Dell Storage with Microsoft Storage Spaces Best Practices Guide
Notes, cautions, and warnings NOTE: A NOTE indicates important information that helps you make better use of your computer. CAUTION: A CAUTION indicates either potential damage to hardware or loss of data and tells you how to avoid the problem. WARNING: A WARNING indicates a potential for property damage, personal injury, or death. Copyright © 2015 Dell Inc. All rights reserved. This product is protected by U.S. and international copyright and intellectual property laws.
Contents 1 Introduction........................................................................................................... 5 2 Terminology...........................................................................................................6 3 Hardware best practices...................................................................................... 7 SMB client servers.............................................................................................................................
8 Scale-Out File Server best practices............................................................... 31 Scale-Out File Share usage................................................................................................................. 31 Creating file shares..............................................................................................................................
Introduction 1 This document describes the best practices for the Dell Storage with Microsoft Storage Spaces (DSMS) solution. The Dell Storage with Microsoft Storage Spaces solution is a software-defined storage (SDS) platform that uses Microsoft Storage Spaces and Dell servers, storage, and networking to create a scalable file server solution.
Terminology 2 • Storage pools — A group of physical disks, which allows you to manage and use the combined disk space of all physical disks in the pool. • Storage Spaces — Virtual disks (VDs) created from free disk space in a storage pool. You can configure each Storage Space VD with a specific resiliency layout and storage tiers. • Resiliency — Refers to the method by using which a VD is protected against physical disk failures.
Hardware best practices 3 SMB client servers For more information about a list of the latest validated servers for SMB client communication, see Dell Storage with Microsoft Storage Spaces Support Matrix available at Dell.com/dsmsmanuals. Recommended switches This solution is intended to provide high availability (HA). When integrating the solution in to your switch infrastructure, ensure that there exists redundant network connections between the SMB clients and the SOFS storage nodes.
Rack requirements When planning to deploy this solution, you must estimate the physical space requirements required in the rack. Dell recommends you to use a 48U rack with a depth of over 1000 mm when deploying this solution to ensure ease of cable management. The four major components of this solution are the client servers, switches, storage nodes, and the storage enclosures.
Figure 1. Example rack space usage 1. Two Switches—Dell Networking S4810 (1U each) 2. Two SMB Client Servers—Dell PowerEdge R630 (1U each) 3. Two Storage Nodes—Dell PowerEdge R730 (2U each) 4.
4 Networking best practices IP address requirements Dell recommends static IP addresses for all ports. There are seven IP addresses required for each storage node for this solution. This solution does not include the IP address requirements for the SMB clients. Subnet requirements Dell recommends you to use five distinct subnets for Scale-Out File Server configurations. Two highspeed storage networks are used for external client communication (L3–routable).
between server cluster nodes Static or DHCP address Subnet mask Static X Gateway DNS Network Subnet 5 Network Interface Cards Configure Network Interface Cards (NICs ) on all SMB clients and storage nodes for jumbo frames with the MTU frame size set to 9 KB. Also, ensure that flow control is enabled on every NIC. Also, it is a best practice to rename all ports in Windows Server based on their function by using names such as Manage, External, and Internal.
Windows Server 2012 R2 best practices 5 MPIO configuration Configuring Multipath I/O (MPIO) on the storage node enables Windows Server to use alternate paths in the event of a failure, and provides load balancing. When properly cabled and using dual-port SAS physical disks, each server has two physical paths to each physical disk. The required MPIO load balancing policy for this solution is Round Robin. Round Robin is the optimal MPIO load balancing policy for workloads with random I/O.
Feature Summary Improved SMB event messages Provides more information in the SMB event logging, which is used to assist in troubleshooting some network failure conditions. Hyper-V Live Migration over SMB Enables the use of advanced SMB features for Live Migration, such as SMB Direct and SMB Multichannel. Improved SMB bandwidth management Expands SMB bandwidth management capability to control different types of SMB traffic.
Storage Spaces best practices 6 Storage pools When setting up pools, ensure that the physical disks are spread evenly across all the enclosures by using multiple pools, rather than assigning a pool to one enclosure. For example, if you have two Dell Storage MD1420 enclosures full of HDDs and SSDs. Each enclosure has 20 HDDs and four SSDs installed. 40 HDDs and eight SSDs are available for configuring in the pool. In this example, you require a minimum of two pools.
Dell recommends you to assign the storage solution to allow for automatic rebuilds. Run the following PowerShell command to enable automatic rebuilds (configured for each pool). Set-StoragePool –FriendlyName -RetireMissingPhysicalDisks Always NOTE: Before performing maintenance operations (when storage pools are configured to automatically rebuild), you must turn off the storage enclosure. Also, you must change the RetireMissingPhysicalDisks attribute of all affected storage pools to Never.
Table 4. Pool fault-tolerance and efficiency Resiliency type Disk space efficiency Fault-tolerance for each storage pool Example enclosures with 60 x 4 TB HDD (3.64 TiB) Simple 100% 0 Disks 218.4 TiB raw disk space 218.4 TiB usable disk space Two-way mirror 50% 1 Disk 218.4 TiB raw disk space 109.2 TiB usable disk space Three-way mirror 33% 2 Disks 218.4 TiB raw disk space 72.07 TiB usable disk space Parity 1 Disk Example with column count of 7: 218.4 TiB raw disk space 187.
A new VD called exampleVD1 was created in storage pool MyPool1. This VD was a Three-way mirror so the ResiliencySettingName attribute was set to Mirror and the PhysicalDiskRedundancy attribute was set to 2. Because storage tiers are used, the disk space sizes of the tiers are specified as part of the VD.
