First, executives DNA is impressive with, all coming from famous companies they have helped to develop and grow for several years. This is a huge addition for the company as it enters in a new phase.
In fact, the story has started in 2013 with the creation of the open source platform for metrics and events. As of today 70,000+ active servers have been deployed representing 250+ customers and many famous names relies on InfluxData like Slack, Amex, Cisco, Comcast, BNP Paribas, SAP, Barclays or Nvidia. But what is it and why do we need such solutions? the team has insisted one 3 major changes that had an impact on the technology landscape:
- At the applications level, we came from large monolithic application model to a very distributed and granular one based today on microservices, that offers huge flexibility but also make things more complex.
- At the infrastructure level, we started with mainframe several decades ago to finally today rely on containers that make things super agile and also create some confusion and complexity just by the number of entities and elements you must control, manage an orchestrate.
- And last wave is IoT with dozens of billions of devices and sensors that contribute to the explosion of volume of data, analyst has projected 44ZB in 2020 so almost tomorrow.
But this approach has existed so what is really new with InfluxData ?
- New workload requirements for new high volume of real-time writes and a need for retention,
- Times-based queries with aggregation and summation time-based functions, ordering, ranking and limiting,
- Scalability and Availability with a distributed approach, fast and consistent.
Also InfluxData promotes a complete platform with 4 main components: Collect with Telegraf, Analyze with Chronograf and InfluxDB with an option for clustering, Monitor/Act with Kapacitor sustained with an ecosystem.
At the heart of the solution is InfluxDB, a database designed and dedicated for Time Series, and finally reflects a new category as the requirements can't be met by Oracle, Elastic or Hadoop. InfluxDB got some new design iterations since the beginning to offer more throughput and to reach new levels of availability. The database service has 2 layers, first for meta-data and this cluster is designed as CP system, and second for data with a AP model to support high volumes of write and reads. The metadata service is written in Go using the Raft consensus algorithm implemented by Hashicorp and BoltDB, a K/V store based on B+tree model. Globally data are sharded controlled by the metadata service. InfluxDB is finally a hybrid model providing HA and operate as an eventually consistent entity.
In term of go to market strategy, InfluxData wishes to be adopted and visible by developers and then upsell some enterprise functionality to make revenue such HA, security and scalability.
Super session, all the team has enjoyed the meeting.