A Business Intelligence Platform is expected to evolve thru user adoption as BI consumers incorporate Business Intelligence into their daily roles. Ultimately, Business Intelligence has to drive effective business activity, gain business insight or achieve competitive advantage. A true catalyst to user adoption is actionable and quantifiable BI which demonstrates Return on Investment or Cost Savings. Achieving this kind of outcome is something I deeply believe in and have seen countless times in my 25+ year career as a practitioner.
Understanding departmental “entry points” when introducing BI tools into an organization is a viable perspective adopted by many organizations. It’s not always desirable to go “big bang”, to invest in lots of software licenses, to build elaborate data architectures, allocate large CAPEX and recurring OPEX requests, to invest in more staff, training, etc. Business Intelligence is a direct product of business need, of proven success which may start with 1 individual, 1 department, 1 functional area. I would be the first to admit that organically evolving something as complex as a BI program without a plan is like driving thru a strange city without GPS or a map, the fact remains that many organizations are just not ready to begin a BI Program at an enterprise level. You can’t. I always say “Think Big, but Start Small” – it is more cost effective. Don’t set your break-even so far out with a big initial investment and no plan for quantifiable return.
I am getting to my point with the specific topic – Scalability. If you visit the Tableau website (http://www.tableau.com) you will quickly find the trial or download area. Don’t think that some savvy or progressive analyst of business process manager hasn’t caught word of Tableau, and have downloaded product and have already started building scenarios against some of those Excel workflows. Whatever the case, it is quite possible that the Information Technology department has needed to be involved or as involved as they would have liked. Adoption happens, and this needs to scale.
Concerned? Don’t be.
Tableau can be configured several ways depending on data infrastructure, user load, usage profile and device strategy, and can be clustered with any # of machines. Common configurations include:
- Simple Configuration. Single server with a recommended HW configuration of 8 CPU cores AMD 32GB of main memory. The minimum requirement is useful for a POC for a larger deployment, or for a departmental server. Tableau recommends running two instances each of Data Server, App Server, VizQL Server and Backgrounder on a single server 8-core deployment of Tableau Server;
- 3-Server (24-Core Cluster). Supports heavier user loads will require clustering additional servers. The Gateway or Primary Server will the Backgrounder, Repository and Extract Host, and will send application server requests to worker machines. The other 2 worker servers will EACH have VizQL (1), App Server (2) and Data Server (1).
- 5-Server (40-Core Cluster). More worker machines can be added to a cluster to support heavier data usage or higher user load. In a larger cluster using data extracts. Repository and Extract Host on one machine, Backgrounders on another, VizQL (2), App Server (2), Data Server (1) on each of 3 Worker Servers;
- High Availability Cluster. Tableau’s High Availability Solution provides automatic failover for the repostory and data engine components. A minimum of 3 nodes is required – a Primary Node serving as the Gateway / Load Balancer and 2 additional nodes hosting the active processes. Gateway failover is a manual step. On failover, Tableau Server sends email alerts to specified Administrators;
- Other Considerations – Virtual Machines or Cloud-based Deployment. No special considerations when running on these environments. If running in the cloud, note that Tableau Server requires static IP Addresses;
Tableau can support large enterprises with 100s of 1000s of users. General Motors, Wells Fargo eBay and Bank of America are using Tableau. Since 2009, Tableau Server has been running at a high scale at the Tableau Data Center to support Tableau Public (free service). Tableau Public supports over 20m distinct users and reportedly serving 800,000 views per week. One event hit a record of over 94,000 views in one hour.
If you are reading this last line, I thank you for spending the time to read. Hope you’ve enjoyed the article, let us know! More to come.