NVMe SSDs are seeing rapid adoption in the cloud, enterprise data centers, and edge computing for data-intensive applications such as relational databases and NoSQL environments. However, the data in these SSDs is temporary and will persist only during the lifetime of their associated virtual machine instances.
NVMe SSD storage is ephemeral for all of the major cloud service providers including Amazon Web Services, Microsoft Azure, and Google Cloud Platform. If customers want to protect data in these SSDs, they have to opt for Elastic Block Storage in AWS or equivalent persistent disk options in Microsoft Azure and Google Cloud Platform. The problem is that these options often require sacrificing performance in exchange for data protection and availability.
To protect the data on high-performance SSDs, most server OEMs use traditional RAID controllers, which were originally designed for hard disk drives. When traditional RAID is deployed with modern SSDs, it behaves in a way that massively reduces SSD performance and endurance, while requiring additional SSD drives for redundancy, reducing available capacity.
Simply put, traditional RAID solutions limit the performance benefits of SSDs in enterprise data center deployments.
Performance issues at the edge
When it comes to edge computing, the objective is to bring data and storage closer to wherever it is needed. Storage resiliency to tolerate SSD failure at the edge is even more important due to the lack of people at edge data centers to repair or replace failed SSDs. Employing traditional RAID solutions all but nullifies the benefits of edge server performance and capacity in the name of data protection.
The Pliops Extreme Data Processor (XDP) was designed to meet this challenge. Pliops XDP provides reliable RAID data protection without any associated performance penalty, unlocking the full performance potential of SSDs by eliminating inefficient storage operations and offloading compute-intensive data processing activities from the CPU.
Pliops XDP optimizes the data flow between applications and SSDs by compressing, indexing and sorting, and then writing data to SSDs in sequential patterns with parity for data protection. XDP fundamentally improves how data is processed, managed, and stored, leading to accelerated application performance, higher reliability, and greater scalability at a reduced TCO (total cost of ownership).
Pliops XDP accelerates random writes by transforming them into compressed sequential writes. This improves storage performance, utilization, and SSD longevity compared with traditional RAID 5/6 solutions. Compression also shrinks the amount of data to be read, enabling better read throughput and latency while reducing IO amplification impacts. Unlike traditional RAID, which uses a dedicated hot spare disk, Pliops XDP Drive Failure Protection (DFP) makes use of all SSDs with a virtual hot capacity (VHC) to drive performance and capacity benefits.
The superior performance offered by Pliops XDP can be attributed to doing most functions in hardware. Line-rate compression is performed using a hardware-accelerated engine. The key-value storage engine is also hardware-based. Like RocksDB on a chip, the key-value engine is the true workhorse of XDP, performing much of the magic behind its performance benefits.
Independent engineering validations
When a new product disrupts long-held beliefs about enterprise workloads, it’s standard practice for well-known engineering and validation firms to evaluate it. One such firm, Principled Technologies, evaluated the Pliops XDP by comparing it with traditional RAID controllers in the Dell PowerEdge server used in traditional data centers with identical system and SSD storage configurations.
The objective of the Principled Technologies tests was to assess the nonfunctional requirements of enterprise applications for sustained performance, availability goals, and SLA compliance by comparing the traditionally deployed RAID with Pliops XDP-RAIDplus. The complete details of the system under tests (SUT), its specification, and reports can be accessed here. The report highlights the performance optimization benefits of XDP and how to meet SLA goals when SSD storage failure is encountered.
StorageReview, a leading independent storage authority with its own test lab, also carried out an in-depth technical assessment of Pliops XDP for enterprise IoT applications. Testing reliability, storage capacity, and efficiency while using high-density QLC SSDs were the main objectives. The StorageReview report articulates how architects design enterprise systems to achieve consistently higher performance, increased reliability, and extend QLC SSD life span by reducing the write amplification to reduce the cost of deployments.
One of the best ways to take a look at new technologies is seeing how they perform in real-life situations. Read on for two use case examples.
Billions of queries per hour
A leading software-as-a-service (SaaS) provider runs tens of thousands of MariaDB instances to serve thousands of customers by delivering billions of queries per hour. This setup was designed to maximize SSD storage performance in order to drive higher MariaDB queries, but the trade-off was the availability of MariaDB applications. Any SSD failure would result in a larger number of MariaDB instances and users failing over to secondary servers, not only affecting customer satisfaction but increasing the operational overhead of rebalancing traffic.
After employing Pliops XDP, this SaaS provider rearchitected MariaDB instances with fully protected storage and could now tolerate SSD failures without compromising on application performance. In return, the company realized additional performance and capacity gains due to line-rate compression, which enabled them to increase MariaDB instances per server.
Enabling dense, performant, and reliable AI infrastructure
Paperspace, a high-performance cloud computing and machine learning development company, faced three major challenges with their existing NFS storage infrastructure: growth, performance, and availability (see case study). Paperspace needed a denser and more performant storage solution that also offered better reliability and recovery times. The company turned to the Pliops XDP on Supermicro servers to replace RAID controllers in order to support high-performance NVMe workloads for AI infrastructure in their data centers. Pliops XDP provided Paperspace with 5x higher performance with built-in data protection and 2x faster recovery time with 33% lower cost per terabyte, compared to traditional RAID solutions.
For additional customer case studies and Pliops XDP solution briefs, visit the Pliops Resource Center.
Full performance for NVMe SSDs
Pliops XDP offers a solution for system, storage, and data architects who need to design cost-efficient data protection across edge and cloud platforms to meet the increasing demands of business applications. Customer deployments and independent engineering validations have demonstrated how Pliops XDP improves application performance, increases SSD endurance and usable life, and addresses data growth challenges with built-in data compression.
Pliops XDP allows architects to configure applications to meet availability needs, reduce application downtime, and improve recoverability. This results in lower total cost of ownership for data infrastructure in edge, data center, and cloud deployments, while preserving the high performance that companies seek in NVMe SSD storage.
Prasad Venkatachar is senior director of solutions and product at Pliops.
—
New Tech Forum provides a venue to explore and discuss emerging enterprise technology in unprecedented depth and breadth. The selection is subjective, based on our pick of the technologies we believe to be important and of greatest interest to InfoWorld readers. InfoWorld does not accept marketing collateral for publication and reserves the right to edit all contributed content. Send all inquiries to newtechforum@infoworld.com.
Copyright © 2023 IDG Communications, Inc.
