Describe Azure Stack HCI storage drives
As the lead systems engineer, you’re tasked with researching and evaluating Azure Stack HCI software-defined storage. Your first step is to understand its architecture and explore its main components, beginning with storage drives.
Describe Azure Stack HCI storage requirements
Azure Stack HCI storage combines and uses several interrelated groups of technologies. These include:
A file system, which provides the core functionality for storage volumes.
Failover clustering, which provides core storage resiliency capabilities.
Network hardware, which supports intra-cluster and external storage traffic.
Storage-related network protocols that implement and optimize intra-cluster and external storage traffic.
Software-defined storage components, including Software Storage Bus, Storage Bus Layer Cache, Storage Spaces, and Storage Spaces Direct. These form the abstraction layer over physical hardware, resulting in simplified configuration in addition to optimized resiliency and performance.
Storage hardware that provides different performance and capacity characteristics.
You can identify the components representing these technologies in the following diagram, which illustrates the basic Azure Stack HCI software-defined storage architecture. Virtual machines (VMs)–based workloads store their files on Resilient File System (ReFS)-formatted Cluster Shared Volumes (CSVs), using storage spaces hosted by a failover cluster. Cluster nodes are connected by high-speed, resilient Ethernet and Remote Direct Memory Access (RDMA) networking. Software Storage Bus provides access to a storage pool composed of different types of drives.
The physical drives used by Azure Stack HCI are subject to the following rules:
Azure Stack HCI supports drives that are attached directly to individual cluster nodes.
Drives can be internal or hosted in an external enclosure, providing they’re connected only to a single server.
Each server in the cluster should use the same number of drives, and each drive should have the same model, size, and sector size.
There is no support for Redundant Array of Independent Disks (RAID) controllers or SAN-based drives, shared disk enclosures connected to multiple servers, or any form of Multipath I/O (MPIO) where drives are accessible by multiple paths.
It’s possible to use host bus adapter (HBA) cards that implement the simple pass-through mode.
Identify drive types for Azure Stack HCI
Azure Stack HCI supports four types of non-volatile drives:
Hard disk drive (HDD). HDDs are rotational, magnetic disks. Their key benefits are their low price and support for large disk sizes. Their primary disadvantage is susceptibility to physical wear inherent to their design, which leads to relatively frequent failures. This translates into lower mean time between failures (MTBF).
Solid-state drive (SSD). SSDs connect via conventional Serial ATA (SATA), or serial attached SCSI (SAS) connections common to HDDs. SSDs rely on integrated circuit assemblies to persist data and commonly include flash memory. The biggest advantage of the SSD is the speed at which data can be written and retrieved. In addition, the use of electronic components in lieu of physical disks results in higher MTBF.
Non-Volatile Memory Express (NVMe). NVMe refers to SSDs that attach directly to the PCI express (PCIe) bus. NVMe offers higher input/output (I/O) operations per second (IOPS) and I/O throughput with lower latency when compared with SSDs.
Persistent memory (PMEM). PMEM is a non-volatile dual in-line memory module, which justifies its nickname of storage-class memory. It offers the best performance out of the four drive types supported by Azure Stack HCI, but that's reflected by its considerably higher cost and limited capacity.
Note
The non-volatile designation means that a drive retains its contents when electrical power is removed.
Describe drive type pricing, performance, durability, and capacity characteristics
The choice of the drive types impacts the Azure Stack HCI storage performance and capacity, based on the performance and capacity characteristics of individual drive types. However, that relationship is also affected by the caching mechanism, which constitutes an integral part of the Storage Spaces Direct technology used by Azure Stack HCI.
The following diagram provides a comparison between non-volatile storage technologies available in Azure Stack HCI—HDD, SSD, NVMe, and PMEM. Horizontal bars represent relative values of price per gigabyte (GB), throughput and IOPS, MTBF, latency, and maximum capacity. Overall, latency and maximum capacity are highest in HDD, and lowest in PMEM. All four technologies have relatively high MTBF values.
