2. Who we are?
Founded in late 2013, tixCraft.com is the most advanced
could-based ticketing system in Taiwan.
- to the FANS
- to the PROMOTERS
- to the VENUES
tixCraft is the latest venture from
ticketing and promotions specialists
who are in the industry more than
a decade.
3. What have we done?
2008 – IBAF Final Olympic Qualifying Tournament
2009 – NBA Taipei Games, The World Games
2010 – MLB Taiwan Games, Golden Horse Film Festival
2012 – Jacky Cheung 1/2 Century Tour, Girls Generation Tour
2013 – Mayday, Super Junior, G-Dragon, NBA Taipei Games
2014 – YG Family Galaxy Tour, aMEI|AMIT Utopia World Tour
2015 – BigBang World Tour [MADE], ONE OK ROCK “35xxxv”
Asia Tour, WBSC 2015 Premier 12, Golden Horse Film Festival,
Sodagreen Endless Story
4. What do we want to do?
Situation: we failed every time when selling hot tickets.
Goal: selling tickets smoothly on internet.
Possible solution: large-scale multi-tier system, only
possible by using cloud service
5. Why AWS?
Evaluation: stability, easy to use, scalability, billing method
Target: ISPs in Taiwan, AWS, and other cloud services
AWS: the most mature, stable, user-friendly and billing by
hours. Almost no differences where we host.
6. When we started it, we found…
Much, much easier and faster for installation new virtual
servers than physical ones.
Much cheaper for building complex structure when using
small instances at beginning.
Handling services easily like load balancing, moving to
RDS for MySQL, creating Read Replicas.
7. Stress testing and architecture adjustment
Starting with 10,000 t2.micro instances, it costs only
$0.013*10,000=$130 for one hour.
Write simple scripts to simulate users behaviors.
It’s much easier to redesign the whole system when
failed.
Relatively simple to rebuild the new system, and test it all
over again.
10. Benefits with AWS
Lower cost for building larger structure.
Highly scalable, if you build it right.
Agility – you can fail faster, fail cheaper, fail forward
Helping us optimizing and lowering cost. Ex. RDS-
>DynamoDB, EC2->Spot/Block Spot.
11. By moving to AWS, we can
1) lower time of deployment: from 1,600 to 20 minutes
2) increase more than 130 times of resources in 30 minutes
3) handle 100,000+ concurrent users selecting seats
4) handle 70,000+ ordering requests for tickets per minute, max
2,500+ ordering requests per second
5) have more than 30,000 tickets sold out in 3 minutes
6) pay less than 2 MacBook Pro to accomplish this mission
7) reduce our server cost to 0.26%
8) Accept IOPS from 20 to 135,000 (average number in every minute)
by using DynamoDB easily
1) have much lower time of deployment: from 1,600 man-hours[2(man-hours)*800(servers)] to 20 minutes
2) increase more than 130 times[from 6 to 800] of resources in 30 minutes
3) handle 100,000 concurrent users selecting seats at the same minute
4) handle 70,000 ordering requests for tickets per minute, max 2,500 ordering requests per second
5) have more than 30,000 tickets sold out in 3 minutes
6) pay less than 2 iMacs to accomplish this mission (4,200 USD)
7) reduce our server cost to 0.26% - 4,200(this one-time event)/800*2,000(server price)
8) We're able to accept R/W IOPS from 20 to 135,000 (average number in every minute) by using DynamoDB easily( normal time IOPS consumed: 20, peak time IOPS on 2015/8/29: 135,000)