurate pricing data in SE Asia, this is the problem Save22 is in the business of solving for the consumer Tuesday, February 25, 14
3. Platforms • IPhone • Android • Web Tuesday, February 25, 14
4. Company Milestones • Company founded in March 2010 • Launched iPhone app August 2011. 30000+ downloads in Singapore • Raised initial funding November 2011 • Jan 2012 transferred engineering operations from Singapore to Manila • Release of Android app September 2011 • Release of redesigned web platform December 2012 • We currently have a team of 10 working from Ortigas Tuesday, February 25, 14
5. Comparison of Price Comparison Apps • US based Price Comparison Apps such as Red Laser and Shop Savvy all source data from other e-Commerce platforms • Save22 price listings are from traditional physical stores such as SM Tuesday, February 25, 14
6. Technology at Save22 • We use a diverse technology stack to support our web/mobile platform Tuesday, February 25, 14
7. Mobile • We use native implementations for both Android and iPhone ( Java and Objective-C) • We have requirements such as barcode scanning using the phone’s camera that require native implementation • Clients get data from our backend via a Web Service API Tuesday, February 25, 14
8. Web / Backend • Python used as primary language • Django used as a web framework • Apache / mod_wsgi used as web server • Postgresql used as relational database • Amazon Web Services ( AWS ) used for cloud hosting • Ubuntu OS • Memcached used for general purpose caching • RabbitMQ / celeryd used for queuing • Tornado used to support Long Polling applications • Fabric used for deployment • GIT used for source control Tuesday, February 25, 14
9. “Traditional” enterprise stack • Java used as primary language • SQL Server / Oracle used as relational database • Physical server • Windows / Unix OS Tuesday, February 25, 14
10. Python • Interpreted rather than compiled Faster speed of development but performance can be slower than compiled code • Expressive Power: More functions with less code Hello World in Java: public class HelloWorld { public static void main(String[] args) { System.out.println(“Hello, World”); } } Hello World in Python: print “Hello, World” Tuesday, February 25, 14
11. Python • Positional and Keyword arguments for function definition Easy to change function definition without extensive refactoring def cheeseshop(kind, *arguments, **keywords): print "-- Do you have any", kind, "?" print "-- I'm sorry, we're all out of", kind for arg in arguments: print arg print "-" * 40 keys = sorted(keywords.keys()) for kw in keys: print kw, ":", keywords[kw] cheeseshop("Limburger", "It's very runny, sir.", "It's really very, VERY runny, sir.", shopkeeper='Michael Palin', client="John Cleese", sketch="Cheese Shop Sketch") Output: -- Do you have any Limburger ? -- I'm sorry, we're all out of Limburger It's very runny, sir. It's really very, VERY runny, sir. ---------------------------------------- client : John Cleese shopkeeper : Michael Palin sketch : Cheese Shop Sketch Tuesday, February 25, 14
12. Python • List Comprehension • Built in data structures: Dictionary Tuples and Sequences Sets Lists • Has a wide array of third party libraries Using a for loop: >>> squares = [] >>> for x in range(10): ... squares.append(x**2) ... >>> squares [0, 1, 4, 9, 16, 25, 36, 49, 64, 81] List comprehension: squares = [x**2 for x in range(10)] Instantiating a Python Dictionary: >>> tel = {'jack': 4098, 'sape': 4139} >>> tel['guido'] = 4127 >>> tel {'sape': 4139, 'guido': 4127, 'jack': 4098} Tuesday, February 25, 14
13. Python • Python resources: “Dive into Python” “Learn Python the hard way” • Trivia: The largest user of python in the world is Google Tuesday, February 25, 14
14. Django • Django is a Python web framework • Web frameworks provides a single pattern to implement common web functions such as defining forms, mapping form input parameters, mapping urls to functions. • Avoids multiple ways of doing the same thing, any Django developer can quickly understand the code without extensive familiarization. Tuesday, February 25, 14
