Learn more about why LeanXcale is the best database for monitoring applications in this series of blog posts!
LeanXcale for monitoring tools
Within IT systems, monitoring tools have become popular, covering a wide spectrum from network equipment and processes to the experience and behavior of web application users, and including technical monitoring of applications.
Monitoring tools help us to know the current state, whether it is normal or exceptional. They allow us to know the causes that led us to it and even allow us to forecast the future based on data.
All these tools have similar needs that mean a set of technological challenges:
- High insertion rates
- Real-time KPI calculation
- Insertion/Recovery of historical time series
- Scaling
A monitoring platform using LeanXcale
1) High ingestion rates: Monitoring tools such are continuously receiving information from agents, probes and applications. And the more time required for ingestion, the fewer elements can be monitored per collector. KiVi is LeanXcale's relational key-value store engine that was designed from the ground up to deliver an optimal ingestion rate based on a NUMA efficient architecture, a new hybrid data structure, as well as bi-dimensional partitioning that enables more efficient ingestion of time series data.
2) Real-time KPI calculation: Most KPIs are based on aggregation operations. As a disruptive innovation, online aggregations in LeanXcale prevent parallel query conflicts and allows for the costless computation of aggregated KPIs as data is ingested. The computation of any switching operation is done at the moment of insertion, in real time, and without generating any kind of contention or conflict.
3) Time series recovery: To optimize time series management, LeanXcale introduces two innovations: a new
data structure that is as efficient as both B+ trees for a broad range of queries and LSM trees for updates and random insertions; and bi-dimensional partitioning, which divides data in such a way that memory is used efficiently, prevents continuous I/O access, and increases the location of information in subsequent searches .
4) ACID linear scalability: Generally, the scalability of monitoring platforms is limited by the underlying databases. With its patented Iguazu algorithm, that allows scaling from 1 to 100s of nodes with each node delivering the same performance, LeanXcale allows processing at any scale with a single cluster, which is capable of storing any size of data and requests.
5) Fully elastic: A cluster in LeanXcale can grow or shrink instantly to exactly the size necessary. Overprovisioning to respond to demand peaks is never needed, and, instead, customers simply use and pay for what is needed at every instant.
6) Any type of query: LeanXcale supports queries for all operational or analytical applications, including dashboarding with any BI tool through JDBC or ODBC, as well as machine learning through Python's SQL Alchemy or Spark.