Kingston Technology252 тыс
Опубликовано 30 июля 2021, 17:00
For some IT professionals, there’s an impression that all SSDs are created equal with no real difference because they're only flash memory, but that is not the case. The way SSDs are built for notebooks or desktops versus how one built for a data center application is vastly different.
A server workload usually maintains 100-percent duty cycle 24 hours a day, seven days a week, year-round. But when we work on our computers throughout the day, even on a virtual machine, we’re only using minimal resources with a lot of idle time. Server drives need to withstand the constant demand and performance requisites for heavy-duty use.
The Right SSD Matters
You can’t take an SSD out of a laptop and expect it to operate in a data center. That’s like expecting a standard car engine to power a semi-truck. Data centers need drives that are enterprise-grade and purpose-built to deliver predictable performance at scale.
Delivering on SSD Performance Consistency
SSDs are used in more applications in the data center than ever before. SSDs are shifting into more high-performance applications, and that's where performance consistency becomes important.
To create SSDs tuned to the demands of a data center, the focus is on consistent IOPs delivery and latency. To test this, performance scripts are created that look at every I/O over a very long test pattern.
What we tune for, and what IT architects want to see, are straight-line of consistency when benchmarked with their workloads. There is never 100-percent perfect consistency, but we want to avoid the sawtooth pattern or what’s referred to as a “Christmas Tree” when the data is graphed.
A drive that is not properly tuned may experience big swings in performance. At one point, the drive may perform 50,000 IOPs, and then it drops to 20,000 IOPs before bouncing back up 60,000 IOPs. While the high numbers look excellent for a spec sheet and sales literature, spikes don’t tell the whole performance story.
Kingston’s enterprise SSDs are designed to deliver consistent levels of performance, even if it means sacrificing peak levels of performance. Consistency ensures that customers aren’t disappointed by the lack of performance, and have a predictable advantage for managing their storage clusters.
A lot of what makes this straight line and the consistency happen goes into the SSD firmware design, the size of the over-provision area and the write cache size. With better predictability, our customers are able to build applications around the consistent performance, and often meet aggressive service-level agreements.
What about latency?
Latency acts in the same way as I/O. For example, if the latency is cruising along at sub-5ms and then jumps up to 500ms before coming back down to 5ms, there is a problem as this performance inconsistency could show up in the application somewhere.
Data center customers have become much more sophisticated. In the past a supplier could provide them a good client-level SSD and they could simply rip it out and replace it with another cheapo drive when something went wrong.
But today, datacenter customers are peeling back the onion on SSDs because they know inconsistencies exist and how to test for them. The best enterprise SSDs today utilize large DRAM caches and are tuned within the firmware to deliver on consistency. In fact, about 90 percent of what smooths out these lines is done in code in firmware – which is a big differentiator for Kingston products.
In the market today, roughly 80 percent of the SSDs that go into service are still SATA. They have relatively fast transfer rates and use a relatively small footprint on the motherboard - which means more drives with excellent RAID profiles. But now there’s a big to move away from SATA to use NVMe drive across the PCIe bus. NVMe was designed from the ground up for Flash based SSDs and promises to break the limitations associated with legacy interfaces. But making the switch isn’t that easy.
PCIe is a great interface for SSDs but many servers are limited today in their capacity and sheer number of available lanes. The simple fact is that people can’t turn that switch overnight to migrate to NVMe. NVMe inherently provides lower latency and higher IO delivery, so customers expect more performance obviously out of an NVMe drive. Whether building for a SATA or NVMe interface, the performance criteria remains the same.
When choosing the right enterprise SSD, data centers should look for SSDs that will reduce latency and limit any possibility of IO cliffs. Lastly, the customer has to consider if their infrastructure is ready to utilize the newer faster storage interfaces. SSDs matter, and the selection customers make for their data centers is becoming increasingly critical. So when looking for the right SSD remember to look to Kingston.
Kingston's Cameron Crandall talks about why the right type of SSD is important when it comes to use in data centers.
#KingstonIsWithYou
A server workload usually maintains 100-percent duty cycle 24 hours a day, seven days a week, year-round. But when we work on our computers throughout the day, even on a virtual machine, we’re only using minimal resources with a lot of idle time. Server drives need to withstand the constant demand and performance requisites for heavy-duty use.
The Right SSD Matters
You can’t take an SSD out of a laptop and expect it to operate in a data center. That’s like expecting a standard car engine to power a semi-truck. Data centers need drives that are enterprise-grade and purpose-built to deliver predictable performance at scale.
Delivering on SSD Performance Consistency
SSDs are used in more applications in the data center than ever before. SSDs are shifting into more high-performance applications, and that's where performance consistency becomes important.
To create SSDs tuned to the demands of a data center, the focus is on consistent IOPs delivery and latency. To test this, performance scripts are created that look at every I/O over a very long test pattern.
What we tune for, and what IT architects want to see, are straight-line of consistency when benchmarked with their workloads. There is never 100-percent perfect consistency, but we want to avoid the sawtooth pattern or what’s referred to as a “Christmas Tree” when the data is graphed.
A drive that is not properly tuned may experience big swings in performance. At one point, the drive may perform 50,000 IOPs, and then it drops to 20,000 IOPs before bouncing back up 60,000 IOPs. While the high numbers look excellent for a spec sheet and sales literature, spikes don’t tell the whole performance story.
Kingston’s enterprise SSDs are designed to deliver consistent levels of performance, even if it means sacrificing peak levels of performance. Consistency ensures that customers aren’t disappointed by the lack of performance, and have a predictable advantage for managing their storage clusters.
A lot of what makes this straight line and the consistency happen goes into the SSD firmware design, the size of the over-provision area and the write cache size. With better predictability, our customers are able to build applications around the consistent performance, and often meet aggressive service-level agreements.
What about latency?
Latency acts in the same way as I/O. For example, if the latency is cruising along at sub-5ms and then jumps up to 500ms before coming back down to 5ms, there is a problem as this performance inconsistency could show up in the application somewhere.
Data center customers have become much more sophisticated. In the past a supplier could provide them a good client-level SSD and they could simply rip it out and replace it with another cheapo drive when something went wrong.
But today, datacenter customers are peeling back the onion on SSDs because they know inconsistencies exist and how to test for them. The best enterprise SSDs today utilize large DRAM caches and are tuned within the firmware to deliver on consistency. In fact, about 90 percent of what smooths out these lines is done in code in firmware – which is a big differentiator for Kingston products.
In the market today, roughly 80 percent of the SSDs that go into service are still SATA. They have relatively fast transfer rates and use a relatively small footprint on the motherboard - which means more drives with excellent RAID profiles. But now there’s a big to move away from SATA to use NVMe drive across the PCIe bus. NVMe was designed from the ground up for Flash based SSDs and promises to break the limitations associated with legacy interfaces. But making the switch isn’t that easy.
PCIe is a great interface for SSDs but many servers are limited today in their capacity and sheer number of available lanes. The simple fact is that people can’t turn that switch overnight to migrate to NVMe. NVMe inherently provides lower latency and higher IO delivery, so customers expect more performance obviously out of an NVMe drive. Whether building for a SATA or NVMe interface, the performance criteria remains the same.
When choosing the right enterprise SSD, data centers should look for SSDs that will reduce latency and limit any possibility of IO cliffs. Lastly, the customer has to consider if their infrastructure is ready to utilize the newer faster storage interfaces. SSDs matter, and the selection customers make for their data centers is becoming increasingly critical. So when looking for the right SSD remember to look to Kingston.
Kingston's Cameron Crandall talks about why the right type of SSD is important when it comes to use in data centers.
#KingstonIsWithYou