Flash FAQ: Flash Memory Basics

As flash-based architectures accelerate everything from databases to virtualization and cloud computing, this Flash FAQ is designed to provide background on many common flash questions.

If you want to know something not answered in the Flash FAQ pages, click on "ASK TODAY!" below and we will do our best to answer your question quickly and thoroughly. We are committed to helping IT professionals continue to develop their expertise in flash.

Other Flash FAQ Categories

Best Practices for Optimizing Flash

Flash Acceleration

Non-volatile Memory (NVM) Technology

NVM Questions

About Fusion

Go to Q&A

Flash 101

1. What is flash memory?

Flash, or NAND, or NAND flash memory, is a digital storage medium made of silicon chips. Flash stores information electronically as positive or negative charges, mirroring how computers use binary code in zeroes or ones. Flash memory is used widely in today’s smartphones, tablets and laptops. With the right software to manage flash reliability and longevity, flash is also being used in datacenter servers to speed up application performance.

There are many different varieties of flash. Single-level cell, or SLC flash, offers the most endurance because it is less dense than other flash varieties. However, SLC flash is limited in capacity. Multi-level cell, or MLC flash, offers more capacity and is often more useful for enterprise datacenters when flash companies use it with software features that ensure longer endurance in servers.

Manufacturers continue to increase the capacity of their NAND flash chips by shrinking the size of the cells that hold the electronic charges to store data. They can also increase the capacity by adding more layers, or even more axis to the flash chips. As density increases, sophisticated software to manage the flash becomes increasingly critical to ensuring that flash memory will perform as expected in enterprise servers.

2. What are the benefits of flash?

Flash memory makes applications faster—up to 10x faster compared to applications relying on hard disk drives. This helps companies find the business opportunity and intelligence in their information at the real-time speeds of our digital economy. In our personal lives, flash ensures that we enjoy fast, responsive website experiences, whether we’re waiting for data to load while banking, shopping or downloading music online, as well as when we are checking in on our friends and family on social networks.

Flash also offers a number of significant cost savings that often help enterprises realize an immediate return on their flash investment. With flash, enterprises need far less space for storage sprawl because racks of disks are no longer needed to provide application performance. Since flash has no mechanical parts, it generates far less heat than spinning disk storage, saving money that would have been spent on cooling. Flash also reduces the amount of time IT professionals need to spend servicing disk drives, helping to free up more time to work on strategic business objectives.

3. Why is flash faster than disk?

Flash memory chips store information electronically on silicon rather than on magnetic spinning disks. Flash stores data on electrons, not atoms, and electrons move much faster than atoms. Disk drives use the same basic principles as a phonograph from the 1800s. Response times become even slower as disks fill up and the read head has to travel further inside the platter. Although flash is fundamentally faster than disk, memory products made from flash differ in their architectures, producing differences in speed and reliability.

4. How much faster than disk is flash?

Flash memory is thousands of times faster than spinning disk drives. At the application level, the amount that flash will outperform disk will vary depending on your specific application and architecture. For direct application acceleration, many companies have reported that flash can outperform disk drives by 10x and more. When using flash to accelerate data-intensive solutions like server virtualization and virtual desktop infrastructure (VDI), companies adding flash performance to their servers have reported performance gains up to 40x.

5. How reliable is flash?

Reliability can vary among flash solutions, but leading products provide software features that ensure enterprise data integrity even if an individual NAND chip should fail. For example, Fusion-io products feature a technology called Adaptive Flashback that delivers built-in redundancy at the block and chip level to protect against NAND failure in the media itself, without having to rely on system-level protection. This enables peak performance and capacity, while simultaneously extending the life of the device far beyond what is possible with other approaches. For more details on these features in Fusion-io products, you can also read our blog post on enterprise reliability.

6. How long does flash last?

Like all storage media, flash can wear out over time, but most enterprise flash products are designed to last for years of use. When evaluating a flash solution, it's a good idea to determine the read and write requirements of your applications to make sure you choose a flash product that offers endurance that meets your needs.

