Tableau Server Two-Node Cluster Performance Optimization
Key Measurable Results
Reduction in extract refresh time (60+ min → 15 min)
Peak-hour server outages eliminated (down from 3/month)
Improvement in interactive dashboard load speed
1The Problem
The client's two-node Tableau Server cluster was suffering from 1-hour+ extract refresh bottlenecks, frequent peak-hour crashes, and slow dashboard loads because both nodes were running heavy Backgrounder jobs alongside interactive VizQL user traffic.
2What We Built
Audited the cluster and re-architected the two-node topology via TSM: dedicated Node 1 ("Head") exclusively to interactive user workloads (Gateway, VizQL, Repository) and isolated Node 2 ("Body") for heavy extract Backgrounders with local SSD File Store locality.
How It Works — Step by Step
Analyzed two-node resource utilization and identified CPU/RAM contention between VizQL and Backgrounder processes
Reconfigured Node 1 as the dedicated interactive "Head" node (Gateway, VizQL, Repository, zero Backgrounders)
Configured Node 2 as the heavy-lifting "Body" node (Data Engine, local SSD File Store, tuned Backgrounders)
Reduced Backgrounder concurrency from 8 to 4 on Node 2 to eliminate CPU thrashing
Colocated the File Store with Data Engine on Node 2 to remove cross-node network I/O bottlenecks
✓The Result
Reduction in extract refresh time (60+ min → 15 min)
Peak-hour server outages eliminated (down from 3/month)
Improvement in interactive dashboard load speed
Skills & Deliverables
Project Details
Our Role
BI Infrastructure Architect
Client
Enterprise Analytics Team
Industry
Data Engineering
Delivery Timeline
2 weeks
Need a similar system built for your team?
Book a free discovery call and we'll map out how to build this for your exact data sources and tools.
More Data Engineering Case Studies
View All
VideoSocial Media Data Stack: Airbyte, Fivetran, BigQuery & dbt
The client's analysts were pulling social media data by hand every week—logging into YouTube, Facebook, Instagram, TikTok, and other platforms one by one, exporting CSVs, and pasting numbers into reports. That took 8+ hours every week, and because each platform names and counts metrics differently, reports never matched.
View Case Study
Gifting Platform Cloud Data Warehouse (Snowflake, Fivetran & dbt)
A fast-growing corporate gifting SaaS platform had data scattered across its operational PostgreSQL database, legacy Snowflake tables, and third-party APIs. The executive team lacked visibility into key business metrics and had to ask product engineers to pull custom SQL reports for every question.
View Case Study
GMV & SaaS ARR Sales Performance Tracker
Sales leadership couldn't track real-time progress toward quarterly and monthly GMV and SaaS ARR targets. Pipeline forecasting lived in manual spreadsheets with no rep-level deal visibility.
View Case Study
VideoAI Invoice PDF to Excel Comparison Tool
The client received invoices from many different suppliers, and every supplier used a different PDF layout. To compare product unit prices or catch billing errors, someone had to type every line item into a spreadsheet by hand—which was slow and made price changes easy to miss.
View Case Study
VideoDiscord Data Bots: Fix Google Sheets Errors & Answer Data Questions
The team managed large operational datasets in Google Sheets while coordinating daily work in Discord. Finding mismatched rows in the sheets required tedious manual reviews, and whenever team members had a question about the data, they had to wait for a spreadsheet expert to look it up.
View Case Study
E-Commerce Data Architecture: 8-Platform BigQuery & dbt Stack
Data was siloed across 8 different platforms (WooCommerce, Meta Ads, TikTok Ads, Google Ads, GA4, Klaviyo, Stripe, and ERP inventory) with no way to see the complete customer journey or calculate true customer acquisition cost.
View Case Study