The Fifth Elephant 2019

We are a gaming company making mass market social games. Since being in a consumer market where user experience is the the key, we had to rely heavily on data from Day 1 of game/product launches. This is the reason we actually built our data infrastructure in parallel to games/products and had it ready for production usage from begining itself. We relied heavily on ready-to-use systems but at the same time had to be cost sensitive being a startup. Setting up whole data-lake and heavy duty hdfs cluster was ruled out due to cost and maintenance overhead. We setup a lightweight data collection pipline to central queues which is then ingested in realtime to our warehouse of choice Redshift (reason being ease-of-use). Also, scaling such a system has its cost overheads when your product grows. So we had to design data retention and data querying capabilities such that we aren’t paying hefty bills as well as aren’t being limited in terms of querying real-time data from our users.

Outline

1. Be clear of Requirements and Constraints

• Having a scalable system for data ingestion
• Data design (Specific or Generic)
• Querying interface - why stick to SQL?

2. Take time to Design Data

• Walking through example of generic table design

3. Sort out Data production part first

• Identify all possible data producers (and understand requirements). In our case -
• Android/iOS app
• Cannot keep sending each event over network
• Cannot lose data even if app crashes or is killed
• Keep out of context from the application itself
• Microservice(s)
• Cannot keep sending each event over network
• Keep data collection agnostic of microservice itself

4. Design v1.0 of Data pipeline

• How and why we chose “anti-pattern”

5. Choose/Design Data warehouse

• Data design in Redshift
• Compression ON for certain columns
• Tuning for scale
• Taking care of Querying patterns of Product Managers and Data scientists

6. Open up: Enable many Data Interfaces

• On demand Data loading and querying: OnDemand Table(s)
• Flexibility for complicated analysis: Adhoc redshift cluster(s)

7. Understand, Tune & Repeat
8. Optimize for Usage

• Added more columns at generic level e.g.
• More examples

9. Optimize for Cost & Ops

• Retention policies of data
• Not all events are of same importance
• But all events should be accessible if required

10. Upgrade to v2.0 of Data pipeline

Speaker bio

I am Kumar Puspesh, CTO and Co-Founder of Moonfrog, India’s top mobile gaming company. We had to design a large scale data infrastructre from day 1 of our company to cater to our product needs. Having a cost sensitive as well as scalable approach helped us achieve large scale as a gaming company in India in short amount of time. At the same time taught us a lot of ingenious ways of building large scale infra customized for business and its users (rather than a generic paid solution and then changing your usage/requirements based on that).

Slides

https://docs.google.com/presentation/d/1qYkGQLzK8UO-f59TFgkj1GROAmFLkotoQWMZNYXJbFQ/edit#slide=id.g4a9ee349ba_2_75