Design a storage architecture

Designing a storage architecture

SharePoint Server farms are data-intensive, requiring both large storage capacities and solid in input/output (I/O) storage design to ensure the best performance possible. This back-end storage is attached to the SQL (data tier) instance of the SharePoint farm.

Three storage architectures are supported within a SharePoint environment: Direct Attached Storage (DAS), Storage Area Network (SAN), and Network Attached Storage (NAS). DAS and SAN storage architectures are fully supported; NAS storage is supported only for content databases configured to use Remote Blob Storage (RBS).

MORE INFO DESIGNING THE STORAGE SUBSYSTEM

Planning and configuring the storage layer of a SharePoint farm infrastructure can be a complex task. For more detailed information about each storage architecture, see the TechNet article “Storage and SQL Server Capacity Planning and Configuration” at http:// technet.microsoft.com/en-us/library/cc298801.aspx.

Each storage architecture has an associated set of hardware and administrative costs; the storage type you choose often has to do with the hardware and administrative structure you have available within your enterprise.

Direct attached Storage (DaS)

DAS describes an environment in which each server maintains its own discrete storage without benefit of a storage network. Modern servers support two distinct types of drives: Serial Attached SCSI (SAS) and Serial Attached ATA (SATA).

Choosing this type of storage has the following benefits and drawbacks:

■■ The initial setup and administration of DAS is pretty straightforward, requiring only the attachment of the storage hardware to the server(s).

■■ The future storage growth of this solution is limited to the available interconnects to the server’s motherboard:

■■ After there are no more connections available to the server backplane, the server must be replaced to allow for future growth.

■■ After the I/O capacity of the motherboard is saturated, the server must be replaced with a larger server to allow for future growth.

■■ Long-term administration of the storage subsystems must be completed on a perserver basis, with no provision for centralized storage.

■■ Any remaining space available in the storage subsystem is unusable for other servers in the farm/enterprise.

Storage area Network (SaN)

SAN utilizes a storage network to abstract the storage subsystem from the server hosts to which they are attached. The benefits of this abstraction are immediate: the storage subsystem can be centrally managed and expanded as desired.

The Fibre Channel connections between the storage and host are attached using either twisted-pair copper wire or fiber-optic cables. A host connected to the SAN uses a host-based adapter to transfer SCSI commands to the storage using the Fibre Channel Protocol (FCP) for transport.

Choosing this type of storage has the following benefits and drawbacks:

■■ Available storage within the SAN can be assigned as required to any participating host because there is no longer a 1:1 relationship between storage and host.

■■ SAN connectivity eases the configuration of cluster storage, enabling storage to be presented to and more easily moved back and forth between cluster nodes.

■■ The initial setup of the storage layer requires in-depth administrative knowledge of the storage vendor subsystem utilized in the SAN.

■■ Incorrect configuration of the SAN can have a far-reaching effect; degraded performance or corruption can occur when improper configuration occurs.

■■ SAN I/O metrics should be watched carefully to ensure that performance is not degraded by saturation of SAN connectivity.

Network attached Storage (NaS)

NAS storage provides file-based data storage to other devices on the network. Connectivity and I/O metrics from such a system are often subpar when compared to DAS- or SANconnected storage. As such, this type of storage is supported only for content databases that have been configured to utilize RBS.






EXAM TIP







 

all network storage connected to SharePoint is required to meet two criteria. the first criterion is that the storage must respond to a ping within 1 millisecond (ms), meaning that the storage will most likely be located within the same datacenter as the host. the second criterion is that the first byte of requested data must be returned within 20 ms (this is true regardless of the disk subsystem chosen).

Disk and RaID types

The types of disks chosen can have an effect on the performance of your storage subsystem. Additionally, the redundant array of independent disks (RAID) configuration of the drives can have a dramatic effect on the performance characteristics of storage.

SharePoint 2013 supports several types of disks:

■■ Small Computer System Interface (SCSI)

■■ Serial Advanced Technology Attachment (SATA)

■■ Serial Attached SCSI (SAS)

■■ Fibre Channel (FC)

■■ Integrated Drive Electronics (IDE)

■■ Solid State Drive (SSD)

Without going into on-disk caching, rotation speed, or other in-depth storage tuning discussions, you can pretty much break down this list in terms of newer, faster drive technologies (SSD, SAS, SATA, FC) and older legacy technologies (SCSI and IDE). Often, the type of drive you choose will simply come down to the available interface types provided by your storage subsystem.

SharePoint 2013 supports all RAID types, but the recommendation for best performance characteristics is to implement RAID 1+0 (also known as RAID 10).

This RAID type configures drives in a striped set of mirrored drives—the mirroring component provides fault tolerance, and the striped component maximizes performance. In such a system, multiple drives can sustain losses but the RAID does not fail unless a mirror loses all its drives.