Introduction
PropFinder aggregates real estate listings from 50+ sources. Their existing MySQL search was slow (2 seconds). This case study covers their Elasticsearch migration on Hostxpeed.
The Search Problem
MySQL full‑text search on 5 million rows took 1.8s average. Complex filters (price, bedrooms, location) made it worse. User bounce rate was 65% on search results page.
Hostxpeed Elasticsearch Cluster
5 x NVME-3 VPS ($49.99 each) forming Elasticsearch cluster (3 data nodes, 2 master/coordinate). 16GB RAM per node, total heap 12GB each. Indexes sharded across nodes.
Performance Results
Search time improved from 1.8s to 78ms (23x faster). Bounce rate dropped to 31%. They can now support fuzzy search, geo‑distance, and autocomplete.
Costs
Elastic Cloud would cost $1,500/month for similar cluster. Hostxpeed solution: $250/month + free support. Savings $1,250/month (83%).
Operational Notes
Used curator for index management. Hostxpeed support helped tune kernel parameters for high file descriptor usage (needed 65535). Snapshot backups to object storage.
Conclusion
PropFinder now serves 2 million search queries monthly. Elasticsearch on Hostxpeed exceeded performance expectations at fraction of managed service cost.