DiskProphet® is a purpose-built disk failure prediction solution. It applies state-of-art machine learning technology for alerting disks replacement before they are failed..
Most of the other disk monitoring tools are usually applications running on a computer for a few disks. DiskProphet® is the first software of this category that can scale-out by only adding new machines to predict growing amount of disks.
DiskProphet® is not only an added protection to various redundant technologies to data protection, such as RAID, multiple-copy, etc. It can help data centers take an entirely different approach to operate their storage. With DiskProphet’s high failure prediction accuracy, DiskProphet helps a data center spend less money on data redundancy. A disk is replaced before it becomes failed and storage spends less time on data rebuilding. The data center can provide higher service reliability and doesn’t have to preserve extra I/O for data rebuild.
Most of the modern data protection methods, such as RAID, backup, or 3 copies, need data rebuild when disks fail. During the data building, there are problems of degraded I/O performance and a higher risk of data loss when more disks fail. Please see the diagram below.
By adopting DiskProphet®, the aforementioned problems are gone. DiskProphet provides disk replacement plan based on the disk failure prediction and storage I/O prediction. The replacement plan informs the timing of lowest predictive IOPS and recommends replacing near-failure disks at the time. Therefore, DiskProphet minimizes the disk failure impact to the workload. Please see the diagram below.
DiskProphet® has no limit on the number of disks it can predict. DiskProphet® prediction capacity depends on the server and storage a customer can provide. DiskProphet® can be deployed on a single machine and scale out to multi-machine for large-scale disk prediction.
DiskProphet® can be installed on both physical machines and virtual machines. One server is the minimum requirement.
Some types of storage inform users after disk fail. Other types of storage provide disk failure prediction. Only DiskProphet’s predictive accuracy is over 95% (for the next 7 days). With such a high accuracy, replacing disks in advanced make sense.
DiskProphet® collects SMART data and performance metrics of storage servers. S.M.A.R.T. is a disk monitoring system built with both HDDs and SSDs. Data provided by S.M.A.R.T. help DiskProphet® predict the likelihood of disk failure. DiskProphet® further uses performance metrics, such Storage I/O, and storage server I/O to provide disk failure prediction service.
There are three types of storage. (1) x86-based storage, such as Ceph, Nexenta, (2) Hyper-converged, such as Nutanix, (3) traditional storage, such as NetApp, EMC, HP 3PAR, etc. DiskProphet supports (1) and (2). (3) is complicated. DiskProphet depends on (3)’s command line tools or API for retrieving its SMART data and performance metrics. Please contact our sales office for more information.
Most of the x86-based storage and hyper-converged storage, there are some open source or vendor proprietary software support them. Currently, DiskProphet® supports CollectD and Telegraf for collecting host metrics. And DiskProphet® supports smartmontools for collecting SMART data. All the three software are popular open source projects and are used by many data centers.