While Elasticsearch can be an attractive solution for search/filtering/sorting needs, we found the maintenance costs outweighed the benefits. That’s why we opted for a Postgres solution. In this talk, I will outline our situation and business requirements, briefly explore our decision-making process and elaborate on how we determined a Postgres solution would adequately serve our needs without creating unnecessary costs.
This presentation was given during the the Spaces Summit, an internal IT conference by and for the engineers of bol.com.
6. 6
Current Search on Orders
Things you can search for:
467144359
790356082
734724846
890267385
153858964
857894655
105735956
937485025
183649573
265037950
729473560
385620718
275936590
274650374
104836395
690376403
306429853
844377449
The possibilities are endless!
14. ❖ Our use case
❖ Postgres Research and Results
❖ Looking ahead
14
Agenda
15. 15
Test Data Set
❖ 50 Million orders in total
❖ 50k open orders
❖ 5 million (10%) LvB
❖ 45 million (90%) LvR
❖ 2000 sellers
❖ 10,000 products
❖ Each seller had about 25,000 orders
❖ REMEMBER: “average” seller has ~1,000/year