If you’ve used a smartphone then you might be familiar with predictive text technology that tries to predict what you are going to type next. Predictive text technologies use algorithms to identify patterns in what you and other users have typed before to predict and help finish your sentences. Scale Computing has now unveiled similar technology that it has applied toward data storage.
With newly announced “Predictive Ordered Write” technology, Scale Computing has created algorithms to identify patterns in data writes to storage and can now predict what data you and your applications will be writing in the future. While this technology may seem unnerving to some, it is designed to greatly speed up disk performance by writing data to the disk before you decide it even needs to be written.
Before a write operation ever hits the file system, this new technology is already predicting the data patterns it expects and is committing those blocks to disk ahead of time. When the write operation finally does come, nothing actually needs to be written and the write operation is immediately successful. This dramatically increases the response time for write operations since the data already exists.
The predictive algorithms and technology are proprietary but they combine analysis of known data patterns with analysis of user profiles from social media to help predict the type of data they might enter into systems or even what kind of data they might want to see. This technology has a prediction accuracy of 98%. For the other 2% of writes that weren’t predicted, those are simply written to disk after the write operation is received.
Phil White, CTO at Scale Computing described the technology like this, “When you break it down in simple terms, data is really only made up of ones and zeros. It isn’t as hard as you might think to detect patterns and predict future patterns in data.”
Predictive Ordered Write (POW) technology has other useful benefits beyond speeding up disk write performance. Now data can be accessed and processed before it is even entered into the system. Depending on the system, predicted writes can be committed minutes or even hours before the actual write is requested. For example, reports run against a database can reflect up to hours worth of future data.
Jeff Ready, CEO and co-founder of Scale Computing, had this to say, “This POW technology is a game changer. People used to think you needed an input before you could get an output. Not anymore. Welcome to the new world of O/I.”
Scale Computing is also planning an upcoming release of predictive delete technology to preemptively delete data for you, saving on storage costs. No more details are available at this time.