NetApp All-Flash FAS: Performance, Availability, And Efficiency

Introduction to Flash Storage

Flash storage is gaining popularity at a high rate because of its greater speed and the ability to utilize less capacity [1]. Flash allows organizations to enjoy advantages like consolidated apps, reduced power consumption per machine, and increased performance. In the past years, flash storage was being used reservedly because it was quite expensive. Due to the evolution and advancement of technology, efficiencies and improvements have been made on flash storage thus reducing the cost [2]. Nevertheless, there are still some instances where the use of hard disk drives (HDD) is still more economical and efficient. This is one of the major reasons why there is rise of storage solutions such as hybrid flash storage.

Currently, there are many vendors who provide all-flash arrays (AFAs) with high performance in order to meet the high-throughput processes and workloads with low latency requirements and data access with random patterns like online data processing and virtual desktop infrastructure. AFAs provide high performance as compared to HDD-based systems for similar processes although several of them do not have vital abilities in a single crucial area: management of enterprise data [3]. NetApp provides all-flash FAS storage arrays that can be implemented as a section of scaled-out architecture that is unified and that bring numerous advantages to the leading industry data management properties of ONTAP operating system. Combining enterprise data management and flash technology bring benefits to three main areas: storage efficiency, availability, and performance.

Figure 1: NetApp Flash Devices

Source: Adapted from [3]

Solid state disks (SSDs) stores data in all-flash arrays. Flash media is used only for non-volatile and persistent storage- data that cannot be affected by power failures or blackouts. Speed is one key property of all-flash because it has no moving parts making it easier and faster to carry out I/O operations and quick read and write [4]. Because of the high performance offered by all-flash makes it possible to run critical applications that require high-speed data input.

Figure 2: All Flash Array increasing functionality while price is decreasing

Source: Adapted from [4]

In addition to performance, AFA is cot effective because they typically occupy less space and are smaller to fit in the storage rack. It also consumers less power because flash runs on solid state disks that has no moving part thus requires not or less cooling. The decision makers should look at the overall picture that all-flash arrays have to offer even though critics have argued that it is costly. The total cost of ownership is seen when the companies use all-flash arrays as general-purpose storage infrastructure for applications as compared to hard disk drive-based systems. Many C-level employees still have the notion that flash storage is an expensive storage alternative for business because they lack more understanding of the impact that such technology can have at business-level.

Benefits of All-Flash Arrays over HDD-Based Systems

NetApp FAS systems latest generation employs software enhancements and several CPU cores to unlock higher speed in solid state drive as compared to past FAS platform. For instance, FAS8000 family depending on the architecture and PCIe, can be scaled out up to 24 nodes to offer many IOPS (millions) at a very high-speed latency (sub-milli second) and support SSD capacity of up to 5PB [5]. The new hardware features allow NetApp FAS systems to enhance the performance of solid-state drive including:

• De-staged writes: NetApp all-flash FAS removes write limits that SSD has by using NetApp write anywhere file layout (WAFL) that eliminates SSD writes from the vital workload latency path. This is achieved by directly writing to the system memory which is a faster process compared to writing to solid-state drives. To protect incoming writes, FAS system utilizes non-volatile random-access memory (NVRAM) that is backed by a battery. Writes are logged into the NVRAM when it is received in the memory form the host. Once received, the NVRAM sends write acknowledgement to the host immediately confirming that the write process was completed [6]. Writes are then de-stages to the solid-state drives during a perioding consistent point.

Figure 3: How De-staged writes occur

Source: Adapted from [6]

• Proximal Data Patterns: to reduce the effects that random operations have on the SSD, WAFL employs the use of proximal write patterns and a highly flexible SSD write allocation policies as writes are de-staged to the SSD from the memory. No blocks are assigned permanently to disked locations that are fixed by adopting these practices. The throughput of SSD’s write is optimized by this technique and the need for extra write cycles are minimized which are not desired in flash technology, commonly known as ‘write amplification.’ Reducing the effects of random processes and minimizing the number of writes increases SSD overall life span and increases performance.
• Read-ahead caching: read-ahead algorithm of a data ONTAP which is customized has the ability to identify data that is read often (hot data). The algorithm detects detect the read patterns even before the host makes a read request and stage to the system memory pre-emptively the reads that are have higher chances of being accessed. This allows direct reading of cache reads from the memory without having to access the data from SSD directly [7]. By doing so, the number of I/O requests are reduced that should be served directly by the SSD creating more space to support heavy SSD writes periods. Additionally, scheduling algorithms are utilized to ensure that latency-sensitive reads are prioritized over throughput-sensitive writes.

Figure 4: Read ahead caching

Source: Adapted from [7]

Secondly, all-flash arrays, when implemented as a section of a scaled-out cluster, benefits for operations that has been enabled by DATA ONTAP that is clustered and are non-disruptive which has been proven to offer more that 99.99% uptime in production environments, thus ensuring enterprise-class availability [8]. More benefits are reaped by moving the processes within a cluster with the ability to expand to 24 nodes. Many processing can be executed and run without experiencing downtime in terms of storage capabilities including software and hardware upgrades, performance balancing, and key lifecycle processes. The cost of protecting data is also minimized by storing copies of data from every all-flash node to hybrid nodes a cheaper price per terabyte [9]. Moreover, workloads and processes can be moved seamlessly between the high-performance all-flash nodes and hybrid nodes in order to optimize performance and cost while responding to the changing requirements.

NetApp solid state drives have a high rating based on daily write operation and has been proven and warrantied by world support organization. It is also easy to monitor the status of the SSD using CLI commands that are simple and easy to remember and which provides important information concerning SSD wear. The event management system on the Data ONTAP records an auto support event and notifies to facilitated preventive measures in the event that the wear threshold of SSD has been surpassed.

