SSDs with application-specific optimization

Numerous highways in Germany have no posted speed limit. Cars typically move at a speed of 90 mph, with the occasional driver reaching speeds of 150 mph — if the weather and traffic conditions allow for such high-adrenaline driving. The "tool" needs to be appropriate for the task; yet, an off-road vehicle will get you to your goal more quickly than any sports car on rocky wilderness roads. Data travels at the speed of light at data centers, which is much quicker than any Audi, BMW, Mercedes, or Porsche. However, the data center's traffic patterns are just as diverse as those on real roads: while some data can move freely on multi-lane fiber highways, others may need to constantly travel back and forth along multi-tier application landscape alleyways or deal with the constant rush hour of data storage.

How quickly and effectively one moves from point A to point B will depend on choosing the right equipment in the data center as well. Because they read and write data up to 10 times faster than hard disk drives (HDDs), solid-state drives (SSDs) have largely taken the place of HDDs in computing.

When compared to traditional spinning storage media, SSDs are in a class of their own, yet there can be significant variations in read/write speeds. Streaming video has different requirements for how to interface with SSDs than interactive cloud apps or machine learning, thus a lot of this depends on the application that gets the data. This is one of the explanations given by the major cloud service providers, or so-called hyperscalers, such as AWS or Microsoft, for spending the time, money, and effort necessary to setup and occasionally even construct their own data center hardware.

Smaller service providers and data center operators must rely on off-the-shelf hardware since they lack the luxury of having servers and storage customized to their specific needs. This presents numerous difficulties. For instance, conventional SSDs encounter spikes in delay when performing memory recovery procedures that are aptly and colorfully referred to as "garbage collection". These spikes may slow down e-commerce transactions or generate annoyance-inducing interruptions when streaming movies. Therefore, the ability to modify SSD hardware to manage such challenges gives companies like Amazon and Microsoft a competitive advantage.

However, there is another approach for data center managers to adapt SSDs to the needs of certain applications. Modern SSDs can be used in conjunction with specialist software that examines how an application uses them, including how frequently and quickly it writes data. Do data writes happen at random or in a specific order, etc.? The SSD firmware can be adjusted to provide exactly what the application requires by examining app activity. The solid-state drives are protected from unneeded wear and tear at the same time.

Although the goal of optimizing SSDs for application-specific use is to boost performance, data center operators can also benefit from lower TCO. Operators gain from this innovation in three specific ways:

1. Reduced latency: SSD firmware tailored to specific applications can up to two times faster read/write response times. In the aforementioned illustration, a colocation provider for a video streaming service may modify the SSD firmware to forgo garbage collection while active video streaming is in progress. As a result, the colo provider could guarantee response times for streaming, turning a minor technical improvement into a salable commercial advantage.
2. Durability: SSDs must be updated frequently, typically every three years. According to stress studies, SSDs can have a five-year longevity by being optimized for apps. The SSD data space has more uniformly distributed data writes as a result. Hyperscalers accomplish this using unique techniques. However, with application-optimized SSDs, these modifications are continuously and automatically made without additional user input.
3. Performance stability: SSD read/write speeds tend to drop off quickly at first. After just twelve to eighteen months, it will typically be significantly lower, occasionally falling to just one-third of its initial capability. Using app-specific modifications to improve data write handling, application-optimized SSDs also perform significantly better in this area. Their optimized firmware will not only increase SSD longevity but also maintain read/write performance much closer to the initial level, with a performance drop of less than 10%. This performance improvement makes app-optimized SSDs far more cost-effective when coupled with the prolonged lifecycle.

Data center operators must modify their storage infrastructure to as closely match the application load as feasible in their fiercely competitive market. When purchasing a vehicle, it is obvious that you need to choose one that is appropriate for both your unique driving style and the types of road surfaces you will likely encounter, such as asphalt highways and city streets or gravel and rock roads in the wilderness. The user experience of millions of consumers may be at stake in data center storage, making it even more crucial to adapt the machinery to the work at hand. Data center operators can dominate their respective market segments by customizing the SSD firmware for specific applications, and quicker, more robust, and more dependable solid-state SSDs give them an advantage over rivals.

Contact our experts