DFS

Introduction

I like to have quite a bit of online storage available which is used for a number of purposes. In addition to the storage of user files, the network hosts computer backups, serves as a repository for media servers, etc. Although a single server can have some degree of fault tolerance, including hardware RAID and redundant power supplies, I do not want to be reliant on a single file server, so have chosen to implement a two node Distributed File System (DFS). In my setup, I have 2 x Windows Server 2012 DFS nodes with the file shares replicated between them.

NB : Whilst DFS provides a level of file redundancy, it is not a substitute for having a good backup, for example, if a file is accidentally deleted from one DFS node, file replication will delete if from the other DFS member server(s). If you need to recover the file, you need to retrieve it from a backup. Similarly, DFS does not provide protection from overzealous anti-virus programs or ransom ware attacks - off-line backups are still required.

Hardware

The hardware details of my DFS servers are shown on the Server Hardware page and consists of a pair of rack mounted Buffalo Terastation WS5400 servers, each with 4 x 2 TB SATA disks configured as RAID 5 arrays. Each array is partitioned with a 100GB system disk running Windows Storage Server 2012 and a 5.2GB data disk.

These servers are a relatively low cost way of implementing quite large file stores, but they are not the fastest machines in the world. In normal operation, file storage and retrieval performance is more than acceptable, but disk intensive operations, such as initial replication, can be quite time consuming and disk performance suffers during such operations.

The biggest issue though, is if a disk fault or power loss occurs and the server must rebuild a degraded array to restore fault tolerance. In a RAID 5 array, a single disk can fail with no loss of data; the disk can be replaced with a new one and the system will rebuild the array to restore full redundancy. However, this can take an age with these servers! The time required for Rebuild and Resynchronisation of the array is of the order of 10 hours per terabyte, leading to well over two days required to rebuild the disk. During this time, another disk failure would result in the complete loss of data from that server.

 

DFS1	Buffalo Terastation 5400R (WSS 2012) Processor Intel® Atom™ Dual Core 2.13 GHz Memory 4 GB DDR3 Disks 4 x 2 TB SATA (RAID 5) Network 2 x GigaBit
DFS2	Buffalo Terastation WSS 5400RR2 (WSS 2012 R2) Processor Intel® Atom™ Dual Core 2.13 GHz Memory 4 GB DDR3 Disks 4 x 2 TB SATA (RAID 5) Network 2 x GigaBit