Topics In Demand
Notification
New

No notification found.

The Next Generation of Enterprise NVMe SSD Hosting
The Next Generation of Enterprise NVMe SSD Hosting

August 23, 2024

23

0

Whether you’re a business blogger, an online store owner, or a website designer, one thing is clear: your website needs to deliver strong performance, lightning-fast speed, and a visually appealing experience. While beauty might be subjective, the performance (including speed) directly ties to the storage technology provided by your cloud hosting provider. Therefore, it’s crucial to choose a hosting service that offers a feature-rich yet budget-friendly solution without compromising on performance.

As technology rapidly evolves, the web hosting industry is also undergoing significant shifts. One of the latest trends is NVMe SSD hosting, a technology that's setting new standards in speed and performance. You might have noticed many VM hosting providers now use NVMe in their servers, and there's a good reason for it—NVMe offers exceptional performance.

The Rise of NVMe

Initially, most SSDs relied on traditional interfaces like SATA or SAS to connect to a computer’s data bus. However, these interfaces, designed originally for mechanical hard drives, quickly became bottlenecks when paired with the NAND flash memory used in SSDs. This led to the development of the NVMe (Non-Volatile Memory Express) interface in 2009 by a working group led by Intel.

Unlike conventional setups where multiple hard drives share a single PCIe controller through SATA, NVMe drives connect directly to the host system via the high-speed PCIe (Peripheral Component Interconnect Express) interface. NVMe significantly increases the number and depth of queues, allowing systems to leverage flash memory's high concurrency and low latency fully.

This advancement has led to the widespread adoption of NVMe SSDs in computing applications that demand high IO performance. However, maintaining such high performance while implementing RAID data protection systems presents new challenges, as this technology was initially designed for mechanical hard drives.

Next Generation of Enterprise NVMe SSD Hosting

Several key features characterize the next generation of enterprise NVMe SSD hosting:

  • NVMe Technology: Utilizes NVMe for ultra-fast data transfer rates, significantly reducing latency.
  • Enhanced Speed: Delivers read and write speeds that are several times faster than traditional SSDs, maximizing system performance.
  • Increased Capacity: Offers larger storage capacities, allowing for vast amounts of data without compromising speed.
  • Improved Durability: Ensures robust durability and endurance, extending the lifespan and reliability of SSDs for enterprise use.
  • Enhanced I/O Operations: Optimizes input/output operations per second (IOPS), enabling faster data access and processing.

Current NVMe Data Protection Solutions

Due to NVMe's high performance and low-latency characteristics, many organizations are now using NVMe SSDs as the primary storage medium in their servers. However, when NVMe SSDs are directly connected to the host, they typically rely on traditional RAID technologies for data protection, such as Software RAID and Hardware RAID.

Software RAID:

The concept of Software RAID for NVMe is similar to that used with mechanical hard drives. It uses the host system's CPU to manage NVMe operations and calculate checksums. The key difference is that, since NVMe connects to storage devices via PCIe, bandwidth is higher, latency is lower, and the instruction set design is simpler. This makes it highly efficient to directly handle RAID through the CPU.

However, a significant challenge with Software RAID arises when RAID modes requiring checksum calculations, like RAID5 or RAID6, are used. For example, a 4K random write request in RAID5 generates additional read and write commands, as well as a checksum calculation. This process can consume a substantial portion of CPU resources, making it necessary to adopt high-end CPUs to fully maximize NVMe SSD performance, which increases system costs.

Hardware RAID:

Hardware RAID is a good solution for traditional hard drives, as all RAID logic is handled by a separate hardware controller, offloading the computation from the host CPU. However, with NVMe SSDs, this approach can become a performance bottleneck due to the high-speed nature of NVMe drives and the limited number of PCIe lanes on the RAID controller itself. This limitation can directly restrict the number of SSDs that a controller can utilize to set up RAID unless a PCIe switch is added, complicating the server design and increasing costs.

Both Software and Hardware RAID have their pros and cons. However, when using NVMe SSD storage, applications tend to consume a large amount of CPU resources and cannot compromise on performance. The industry urgently needs a new type of RAID technology that can provide data protection while fully utilizing the performance of NVMe SSDs.