15. Django • Object-relational mapper (ORM) Avoids direct interaction with the database layer, provides a way to map domain objects to database structures • URL design A regular expression based mechanism to map URL patterns to functions • Template system Provides a templating language to separate presentation layer logic from business logic from data access ( MVC design pattern ) • Out of the box admin interface Provides default mechanism for users to update, add, delete from tables, avoids building repetitive function • Internationalization Built in support for multi language applications Tuesday, February 25, 14
16. Django • Alternative Python web frameworks: Flask Pylons Grok TurboGears web2py At this point in time Django is still probably the most comprehensive framework • Resources: https://www.djangoproject.com/ Tuesday, February 25, 14
17. Postgresql • Postgresql is an open source DBMS can scale on the same level as SQL Server or Oracle. • Important to handle the data required to build up an index of products, prices, and stores in SE Asia! • Supports the PostGIS library that adds Geographical data to a postgresql relational database. Highly useful to support functions like: - Where is the closest place a user can buy the product they are looking for - Where is the cheapest place I can buy a product within a given polygon ( for example Quezon City is a polygon ) • PostGIS has a convenient Django wrapper ( GeoDjango ) Tuesday, February 25, 14
18. Cloud computing • Physical infrastructure requires a high upfront set up costs including: Purchase of hardware, lease/setup a data center, network connectivity, build in redundancy, hardware end of life and replacement, firmware upgrades, security, hardware firewalls etc • Cloud computing has low upfront cost, pay on a per hour basis, don’t need to worry about hardware maintenance, can scale computing capacity easily Tuesday, February 25, 14
19. Amazon Web Services • AWS is a set of cloud computing services provided by Amazon.com • Elastic Compute Cloud (EC2) allows users create, launch, terminate “virtual machines”. For practical purposes they function like corresponding linux or windows machines. • Different EC2 “instances types” can be started with varying amounts of memory and computing power. • S3 is a service that provides storage at a relatively low cost per GB. • We use EC2 instances to host our Web and App servers, as well as database servers • We use S3 for backups and static media hosting ( for example product images ) Tuesday, February 25, 14
20. Memcached • Memcached is a general purpose key value pair caching mechanism. • Is supported out of the box by Django as a caching backend • We use memcached for query caching to minimize heavy queries on the database when generating frequently hit pages ( for example homepage, categories page ) Tuesday, February 25, 14
21. RabbitMQ / Celery • RabbitMQ is an Open source messaging infrastructure • Celery is an asynchronous queuing mechanism works with RabbitMQ Written in Python! • We use RabbitMQ/Celery for job scheduling (Celery Beat), running batch jobs asynchronously. Tuesday, February 25, 14
22. Tornado Web Server • An open source non-blocking web server originally developed by FriendFeed written in Python • Tornado is event based that makes it distinct from web server such as Apache that are thread based. • It was initially developed to overcome the “C10K problem” of being able to handle 10000+ concurrent socket connections. Threaded models begin to suffer at this point due to context switching between this number of threads becoming a large overhead. • Handles “long polling” style applications such as news feeds that are setup to refresh themselves at certain intervals. • We use Tornado to serve the “news feed” that is constantly being refreshed and requested by our mobile users. Tuesday, February 25, 14
23. Fabric Deployments • Traditional deployment to servers involving SSH to each server and running commands. • Fabric is a tool that streamlines this process by running shell commands from a single terminal to a set of remote servers. • Fabric is also written in Python! Tuesday, February 25, 14
24. GIT • We migrated from SVN to GIT for one overriding reason: git merge Tuesday, February 25, 14
25. The Team • Product Manager / UX • Designer /UI • Industrial Engineer / Process • Testing / QA • Software Engineers x 3 • Software Engineers are expected to be generalists who may work on any technologies in the Save22 technology stack Tuesday, February 25, 14
26. Questions • ronald@save22.com • http://twitter.com/ronaldlcheung • http://www.linkedin.com/in/ronaldlcheung Tuesday, February 25, 14