Write latency is choking your analytics... and your customers notice

A new generation of SSDs is the easiest way to modernize your data center for data-hungry workloads.

Choose a better SSD to serve more customers with more data ?

The single user experience that all technology solutions can be measured on is “time”. When an application or a website runs slower than another, we can all see the difference. It is much harder to feel things like “IOPS” (Input/Outputs Per Second), “MIPS” (Millions of Instructions Per Second), or “bandwidth”. These metrics are important to the system designers, but the users themselves only experience them indirectly. As an example, if the system you are using runs out of bandwidth or IOPS, the user experiences a “slowdown”. Time is the universal constant across all cloud and enterprise systems.?

In the world of storage devices, we use the terms “latency” and “latency consistency” to represent the value of “time”. Customers who want the best, seek out storage devices that can provide the lowest and most consistent latency under the most strenuous workloads. The industry saw huge improvements when Solid State Devices (SSDs) were first introduced and then continued to see great strides as we moved away from legacy interfaces (SATA, SAS) to put flash memory on a high-speed serial computer expansion bus where you could see the real benefits (PCIe w/ NVMe). Most recently, we’ve been pushing the bandwidth as the industry evolves from PCIe Generation 3, through Generation 4, to Generation 5… but as I pointed out above, bandwidth is great for system designers and improves the total work you can do, but these latest changes have negligible reductions in latency.??

What can I do to improve the most universally valued metric…time??

Today’s SSDs can achieve <0.1ms (milliseconds) latencies, which is amazingly fast, especially when compared to Hard Disk Drives (HDDs) with latencies around 3-5ms. However, those nominal values on SSDs drop off when system designers focus on latency consistency. Would you be surprised if I told you that 10% of the time it may take 10x longer to get you the data you need? That may not sound like much, but when system designers plan for hundreds of thousands of transactions per second, they do not want to see thousands of those transactions taking 10 times longer. This explains why many designers look at 99.9%, 99.99%, or even 99.999%+ metrics when planning for latency consistency. ?

The options for a system designer are limited. If they want to improve the latency of their SSDs, the primary option is to dial up the overprovisioning. To learn more about why and how this works, please watch our webinar, 7 Things to do while waiting for SSDs to finish writes. Unfortunately, overprovisioning forces you to increase your costs significantly; you pay the same price but get less usable storage space. An expensive and difficult tradeoff!

What if I want to recover my time, but can’t sacrifice capacity??

ScaleFlux Computational Storage SSDs with Transparent In-Line-Compression solve this challenge by compressing data as it comes into the drive, freeing up space, and dramatically improving latency consistency.

ScaleFlux demonstrates improved latencies across multiple testing environments where you can see multiple orders of magnitude improvement, all with ZERO changes to your systems or software.??

You don't have to take my word for it, you can see it for yourself. Try a ScaleFlux SSD in your environment by requesting 2 free samples. You can tell them Eric sent you ;)?