GRAID: The Next Evolution in NVMe RAID Technology

The concept of hardware-assisted Software RAID already exists, utilizing hardware like an HBA with RAID BIOS or motherboards with integrated RAID BIOS. However, these solutions still rely on the CPU to handle RAID logic, which doesn't resolve the main issues faced by Software RAID in environments using NVMe.

Now that a single NVMe SSD can reach up to 1 million IOPS, developing a high-speed hardware accelerator card that can keep up with this performance is incredibly challenging. As a result, a new solution has emerged: GRAID, which combines Software RAID technology with programmable AI chips.

GRAID works by introducing a virtual NVMe controller into the operating system and integrating a PCIe device into the system, equipped with a high-performance AI processor to manage all RAID operations of the virtual NVMe controller. This solution offers several advantages:

  • Fully leverages NVMe performance, achieving 6 million random IOPS, which is currently the industry-leading performance benchmark.
  • Unlike Software RAID, it doesn’t consume excessive CPU resources.
  • Overcomes many limitations of Hardware RAID cards, such as computational performance and PCIe bandwidth.
  • Plug-and-play capability even for systems without PCIe switches, using SSDs directly with the CPU via PCIe without requiring hardware redesign.
  • Compatible with Software Composable Infrastructure (SCI) and can be used with external SSDs connected via NVMe over Fabrics (NVMeoF).
  • Highly scalable, with the ability to easily add new software features like compression and encryption.

Results

Experimental results show that in random read tests, both RAID5 and RAID10 configurations reached the maximum performance limit of 10 NVMe SSDs combined, while maintaining very low latency. In random read and write tests, RAID10 was able to fully utilize the performance of the NVMe SSDs, with RAID5 achieving up to 1.8 million IOPS—the highest result currently achieved in the industry.

Finally, in sequential read and write tests, RAID10 read and write throughput reached 25 GiB/s and 10 GiB/s, respectively, which equals the total throughput of 10 NVMe SSDs. Even with the write penalties and checksum calculations in RAID5, write throughput still reached 9.68 GiB/s, which is very close to RAID10 performance.

Conclusion

This exploration highlights the impact of NVMe SSDs on traditional RAID technology and identifies the RAID architecture best suited for this storage medium. The experimental results demonstrate that GRAID provides robust data protection while fully utilizing NVMe SSD performance in a highly efficient and streamlined platform. Additionally, GRAID frees up CPU resources, allowing them to be used for other applications, making it ideal for diverse workloads in 5G, IoT, and AI computing.


That the contents of third-party articles/blogs published here on the website, and the interpretation of all information in the article/blogs such as data, maps, numbers, opinions etc. displayed in the article/blogs and views or the opinions expressed within the content are solely of the author's; and do not reflect the opinions and beliefs of NASSCOM or its affiliates in any manner. NASSCOM does not take any liability w.r.t. content in any manner and will not be liable in any manner whatsoever for any kind of liability arising out of any act, error or omission. The contents of third-party article/blogs published, are provided solely as convenience; and the presence of these articles/blogs should not, under any circumstances, be considered as an endorsement of the contents by NASSCOM in any manner; and if you chose to access these articles/blogs , you do so at your own risk.


images
Anuj Bairathi
Founder & CEO

Since 2001, Cyfuture has empowered organizations of all sizes with innovative business solutions, ensuring high performance and an enhanced brand image. Renowned for exceptional service standards and competent IT infrastructure management, our team of over 2,000 experts caters to diverse sectors such as e-commerce, retail, IT, education, banking, and government bodies. With a client-centric approach, we integrate technical expertise with business needs to achieve desired results efficiently. Our vision is to provide an exceptional customer experience, maintaining high standards and embracing state-of-the-art systems. Our services include cloud and infrastructure, big data and analytics, enterprise applications, AI, IoT, and consulting, delivered through modern tier III data centers in India. For more details, visit: https://cyfuture.com/

© Copyright nasscom. All Rights Reserved.