Resiliency type Minimum number of columns Column-to-disk correlation Minimum number of disks Dual Parity 7 1:1 7 When sizing the column count of a VD, consider the following. Will the VDs be configured to automatically rebuild? When a disk fails and an automatic rebuilds are enabled, the VD attempts to repair the degraded VD by using existing free disk space in the pool.
For Three-way mirror VDs: Without enclosure awareness With enclosure awareness For Dual parity VDs: Without enclosure awareness With enclosure awareness For example, if you have three Dell PowerVault MD3060e storage enclosures each with 12 x SSDs and 48 x HDDs and decide to create two storage pools, MyPool1 and MyPool2, this is how you would calculate the column counts for the VDs in different scenarios. Table 6.
Table 7. Three-Way Mirror with enclosure awareness example Pool Name MyPool2 HDDs 72 SSDs 18 VD name columnExample2 Resiliency Three-way mirror Enclosure awareness Yes Automatic rebuild Yes Column count 4 Interleave size Storage Spaces optimizes performance by striping data across multiple physical disks. The stripe size (called interleave size) is set by default to 256 KB. This means that Storage Spaces stores 256 KB of data per stripe on each disk.
Storage tiers Storage tiers enable the mixing of HDDs and SSDs in one pool and take advantage of the faster SSDs to maximize IOPS and throughput by using the SSDs as cache. Storage Spaces constantly analyzes data usage patterns and moves the most frequently used data from the HDD tier to the SSD tier to speed up access. NOTE: Storage tiers are only supported with mirroring or simple spaces. Storage Spaces creates a heat map based on the frequency that the data is used.
Get-FileStorageTier –VolumePath Example: Pinning a VHDX file called myVHDX, which is on exampleShare on a CSV labeled Volume3. The storage tier is called MyPool1_SSD. Set-FileStorageTier –FilePath “C:\ClusterStorage\Volume3\Shares\exampleShare \myVHDX.vhdx” –DesiredStorageTierFriendlyName “MyPool1_SSD” For a correctly designed storage pool, it is recommended that you evenly spread the SSDs and HDDs across the pools and enclosures.
3wayVD2 HDD tier was 25 TiB, which uses 75 TiB of the HDD tier because of the 66 percent resiliency overhead from Three-way mirroring. The SSD tier was 1 TiB, which uses 3 TiB of SSD tier because of resiliency overhead. After both VDs are created, there is 17.72 TiB of disk space remaining in the HDD tier (10.2 percent disk space free) and 0.93 TiB of disk space remaining in the SSD tier (10.6 percent disk space free). Table 8.
New-StorageTier -StoragePoolFriendlyName “tierPool1” -FriendlyName “tierPool1_HDD” –MediaType HDD Storage tiers are sized for each VD when the VD is created. You can size the SSD and HDD tiers in the GUI during VD creation or by using PowerShell. Run the following PowerShell command for creating a VD using existing storage tiers.
For example, A new VD is created called 2wayVD1 in the storage pool tierPool1. This VD is configured as a Two-way mirror. The SSD and HDD tiers for this pool was created in the earlier example. Because the StorageTiers attribute requires objects as its input, the output of the Get-StorageTier PowerShell command is assigned to the $ssd_tier and $hdd_tier variables, and then used when creating the VD. The StorageTierSizes attribute is set to the size of each tier for which the VD uses the tier disk space.
PhysicalDiskRedundancy 1 -StorageTiers $ssd_tier, $hdd_tier -StorageTierSizes 20GB, 10TB Write-back cache Configuring write-back cache (WBC) allows Storage Spaces to use a portion of available disk space of the SSDs to provide low latency cache for incoming ‘write’ commands. All incoming ‘write’ requests are cached on SSDs and immediately respond back to the host with write completion status.
Add-PhysicalDisk -StoragePoolFriendlyName "MyPool1" -PhysicalDisks $pd -Usage Journal Adding disk space to a storage pool When planning to add physical disk space to existing storage pools and VDs there are several considerations to be made before adding the additional physical disks or storage enclosures.
enough physical disks in existing storage enclosures to spread the new data copies in a method which satisfies the enclosure awareness requirements. PowerShell command for adding a new physical disks to an existing pool. Add-PhysicalDisk -StoragePoolFriendlyName -PhysicalDisks -Usage AutoSelect PowerShell command for extending a VD. Resize-VirtualDisk –FriendlyName -Size PowerShell command for extending a volume.
Cluster best practices 7 Cluster shared volumes You must have one cluster shared volume (CSV) for every VD. You must add CSVs in multiples of the storage nodes to allow for the CSVs to be load balanced across the cluster. For example, if you have a 2x2 configuration, which has two storage nodes, you should create CSVs in multiples of 2 (for example 2, 4, 6, and 8 CSVs).
Cluster quorum A witness disk is used when there is a node or network communication failure where the nodes continue to operate but can no longer communicate between one another. The witness disk acts as part of the voting process to determine which nodes remain as part of the cluster to maintain quorum. Use witness disks on clusters with even number of storage nodes to ensure there is a vote majority in case of a splitbrain situation.
Scale-Out File Server best practices 8 Scale-Out File Share usage Scale-Out File Server shares are intended for storage of Hyper-V virtual hard disks and SQL databases. Using Scale-Out File Server shares for end user file shares, containing unstructured data such as home folders, spreadsheets, or PDFs (defined by Microsoft as information worker workloads) is not recommended.