Microsoft's ReFS vs NTFS system Which is Better?

Microsoft always continues to make improvements to their systems and services. This time, Microsoft launch ReFS system as a repair system for NTFS. NTFS itself has been released by Microsoft along with when released Windows XP.

NTFS is the primary file system for recent versions of Windows and Windows Server, that provides a full set of features including security descriptors, encryption, disk quotas, and rich metadata, and can be used with Cluster Shared Volumes (CSV) to provide continuously available volumes that can be accessed simultaneously from multiple nodes of a failover cluster.

So, what is ReFS?

Resilient File System (ReFS) is Microsoft's newest file system, that designed to maximize data availability, scale efficiently to large data sets across diverse workloads, and provide data integrity by means of resiliency to corruption. The ReFS used to detect all types of disk corruption, data striping and also copy-on-write features. ReFS is a newest system of Windows based on NTFS, and in other words, this is an update system to complete the Windows NTFS system.

Basically, the ReFS system is better than the NTFS system. Because the NTFS system has several limitations that have been improved by the system. ReFS supports features that already exist in NTFS such as BitLocker encryption, access control lists for security, USN journaling, change notifications, symbolic links, junction points, mount points, reparse points, volume snapshots, file IDs, and OpLocks.

How does it work?

ReFS uses checksums for metadata and optionally for data files and has the ability to detect damage reliably. Additionally it can automatically repair detected damage using alternative data copies provided by Storage Spaces. ReFS can delete corrupt data when the volume is damaged and a copy of the data is damaged and it can handle most irreparable damage. ReFS introduces a data integrity scanner, known as a scrubber, which can periodically scans volumes, identifies damage and proactively triggers corrupt data repair. However, this system cannot be used for booting PCs or used in external media such as USB flash disks or memory cards.

Some of the advantages of ReFS compared to NTFS include the field of hyper-V and backup applications.

On Hyper-V, ReFS has some advantages:

1. The “Block Cloning” feature can accelerate copy operations, enabling quick, low-impact VM checkpoint merge operations, which in turn reduces the disk I/O load
2. The “Sparse VDL” feature can zero files rapidly, reducing the time needed to create fixed VHDs from 10s of minutes to mere seconds.

However, it’s not suitable for SAN-based cluster shared storage, at least for now, because the Hyper-V cluster will always put the ReFS CSV on “File System Redirected” mode, which caused the owner node to be the only one accessing the SAN, while the other nodes are accessing the storage over the cluster network (I’ve seen no SAN traffic flowing on the other nodes, only on the owner node). See below for more details:

This is on a ReFS CSV:

BlockRedirectedIOReason: NotBlockRedirected

FileSystemRedirectedIOReason: FileSystemReFs

Name: HVC-Prod-Disk3

Node: WOWHVC3

StateInfo: FileSystemRedirected

VolumeFriendlyName: HVC-Prod-Disk3

VolumeName: \\?\Volume{ee02dcb9-8066-458f-8fb4-cc974f8be89a}\

And this is on a NTFS CSV:

BlockRedirectedIOReason: NotBlockRedirected

FileSystemRedirectedIOReason : NotFileSystemRedirected

Name: HVC-Prod-Disk1

Node: WOWHVC3

StateInfo: Direct

VolumeFriendlyName: HVC-Prod-Disk1

VolumeName: \\?\Volume{7973af21-ee14-4d9e-aada-314e8b8d8850}\

With NTFS, all nodes are accessing the SAN storage using DirectIO, which should give us better performance.

But ReFS is okay for Storage Spaces Direct (S2D) configuration, since there is no shared storage anyway (every nodes have direct-attached disks) so all access is essentially redirected.

For Backup Applications, the advantages are as follows:

System Center DPM 2016 is actually using the resilient file system by default for the backup storage, as it utilizes the “block-cloning” technology on ReFS to accelerate the backup process. It’s a part of the “Modern Backup Storage” technology that SCDPM 2016 offers.

DPM 2016 uses block cloning to store backups on ReFS volumes. Instead of using copy-on-write to store backups (which was used by VolSnap in DPM 2012 R2), DPM 2016's block cloning uses allocate-on-write. This change improves IOPS efficiency, making backups nearly 70% faster.

Veeam also recommends ReFS for the backup storage, as it also utilizes the same “block cloning” technology on ReFS (Veeam calls it “fast cloning capability”) to speed up the backup process.

Version 9.5 leverages the fast cloning capability in ReFS API to create synthetic backups without moving the data blocks between files, and instead, references backup file blocks already present on the volume. This means all manipulations associated with synthetic full backups are limited to metadata updates and require no actual I/O operations on backed up data.

Another advantage of using ReFS for backup storage is: it enables us to scale-up the backup storage capacity up to 35 PB (compared to a maximum of 256TB on NTFS)

So, to conclude this story:

1. On Hyper-V Cluster, use NTFS on SAN-based storage, and use ReFS on Storage Spaces Direct (S2D) storage
2. On a Standalone Hyper-V server, use ReFS so that we can take advantage of the “Block Cloning” and “Sparse VDL” feature, which can accelerate VM checkpoint merge operations and fixed VHD creation. However, please be advised that a ReFS partition can be expanded, but cannot be shrunk, so if you know that you will need to shrink the partition in the future, please use NTFS.
3. On Backup servers using SCDPM 2016 or Veeam B&R 9.5, use ReFS on the Backup Disk/Storage to accelerate the backup process

Curious what Wowrack can do for you? Contact us today to learn more!