Workloads such as high-performance transactional databases, high-volume analytics, artificial intelligence (AI) and machine learning (ML) stand to unlock new value for the business, and non-volatile memory express (NVMe) promises to deliver on these levels of performance. However, deploying NVMe at scale to facilitate agile cloud-service like delivery of workloads, such as threat and manufacturing analytics, while keeping costs in check may stress traditional storage architectures to their breaking point.
For its part, Vexata designed the VX-OS software architecture to extract and utilize all the performance available from NVMe storage media. The VX-OS platform disaggregates the storage control and data paths, and then applies an acceleration offload engine to facilitate consistent low latency and high throughput. This acceleration offload engine consists of active-active I/O Controllers that handle host I/O traffic providing hardware acceleration with minimal latency. The IOCs access SSDs through scalable, low latency lossless Ethernet fabric. The architecture supports RAID-5 and RAID-6 groupings, as well as enterprise data services including thin provisioning, snapshots, clones, data at rest encryption and pattern removal. The VX-OS platform may be deployed stand-alone software on industry standard servers or bundled into the Vexata VX-100 Scalable Storage Systems.
The result of this architecture is faster application performance. For example, Vexata claims its VX-100F array has read and write latencies of only 90 and 35 microseconds, specifically, alongside max throughput performance of 80 gigabits per second (Gbps) and eight million input/output operations per second (IOPS).
Vexata’s key differentiator is enabling distributed, low latency parallel access to SSDs for multithreading. As a result, it has built traction serving traditional Oracle databases, distributed databases such as NoSQL, analytics, AI and ML. Most recently, Vexata announced its VX-Cloud Data Acceleration Platform to bring this value proposition to cloud environments.
Introducing Vexata’s VX-Cloud Data Acceleration Platform
The Vexata VX-Cloud Data Acceleration Platform adapts the VX-OS architecture for large cloud service providers, and for large enterprises looking to host private cloud services at massive scale. Whereas the VX-100 is a turnkey hardware platform, VX-Cloud opens up the possibility of using a variety of hardware solutions; ideal for service providers. The IOCs integrate Field Programmable Gate Arrays (FPGAs) for additional performance acceleration that can scale across multiple petabytes (PB) of NVMe capacity. According to Vexata, the result is up to a 20 times improvement to IOPS and bandwidth at consistent low latency for random, mixed read/write traffic when compared to premium SSD tiers currently available from the leading public cloud providers, at a lower upfront cost.
VX-Cloud delivers block and file services, and it is deployed on industry-standard servers that can scale both within and across racks. Unlike the VX-100, the VX-Cloud can scale capacity and performance independently through adding or enabling controller and acceleration nodes into the configuration. Core data services including encryption, thin provisioning, in-line data reduction, input/output (I/O) monitoring, analytics, and volume management are retained. Furthermore, it supports RAID-6 configurations for data protection and the FPGAs can be deployed in clustered configurations for high availability failover.
VX-Cloud supports common virtualized and physical operating systems and integrates with key database, analytics and AI software platforms such as Cassandra, Oracle RAC, Postgres and SQL Server. It is delivered as a reference architecture through hardware vendor partners (currently Fujitsu and SuperMicro). VX-Cloud is currently in directed availability, with general availability planned for the second half of 2019.
So far, Vexata has made many inroads into scale-up database environments, such as NoSQL and, (as Storage Switzerland demonstrated in a recent Lightboard Video), VMware, on the back of its ability to maximize throughput and minimize latency. Now, with VX-Cloud, Vexata is bringing its ability to optimize performance of next-generation analytics and autonomous workloads, and to consolidate mission-critical databases at scale with improved utilization, to high performance cloud environments. Storage professionals should look for reference architectures and proof of concept as the solution moves to general availability, for applicability to their own next-generation workload sets and initiatives.