An interesting advertisement appeared in The SSD Guy’s in-box, so I decided to ask more about it. SSD maker Swissbit is offering to customize one of its enterprise SSD models to match their customers’ workloads.
Matching an SSD to the workload has always been problematic for SSD makers and users alike. Everyone’s workload is different, and an SSD must be designed with a certain workload in mind. This means that the SSD will be a perfect fit in one application but may not perform well in another.
This has a lot to do with a number of variables. The biggest two are the randomness of the accesses, and the read/write balance. An SSD that has been optimized for a lot of sequential writes may not be very good at random reads, and vice-versa. Wear and cost are two other big issues that must be balanced against the others. SSDs with a considerable amount of overprovisioning can tolerate a much higher number of drive writes per day, but will cost a good bit more than drives aimed at read-mostly applications.
The SSD firmware designer must balance these four attributes, plus a large number of others, against each other to try to design an SSD that will fit the workload that the designer expects to make up the bulk of the company’s market, even though that design will not be right for everyone. Some SSD makers even use AI to help address this challenge.
What Swissbit has developed is an SSD, their N4200, whose firmware is custom-tailored to a customer’s application. They can do this by using either of two approaches:
- Customers who already know their applications’ workload can select from predefined workload cluster configurations, including options for Big Data, video streaming, or machine learning, whichever configuration best matches their system.
- If there are additional special customer requirements or needs, then test SSDs can be run in the customer’s system for a certain period to capture the real workload. These SSDs have been designed to capture only the workload data but no user data. Then Swissbit’s team can pull the workload log from the drive and analyze it in what is called the “Swissbit Optimization Process” to determine possible firmware changes that will help the SSD to meet the customer’s requirements. Swissbit then provides the customer new firmware to be flashed onto the drive.
The timeframe for this in-depth analysis process varies from customer to customer depending on the application, roughly between 6-8 weeks.
Even without optimization, the N4200 shows impressive performance, as the company illustrates with the graph below:
(There is not a lot of detail of the conditions under which these measurements were taken.)
The SSD Guy finds it intriguing that a company is putting forth the effort to customize SSDs to specific customer needs, while competitors figure out a workload that they think is best, and optimize for that, hoping that their design will work sufficiently well in a wide range of environments.
It’s an interesting approach. I will be watching to see if Swissbit’s competitors decide to follow the same path.
Objective Analysis focuses significant efforts on understanding the technologies that might change the market’s direction before that change actually occurs. If your company would like to stay at the forefront, and to reap the benefits of leadership, then please contact us to explore ways that we can work together to guarantee your success.