VxRail satellite nodes enable customers to extend the benefits of the common operating model provided by VxRail HCI System Software to expand workloads outside the data center. Rapid digital transformation, workload expansion outside traditional core data centers, and the proliferation of 5G networks impels an immediate need for a small-footprint, low-cost, easy-to-manage infrastructure that provides the same benefits as VxRail. This infrastructure is especially needed in retail, telecommunications, manufacturing, and ROBO environments, where more data collection and data processing happens at the edge. A single-node deployment option, satellite nodes deliver the simplicity and automation of VxRail at the core data center to the edge.

VxRail satellite nodes are based on the existing E660F and V670F models. The satellite nodes have the same E660F and V670F hardware options, enabling various configurations, from a lightweight, single-socket hybrid E Series to a GPU accelerated V670 heavyweight. In addition, VxRail satellite nodes come with Dell PERC H755 to provide local RAID protection. They are managed from an existing VxRail cluster with vSAN, which, in addition to its existing workload, has the supplementary role of managing up to 500 VxRail satellite nodes. 

Secure drive access on VxRail satellite nodes can be applied using encrypted keys through the PERC card. Local key management is performed exclusively within the satellite node, where the keys are stored on the PERC card. While the PERC card is still used to restrict access to the local drives, remote key management provides enhanced protection with the use of a centralized remote key management server. The local key management option is the more affordable option; however, security will be compromised if the entire node is stolen. Remote key management requires additional iDRAC licensing, provides stronger protection in case of theft, and offers a more efficient scale-out solution.

Intel Xeon Scalable processor: Powerful processing for VxRail

Intel Xeon Scalable platforms are powerful infrastructure that represents an evolutionary leap forward in agility and scalability. Disruptive by design, it sets a new benchmark in platform convergence and capabilities across compute, storage, memory, network, and security. An innovative approach to platform design in Intel Xeon Scalable processors unlocks the power of scalable performance for today’s data centers and communications networks—from the smallest workloads to the most mission-critical applications. 

Intel Scalable Processors have up to 40 cores and deliver highly enhanced per-core performance, and significant increases in memory bandwidth (eight memory channels) and I/O bandwidth (64 PCIe lanes). The most data-hungry, latency-sensitive applications—such as in-memory databases and high-performance computing—will see notable improvements enabled by denser compute and faster access to large data volumes. 

The convergence of compute, memory, network, and storage performance combined with software ecosystem optimizations makes Intel Xeon Scalable platforms ideal for fully virtualized, software-defined data centers that dynamically self-provision resources—on-premises, through the network, and in the public cloud—based on workload needs. 

2nd Gen and 3rd Gen AMD EPYC

AMD Infinity Architecture is the foundation of AMD processor technology. It represents a thoughtful design approach to accelerate computation, access data quickly, and help protect against ever-changing security threats. AMD has brought many firsts to the market, including:

• First with 7 nm technology, enabling higher transistor density and energy efficiency
• First with PCIe 4.0, delivering 128 lanes to double the I/O performance over PCIe 3.0
• First with 64 cores (128 threads) in a single socket

AMD EPYC has been engineered for data centers that rely on CPU performance. From oil and gas exploration, in-memory databases, and big data analytics to production rendering to standard data center applications, highly parallel workloads have more cores to work with. Traditional CPUs typically must scale up to a 2-socket server to overcome an imbalance of resources. With AMD EPYC, single-socket servers satisfy many workload needs, helping increase density and reduce capital, power, and cooling expenses.

PowerEdge servers are optimized for the AMD EYPC processors by taking advantage of the additional cores, faster and additional memory channels, and PCIe 4.0 for faster networking. 

VxRail node storage disk drives

Drives that have been integrated, tested, and validated by Dell provide storage capacity for the VxRail system. VxRail configurations use 2.5-inch SSDs, 2.5-inch NVMe drives, and mechanical hard drives. There is a VxRail configuration that uses 3.5-inch drives for dense-storage requirements. Drives are logically organized into disk groups, which are configured in either of two ways:

• All-flash configurations, which contain a single SAS SSD or NVMe cache drive and NVMe, SAS, or SATA SSD for capacity drives
• Hybrid configurations, which contain a single SAS SSD or NVMe cache drive and multiple hard drives for capacity

The flash drives used for caching and capacity have different endurance levels. Endurance level refers to the number of times that an entire flash disk can be written every day for a 5-year period before it must be replaced. A higher-endurance SSD is used for write caching, and capacity-optimized SSDs are used for capacity. All VxRail disk configurations use a carefully designed cache-to-capacity ratio to ensure consistent performance. Capacity SSDs are offered in both higher-endurance SAS and SATA. The SATA SSDs are a lower-cost option, up to 30 percent less per drive, and are great for read-intensive and moderately write-intensive workloads. 

All-NVMe configurations that use either Intel Optane or NVMe drives for cache and NVMe drives for capacity are available. The E660N is a cost-effective 1U platform that can support various workloads, including data warehousing, database, and analytical workloads, with its support for NVIDIA single-width GPUs. The P580N is the only 4-socket VxRail offering. Combined with all-NVMe storage, it is an ideal platform for memory-intensive and compute-heavy application workloads.

VxRail hardware options

VxRail nodes can be configured with a choice of processor, memory, storage (cache and capacity drives), networking, power supply, and GPU (for most, but not all, node types). Customers can be assured that their VxRail is configured to best match their workload requirements prescriptively, with millions of possible configuration combinations in the VxRail Series. With best-match configuration and numerous ways to scale on demand, VxRail provides the agility demanded by today’s modern IT. Upgrade options for VxRail, including memory, GPU, SmartDPUs, NIC cards, cache, and capacity drives, expand workload use-case possibilities.

• GPUs—VxRail supports various NVIDIA data center GPUs. Depending on the GPU model, workloads such as VDI, graphics rendering, machine learning, 3D rendering, and complex visualization computing can be suitable for a VxRail cluster.
• NIC cards—As the demand for high-bandwidth network connectivity grows, VxRail is adding higher-bandwidth NIC card options. Workloads that rely on GPUs, such as AI-powered business operations, will drive more data transfer between nodes and clusters. 
• SmartDPUs—SmartDPUs combine high-performance processors and high-performance network interfaces to accelerate network and storage functions and serve as the hardware underpinning VMware's vSphere Distributed Services Engine.
• NVMe drives—With the economics of NVMe drives becoming more favorable, NVMe cache and capacity can be a cost-effective option for HPC and in-memory database workloads.
• Fibre Channel HBA—Fibre-attached storage can be used as primary storage in VxRail dynamic node clusters. Connecting to external storage arrays as secondary storage can be a valuable use case for repurposing existing investments while customers transition more toward VxRail clusters as their primary platform for virtualized workloads.