Elastic{ON} 2018 - Cost Transparency at Credit Suisse

Slide 1

Slide 1 text

Credit Suisse February 2018 mathias.richter@credit-suisse.com Cost Transparency at Credit Suisse Mathias Richter, ITS Platform Metrics

Slide 2

Slide 2 text

2 Cost Transparency at Credit Suisse 1 What cost and why 2 What we do with Elastic – A fresh view 3 High-level Architecture – Rinse and Repeat

Slide 3

Slide 3 text

THE DATA IS THERE – WE JUST TAKE A FRESHVIEW

Slide 4

Slide 4 text

4 Who we are and what we do • We partner with our IT to evolve our trading platform in line with our business needs • IT cost is one of the key metrics we use to do this What Cost and Why? Cost Generator e.g. servers, routers, VMs, workforce Cost Aggregator Applications or Projects Business Cost Center

Slide 5

Slide 5 text

5 …but static reports „bundle“ a point of view and a specific point in time The Data Is There… How much am I paying now? How much are others paying ? What apps/projects am I paying for? What’s the total cost for an app or project and who else is paying for it? What infrastructure am I paying for? What’s the total cost for a given type of infrastructure and who else is paying for it? Cost Generator e.g. servers, routers, VMs, workforce Cost Aggregator Applications or Projects Business Cost Center Which organisations are charging me?

Slide 6

Slide 6 text

6 The Dashboard

Slide 7

Slide 7 text

7 High-level Architecture DB Connectivity REST APIs Extract Transform Enrich Elasticsearch Kibana X-pack X-pack Instances (1) Master Nodes (3) Ingest Nodes (0) Data Nodes - Hot (5) Data Nodes - Warm (0) Enterprise Systems Hadoop FS or other Bespoke Data Loader Scalable storage of timeseries data Ad-hoc Analytics Files

Slide 8

Slide 8 text

8 Identify data source 1 2 3 Extract, enrich and index Analyse on dashboard Rinse and Repeat Repeat the approach for any enterprise data • Massaging the data: Logstash, Scripted Fields or your own code? • Data formats (JSON is great, but…) • Get your «schema» right – index mappings, timestamps, data partitioning • Using the dashboards – train users • When a dashboard is not enough – R for data analysis • Visualisations: Metrics vs. aggregates on buckets • Keeping up with versions Some of our lessons learned