There are different types of storage, with different protocols, different architectures, different scaling, different capabilities, and also different purposes.
In a virtual environment, you will need a resilient and reliable storage solution, with the expected performance, that can scale for the future. This can only be possible using enterprise storage products, with some exceptions for the ROBO and SMB scenarios, as discussed in Chapter 2, Design and Plan a Virtualization Infrastructure.
Enterprise-class storage can be classified in different ways, but usually different acronyms are used, such as:
- Direct Attached Storage (DAS)
- Network Attached Storage (NAS)
- Storage Area Network (SAN)
- Content Addressable Storage (CAS)/Fixed Content Storage (FCS)
- Object-based storage/cloud storage
For VMware vSphere, the first three storage classes are the most relevant and the only type of solutions actually usable for running VMs, but object-based storage could be used by other solutions (such as backup products), and maybe also by vSphere in the future.
The main difference across these different types of storage are the types of services, the different targets of usage, the performance, and how they can scale, as shown in the following table:
Storage type |
Type of service |
Front-end protocols |
Data/access ratio |
Typical usage |
DAS |
Block |
SCSI, SATA, SAS |
1:1 |
Local storage (inside ESXi) |
SAN |
Block |
FC, FCoE, FCIP, iSCSI |
N:1 – 1:1 |
Shared storage |
NAS |
File |
SMB, NFS |
1:N |
File server |
Object |
Object |
HTTP, APIs |
1:Millions |
Cloud storage |
All the enterprise storage could be classified according to its architecture. A great classification is from Chad Sakac (Dell-EMC), which defines four main types of storage architecture—clustered scale-up and down, tightly coupled scale-out, loosely coupled scale-out, and distributed shared nothing. For more information, see the blog post at http://virtualgeek.typepad.com/virtual_geek/2014/01/understanding-storage-architectures.html.
To keep it simple, we can just focus on two different architectural models:
- Scale-in or scale-up: This is where the storage grows in capacity (and initially also in performance) by adding new disk shelves
- Scale-out: This is where more arrays are managed as a single logical storage performance, and capacity can scale by adding new arrays
A different point of view, more focused on the evolution of the storage market, is provided by Stuart Miniman on Wikibon (http://wikibon.org/wiki/v/Server_SAN_Market_Definition).
But a more simple classification could define different levels of the tiers used to classify the storage:
- Tier 0: This is usually not used, but sometimes define a very high-performance storage, such as the All-Flash Array (AFA)
- Tier 1 or primary storage: This is usually the main storage that corresponds to the VMware side at the different types of datastores
- Tier 2 or secondary storage: This is a storage not (usually) used from VMware, that stores online archives, backups, cold data, and so on
- Tier 3: This could be a long-term and may be offline archival storage repository, such as tapes, or copy on public cloud storage