Transcript
A little bit about myself
Chen Salomon, Data architect– c@Playbuzz.com
First developer @ Playbuzz
The guy that still wakes when shit hits the fan…
Currently leading (data) solutions
Playbuzz
3 years old
4 locations (Israel, NY, London, Hamburg)
100 employees
30 in RnD (we are still small…)
Seconds on a page
60% 78% 15
Source: Wall Street Journal, 2014
Changes in internet content
of internet users don’t
read entire articles
Web publishers
get
of traffic through
referrals from
Facebook on
average
User stays only
Playbuzz is the king of Engaging Content™
7,000+publishers can’t
be wrong
Formats
Gallery QuizTriviaPersonality QuizListRevealRanked List
Countdown ListPoll
Open List
Photo Caption
Video
THE SECRET SAUCE
• Keep it simple and super slim – We always try to keep our
production with as few servers as possible
• Be lazy – We pay a lot on insert and barely work on reads
• Know your traffic
• Enabling the business to scale at minimal cost
How we do it?
• Object caching
• Production usage
• Keeping items in cache (lazy initialization) – objects (strings)
• List management as sorted sets
• Permissions are hashmaps
Where you can find Redis in our system?
• Distributed HTML Output caching
• ETL session management and enrichments
• Session parameters are objects
• IP to location (in a few seconds)
Where you can find Redis in our system?
Most recent Redis project
IP to location
• Location objects are hash-maps
• Convert IP to integers
• e.g. IP 10.0.0.138 -> 10 *(256^3) + 0 *(256^2) + 0 *(256) + 138
*(256^0) =167772298
• ZRANGEBYSCORE to find the range (position) – returns the first item
that larger than the IP.
• HGETALL for the returned position
• Challenges
• IP V6 (same idea –different multiplier)
• Empty subnets – You can’t use only lower/upper bounds
We are hiring – jobs@playbuzz.com
top related