Data has become a new competitive advantage, and business decisions are expected to be backed by data more than ever. This comes with an increased demand for talented big data professionals, and companies are aggressively investing in growing their analytics teams to meet this need. But even with all of this talent, in today’s competitive, fast paced environment, analytics teams must also be able to move in an agile manner.
To scale effectively and quickly generate actionable business insights, users, and their organizations, must adopt self-serve analytics. In recent years, AI-powered solutions have emerged as an effective way to achieve self-serve analytics.
An example of this is what we're doing here at Kyligence. Kyligence has leveraged AI to create a powerful self-serve data platform. Before diving into any specific AI capabilities, let’s talk about Kyligence’s core technology.
The core driver of Kyligence’s high performance is its pre-compute technology. Specifically, OLAP technology. OLAP is a proven technology and has been given a new life with modern advances in compute and storage. Because the data is pre-computed, Kyligence is able deliver sub-second performance at peta-byte scale with hundreds of concurrent users.
Pre-compute technology is the ideal solution to today's big data challenges. The more data and the more concurrent users, the higher the return of pre-compute technology. This is because pre-compute technology follows economies of scale.
The reason being, once a model is created, it does not matter if 10 or 1000 users are hitting the model; it will return the same performance. The data is pre-computed in the model, so the 10 or 1000 people are simply “grabbing” the data.
There is no on-the-fly compute. This leads to the question, is the model static? No, Kyligence supports both batch and streaming processing. New data is continuously added to the models and automatically exposed to the user when ready.
An Intelligent Engine
Traditionally, pre-compute technology requires a high level of technical knowledge. Someone with specific knowledge must build the pre-computed model. However, with Kyligence’s AI-powered capabilities, all users can create the perfect model. The user can simply upload a SQL script, and the engine will create the model automatically. The video below provides a great example of this in action.
People with experience using pre-compute technology also know that models require tuning over time. Before, users would need to know about the tuning parameters and have other specific knowledge to optimize the model. With Kyligence’s smart engine, the engine will learn from users’ past behaviors and automatically tune the model.
From model creation to model optimization, the engine does all the work in the background. The user reaps the benefits without ever having to know the details of how it all works. For example, the engine is smart enough to know whether it needs to build a new model or simply modify an existing model.
Sometimes, it is more resourceful to build on top of an existing model rather than build a new model, in that case, the engine will do just that. Data requirements change over time, if a model has been latent for an extended period of time (no one is querying the data), the user will be asked if they want to purge the model. In another example, let’s say only 1/10 of a model is frequently being used, the model will automatically be reduced to reflect that usage.
The engine intelligently adjusts based on two factors: maximizing performance and saving resources. The engine increases or decreases compute resources elastically. For example, if the system cost is very high and not busy, it will automatically choose to use “cheap nodes” to reduce costs. Contrarily, should a cluster load become very high, the system will automatically add query nodes to expand resources to increase performance.
The only thing the user needs to do is set the optimization parameters and the engine will execute based on those directions. There is an interface which shows optimization recommendations and acceleration choices. The user can choose to exert as little or as much control as they want. The user has complete control of the settings.
Better Over Time
Kyligence’s AI-powered engine is constantly learning, the more behavioral data it logs, the smarter it becomes. Kyligence’s auto modeling and self-optimization capabilities remove a major pain point, analysts no longer need to wait on engineers to fulfill their data requests.
Any user with SQL knowledge is empowered to retrieve data to slice and dice as they wish. Users can also consistently expect their query response times to be under a second. The entire process is made frictionless. Analysts will never be blocked by data retrieval or query latency issues. With more control and a reliable working environment, analysts can simply focus on delivering business insights.