Software features in enterprise flash products help ensure data integrity and flash memory longevity beyond the endurance of individual NAND flash chips. For example, Fusion-io offers a feature called Adaptive Flashback that enables flash to self-heal if an individual NAND die shoud fail. Other features like wear-leveling and predictive management help ensure a long flash lifespan in Fusion-io products. With features like these, enterprise flash products can last as long as the servers they are accelerating.

7. What is the difference between flash reliability and endurance?

Flash reliability refers to a product's ability to operate under enterprise data demands without failure. Endurance refers to how long a flash memory product can be expected to last. Many flash products measure endurance by how many petabytes of data can be written, erased, and written again to the device. The amount of time that it will take for a device to reach that maximum depends on the unique requirements of each application and overall enterprise workloads.

8. How is enterprise flash different from flash for laptops or cell phones?

The raw flash memory used in enterprise flash is often the same flash fabricated for consumer products like cellphones, tablets, and laptops. Software features provide greater endurance and reliability for enterprise flash memory compared to consumer devices.  Features like over-provisioning and grooming help enterprise flash products perform well when almost at full capacity. These features can vary from product to product, so it’s a good idea to learn about offerings of specific flash products before moving forward with a purchase decision.

9. Why does flash cost more than disk drives?

Flash actually costs less than disk when evaluated by dollar per input/output operation ($/IOP). Flash also offers more capacity per square inch compared to disk. When evaluating the cost of a flash investment, we recommend examining the cost per operation ($/IOP) rather than the cost per gigabyte ($/GB). Often, adding a small amount of flash to a server can significantly improve performance, reframing the discussion around performance rather than capacity.  If businesses will benefit from being able to respond more quickly and save money on not having to scale out, power, and cool more storage, it is likely that flash would add business value to that enterprise’s IT infrastructure, in addition to significant savings when examining the total cost of opportunity.

10. How does flash memory save enterprises money?

In the datacenter, flash memory provides cost savings that often deliver an immediate return on the flash investment. Flash offers much more capacity than spinning disks, helping to save the costs of sprawling racks of spinning disk storage that is expensive to power, cool, and maintain. When fewer servers are able to deliver more work, flash can help save on software licensing costs. Flash also reduces the amount of time IT professionals need to spend replacing disk drives, freeing up more time for working on strategic business objectives.

In addition to saving on costs, the speed of flash memory provides value by helping companies get more insight from their information. For example, faster applications can enable employees to quickly get to work on responding to the results of a sales report, rather than waiting for those reports to process. There are many examples of how flash is accelerating business in these case studies.

11. How can flash memory be added to a datacenter?

Flash memory can be added to datacenter servers via the PCIe connection for direct application acceleration and as a cache to accelerate virtual machines and VDI. Flash can also be used in storage arrays and in combination with other storage media, like disks.

Choosing where to begin using flash first depends on an analysis of a company’s current infrastructure, business plans, objectives, cost considerations, and resources. Some companies may choose to begin using flash in one area for ease of migration, later adding flash into other areas to evolve their datacenter in tandem with their business needs. To see how flash may help your specific needs, feel free to browse the Fusion-io solutions webpage for more information or chat live with Fusion-io.

12. Where does flash offer the most performance?

Since flash is an investment that can be added in many locations from the server to the storage array, many enterprises want to make sure they are adding flash where it can provide the most benefits. Flash can speed up applications in all locations, but adding it to servers offers the most raw performance since that puts the flash resource closest to the application and the CPU. Flash can also serve as a caching store that connects to back end storage to accelerate the most active data.

If you want to prepare for a possible migration to an all-flash architecture in the future, it’s a good idea to check into options that allow the storage media to be upgraded from disks to flash in the future. In hybrid storage systems, software that enables IT teams to provision performance to a variety of applications is essential to meeting varying Quality of Service requirements for applications that share flash resources.

