HD from Windows Server if it fails, what to do? [closed]

Question

I first had the idea of performing a RAID 1 mirroring with two disks, but I saw that there was loss in disk write performance, and since the server is a domain/database server/files I would not want to lose performance by Fear of SQL Server having a loss of performance in the data write rate.

So I thought about RAID 5, but there is also loss of write rate. Soon after, I came to the conclusion that RAID 10 would be the ideal option for me, I would not lose performance either in writing or in reading, on the contrary I would gain performance according to documents that I read, and would have a great security in case some disk dies.

But consulting a colleague with a lot of experience in the area, in about 20 years, told me that RAID is a thing of the past and should no longer be used, already had a lot of problems with RAID in which in one case the controller stopped working and damaged the disks, and also another colleague of his had problems with RAID too, he knocks on the table saying that RAID will not give performance no matter which RAID chosen is in use. His suggestion was as follows, if I do not want to lose performance and have good security, I replace the OS HD with an SSD, which will provide me with even more speed, and is much more crashworthy, or else I would do another a server workstation, in case the server fails, put it in the server's place and use it as a substitute, performing some more operations so that it can replace the server if the OS would be experiencing problems.

In your opinion, what is the best alternative for the security of a crash in the HD that the OS is stored in, if it burns after a power failure or for another reason and damages it, what would be the best alternative for a recovery not as time-consuming as pick up a new HD and do the reinstallation of the server and all its configuration, an alternative in which I do not lose performance and in that does not harm the work of the employees of the company?

Answer 1

There's so much wrong here, I scarcely know where to start, but I'll take a shot.

since the server is a domain/database server/files I would not want to lose performance by Fear of SQL Server having a loss of performance in the data write rate

So this is your first problem, right here. It is generally a bad idea to mix server roles on a single server. Based on what I can gather from this rudimentary description, you should have (at least) 4 separate server OSes here. Two domain controllers (because it's a bad idea to ever only have a single domain controller for a domain), one file server and one database server. This doesn't necessarily mean you need separate physical hardware for everything, (virtualization as a technology is robust and mature at this point) but as a rule, you should separate your server roles onto separate sever OSes.

Having said that, it's entirely possible you could end up needing a separate server for SQL performance reasons, but there's no way to know that without further information.

But consulting a colleague with a lot of experience in the area, in about 20 years, told me that RAID is a thing of the past and should no longer be used

He's wrong, or you're misunderstanding. At your scale, there's no question that your servers should use some RAID level for their disks. Seriously, what else are you going to, populate the drive bays, go JBOD and have 8 separate volumes (or whatever number)? Furthermore, these days, it's common, and probably recommended, to separate out the OS disks and the data disks. For example, you'll frequently see servers come with a pair of disks intended to be put in RAID1 for the OS, and a large drive bay for all the other disks to be used for data (also at some RAID level, usually 10 or 6). Another common use case for virutalization hosts is to install the hypervisor on an SD card, and use the disks in the server drive bay for a large RAID array that becomes the datastore which houses the guest VMs.

In your opinion, what is the best alternative for the security of a crash in the HD that the OS is stored in, if it burns after a power failure or for another reason and damages it, what would be the best alternative for a recovery not as time-consuming as pick up a new HD and do the reinstallation of the server and all its configuration, an alternative in which I do not lose performance and in that does not harm the work of the employees of the company?

Well, the first thing here is that you use RAID to provide redundancy. It's not a backup or disaster recovery strategy. RAID may also provide performance benefits, but the primary advantage is that when a disk fails (when, not if), you can replace it and carry on without downtime.

The second thing is that you don't have a backup strategy or disaster recovery plan. What those look like, however, depend on first making other decisions about your general architecture for this server system. Proper strategies for both backups and disaster discovery basically require that the data live on a separate system, ideally in a completely different location from whatever data and systems you're protecting.

In short, you don't need RAID or backups, you need RAID AND backups. But before you get to that point, you need to design your server systems right in the first place, or you're just wasting your time and money. A single server running domain services, SQL and file services is not supported and not manageable, nor are you going to be able to find a good way to backup and restore such a thing.