Thirdly, is storage efficiency; when considering all-flash array, one of the primary concerns is cost. SSD based systems are continuously gaining significant premium compared to hard-disk drive-based system despite the reducing costs. Additionally, the vendors of All-flash arrays are starting to introduce techniques of reducing data such as compression and deduplication to lessen the high SSDs cost. Nevertheless, access to a wider storage suite is provided by deploying an all-flash array with Data ONTAP in order to reduce cost and increase storage efficiency.  Users of NetApp all-flash FAS can use the following technologies to store active data the exceeds storage limit of the system: Snapshot technology, Data compression, RAID-DP, Thin provisioning, storage-efficient replication, FlexClone technology, and deduplication [10].

Enterprise Data Management with NetApp All-Flash FAS

NetApp provides a wide range of storage solutions that are flash-optimized including all-flash arrays and hybrid flash that is designed to enhance the performance of the application while ensuring that availability and reliability are maintained at high level [11].

Hybrid flash arrays is a combination of hard-disk drive and solid-state drive technologies to enable enterprises to take advantage of the HDD predictability and SSD high-performance levels. It makes economic sense to businesses and organizations relying on HDD storage infrastructure to add some flash-storage solutions for specific applications. It allows the to transit slowly to the new storage technology to achieve high level of performance while reducing the cost that could be incurred in direct change-over.

Figure 5: Difference between Hybrid and All-Flash Arrays

Source: Adapted from [12]

Hybrid flash also has the ability to gain efficiency and adaptability [12]. Although HDD can store large volumes of data there are some challenges that may cause it to be slow. A balanced storage infrastructure is achieved by combining the high capacity of HDD and the high-performance of SSD [13].

Conclusion

Both hybrid and al-flash arrays are cost effective and provide enhanced performance regardless of the storage infrastructure that is already in place. They offer greater speed and capacity that the past versions in addition to cloud-ready capabilities and guaranteed availability, satisfaction guarantees and timeless storage. Flash allows organizations to enjoy advantages like consolidated apps, reduced power consumption per machine, and increased performance. In the past years, flash storage was being used reservedly because it was quite expensive. NetApp provides all-flash FAS storage arrays that can be implemented as a section of scaled-out architecture that is unified and that bring numerous advantages to the leading industry data management properties of ONTAP operating system. The decision makers should look at the overall picture that all-flash arrays have to offer even though critics have argued that it is costly.

References

[1] H. Heo, M. Pirahandeh, K. Lee and D. Kim, “All Flash Array Storage Virtualisation using SCST”, KIISE Transactions on Computing Practices, vol. 20, no. 10, pp. 525-533, 2014.

[2] D. Archdeacon, T. Boothby and J. Dinitz, “Tight Heffter Arrays Exist for all Possible Values”, Journal of Combinatorial Designs, vol. 25, no. 1, pp. 5-35, 2016.

[3] A. Chimenton, C. Zambelli and P. Olivo, “A Statistical Model of Erratic Behaviors in Flash Memory Arrays”, IEEE Transactions on Electron Devices, vol. 58, no. 11, pp. 3707-3711, 2011.

[4] M. Shafaei, M. Hajkazemi, P. Desnoyers and A. Aghayev, “Modeling Drive-Managed SMR Performance”, ACM Transactions on Storage, vol. 13, no. 4, pp. 1-22, 2017.

[5] U. Ferner and M. Medard, “Coded-Seeking: A Simple HDD Speed-Up Concept”, IEEE Communications Letters, vol. 19, no. 2, pp. 139-142, 2015.

[6] D. Novak, “Flash! ‘A Flash in the Pan’: Flash, Performance, and Event”, Journal of Victorian Culture, vol. 23, no. 4, pp. 497-502, 2018.

[7] L. Xu, J. Cipar, E. Krevat, A. Tumanov, N. Gupta, M. Kozuch and G. Ganger, “Agility and Performance in Elastic Distributed Storage”, ACM Transactions on Storage, vol. 10, no. 4, pp. 1-27, 2014.

[8] C. Zambelli, T. Vincenzi and P. Olivo, “A Compact Model for Erratic Event Simulation in Flash Memory Arrays”, IEEE Transactions on Electron Devices, vol. 61, no. 11, pp. 3716-3722, 2014.

[9] H. Luo, Q. Liu, Z. Qiao, J. Wang, M. Wang and H. Jiang, “DuoModel: Leveraging Reduced Model for Data Reduction and Re-Computation on HPC Storage”, IEEE Letters of the Computer Society, vol. 1, no. 1, pp. 5-8, 2018.

[10] F. Fujitsu, “NetApp All-Flash FAS – Fujitsu CEMEA&I”, Fujitsu.com, 2018. [Online]. Available: https://www.fujitsu.com/fts/products/computing/storage/all-flash-arrays/netapp-all-flash-fas/. [Accessed: 12- Nov- 2018].

[11] N. NetApp, “All Flash FAS (AFF) | All Flash Storage Arrays | NetApp”, Netapp.com, 2018. [Online]. Available: https://www.netapp.com/us/products/storage-systems/all-flash-array/aff-a-series.aspx. [Accessed: 12- Nov- 2018].

[12] A. Taylor, “All-flash or Hybrid Flash: How to Decide”, Network World, 2018. [Online]. Available: https://www.networkworld.com/article/3282830/storage/all-flash-or-hybrid-flash-how-to-decide.html. [Accessed: 12- Nov- 2018].

[13] D. Robertson, “Problems and Solutions: How Applications Drive Data Converters (and How Changing Data Converter Technology Influences System Architecture)”, IEEE Solid-State Circuits Magazine, vol. 7, no. 3, pp. 47-57, 2015.