13. How do I know if flash will help my application?

From databases to search to analytics and business intelligence, flash memory is widely used today to help enterprise applications perform faster. For raw performance, adding flash to servers offers the most speed because the data stays in the server where the application is hosted. Many companies using databases including Microsoft SQL Server, MySQL, SAP HANA, Cassandra, and MongoDB have reported substantial application performance improvements with flash memory. Flash memory is also proven to improve virtualization of both servers and desktops, or VDI.

Flash memory in shared storage can still help accelerate applications and virtualization with significantly more performance than disk. Sharing flash storage across servers enables the resource to be allocated to many needs, but raw performance won’t be as fast as flash directly attached in the server. If you would like to test flash performance for your applications, please feel free to contact us.

14. What kind of applications benefit most from flash?

Most enterprise applications benefit significantly from flash memory acceleration. From databases like Microsoft SQL Server, Oracle, MySQL, SAP HANA, big data NoSQL databases like MongoDB, Cassandra, and Hbase, to enterprise tools like Microsoft Sharepoint, virtual desktop infrastructure (VDI), and server virtualization, flash helps enterprises turn data into information at real-time speeds.

15. How do applications use flash?

Today, nearly all applications use flash as a block storage device. Fusion-io is developing flash-aware APIs to enable applications to leverage flash as an extension of memory. Early innovators in using flash-aware APIs for breakthrough application performance include MariaDB and Percona, who are adding new flash-aware features to their databases.

16. How can faster data improve business results?

For a national retailer, faster data powered by flash can help analyze a weekend sales report in a few hours rather than days, enabling executives to adapt more quickly to what is working and what is not. For a medical professional, flash loads a patient’s information from a database onto a computer or tablet, giving them instant access to life-saving data whether a doctor is at an office or at the hospital bedside.

Flash can also ensure responsive customer experiences for online companies. For a streaming content company, flash helps keep a website up even as a new hit video goes viral. For an online retailer, flash helps process millions of payments at the speed we click to check out customers in seconds during even the world’s busiest online shopping frenzy.

In today’s real-time economy, examples like these can be found across all business verticals.

17. Is flash a block storage?

Flash is a raw storage medium that can be used as a block storage. How the flash medium can be used depends on the flash controller used to manage the device.

18. What is a flash controller?

A flash controller manages all the flash chips on a flash memory product. It manages where to put and send data needed by applications. In some products, it can also monitor the health of the NAND flash chips to report or even correct errors on the device.

19. What is wear-leveling in flash?

Wear-leveling ensures that all blocks are written equally across all NAND flash cells on the device. This helps to protect the longevity of the flash memory product.

20. What is grooming in flash?

Grooming is the process of continually reorganizing data stored on the NAND flash cells to ensure optimal placement of data for performance. It ensures that there is a block available to write to when a request is made. This helps maintain flash performance during writes.

21. What is over-provisioning and under-provisioning?

Over-provisioning is reserve capacity of NAND flash memory that is provided in addition to the official device capacity. All NAND flash products require some degree of over-provisioning to provide grooming functions. Intelligent flash products provide the ability to adjust the amount of capacity used for provisioning if you understand your application’s read and write requirements. Some intelligent flash memory products also allow you to under-provision to regain the extra capacity to use for applications that mostly read rather than write.

22. How is performance measured for flash SSDs?

Flash memory products are categorized by a number of factors, including capacity, latency, bandwidth, and input/output operations per second, or IOPS. Many enterprises choose multiple flash products to meet different needs within their IT infrastructure.

23. Why do IOPS numbers often differ for reads and writes?

When an application requests a read, the data has to be found by a log file "look up" that determines where the data was stored. Writes are executed directly, without a log file look up. This results in assymetrical performance for reads and writes.

24. What is the difference between GB/s and Gb/s?

GB/s stands for gigabytes per second, which is a measure of data throughput. Gb/s is the acronym for gigabits per second, which is a measure of networking speed.

25. How can I learn more about flash?

Even if you're just starting to consider adding flash to your datacenter, we'd love to help you develop your expertise. Give us a call at 800.578.6007 to set up a meeting, or send us a message to chat live.