A VALUED CUSTOMER OF THIRDEYE DATA
Nimble Storage: predictive maintenance at ten times the data
Nimble Storage, acquired by Hewlett Packard Enterprise, operates InfoSight, a cloud-based predictive analytics service that watches flash storage deployed in server farms around the world and warns customers about failures before they happen. The value of a service like that rests entirely on how early the warning arrives, which means it rests on how much sensor data the platform can take in and how fast it can reason over it.
LOCATION
San Jose, California, United States
COMPANY SIZE
Mid-scale Enterprise
INDUSTRY
Information Technology › Explore solutions for this industry
Extending a predictive analytics platform to real time
InfoSight was already working when we arrived. The problem was a ceiling. The platform could only take in a fraction of the sensor data the installed base was producing, and the analytical queries that turned that data into a prediction were slow enough that the answers arrived after they were useful. More data would have made the predictions better and the queries slower. Doing one without the other was not an option.
ThirdEye Data rebuilt the ingestion and analytical layers so the platform could carry ten times the sensor data points it had been handling, and answer questions about them in near real time rather than in batches. The work covered how data enters the system, how it is held once it arrives, and how the analytical layer reads it, because a change at only one of those points would have moved the bottleneck rather than removed it.
The result was a platform that predicts against a far richer picture of what is actually happening in the field, and returns an answer while the answer still matters.
Ten times the sensor data points ingested. Ingestion rate improved by more than 50 percent. Analytical query response times improved by more than 900 percent.
Related solution: predictive maintenance at fleet scale
Predictive maintenance is easy to demonstrate on one machine and hard to run across a fleet. The difficulty is not the model. It is that the volume of telemetry grows with the installed base, the analytical questions get more expensive as the history deepens, and the value of the prediction collapses if it arrives late. Most platforms that stall do so because they were designed around the model rather than around the data path feeding it.
ThirdEye Data builds the whole path. How telemetry enters the system, how it is stored so it can be queried cheaply, how the analytical layer reads across it, and how a prediction reaches whoever acts on it. For companies whose product is the prediction itself, as it was here, that path is the product. We have done this work for storage hardware in server farms, for industrial equipment on production lines, and for physical assets inspected in the field.
DO YOU HAVE A SIMILAR USE CASE?
Talk to our team to learn more about what we did for Nimble Storage, and how that experience applies to your own environment.
Explore What We Did for Other Customers
