Drive Redundancy on a VPS - does it help us as consumers?
As many on here I'm sure sometimes do, I was going through the list of the VPS I've got idling away dedicated to srs tasks, and trying to figure out which ones to keep. In terms of the storage aspect of them, obviously speed is an important difference for many use-cases, but what about hardware redundancy?
Many of the bargain providers have no redundancy for their cheapest plans, which makes sense, but some have their drives listed under various RAID configurations -- some of the ones I have say they have RAID-6, RAID-10, or RAID-Z2, for example. Does that actually benefit us, as consumers of those products, where if a drive died and they replaced it, I wouldn't see data loss assuming nothing else went wrong? Or does that redundancy only help the provider in some way, thus only loosely or indirectly helping us?
(Insert obligatory comment that RAID is not a backup solution.)
Comments
Drives will fail, and it's a matter of probability whether this will lead to complete loss (which is never zero)
Different raid configurations
1) have different probability of complete loss, and thus (assuming that you always have backups) your downtime in such case.
2) provide different performance which is self-explanatory
Really depends on your backups, drive failure is real, we wouldn’t use anything less than RAID 1. Using RAID 0 on anything production is pretty terrifying the chance of failure is increased for each disk in the array.
RAID 1 half’s write speeds but theoretically doubles read speeds.
RAID 10 gives you similar performance to RAID 0 with a maximum of two tolerated failures per 4 disk array depending if the failed disks are in the mirror or stripe. A single disk failure on a RAID 10 array is completely safe.
The issue here is more down to how a host recovers from a disk failure, RAID is not a backup. There is every chance that smart hands could pull the wrong disk during replacement corrupting the array. Controller failures are also real and can completely corrupt an array.
Does having a redundant RAID level mean you don’t need to take backups? Absolutely not. But it is a layer of redundancy and can prevent data loss in the right circumstances.
Where possible you want a redundant RAID array and regular backups where data and uptime is important. In most cases disks can be hot swapped in the event of failure with 0 downtime (but slightly reduced performance during the rebuild period).
Edit: This is a very brief explanation and it’s worth reading into the various RAID technologies they each come with pros and cons. RAID 10 for example is more expensive as you get around half the storage of an equivalent RAID 0 array although you get a level of redundancy. RAID 5&6 give you similar redundancy with more storage for a sacrifice in performance.
RAID 0 gives you the combined storage of all disks in the array with the performance equivalent of each disks max read/write combined (depending on the controller) but with absolute no redundancy - a single disk failure usually causes total data loss depending on the stripe size. Misconfigration of RAID 0 with a large stripe size can sometimes have data recovered where files are not striped across disks although this absolutely shouldn’t be relied on and is a data recovery task at best.
Tldr dont rely on a single disk or RAID 0 unless the data and stability is absolutely non critical.
As long as it's not RAID0 it's fine.
Disasters can come in many ways, misconfiguration, human error, hackers, bad juju, bad karma, nuclear war, an act of God, zombie pandemic virus...
Usually disasters involve cutting of costs in both human and actual resources.
Any standard plan and not super cheap discounted promo plans should be on sone mirrored raid. Plus this also depends on how competent the host is. Mirrored raid isn't useful if the host fucks up and loses data or has extended downtime when it's completely avoidable.
Modern day NVMe failures on a cheaper brand (i.e. not enterprise cloud service, not Linode, DO, etc.) should only have a few minutes of downtime at most for the non-hotswappable drive to be replaced. No clients would notice raid rebuilds, drive failure, etc.
As a VPS customer, you shouldn't need to worry.
ExtraVM
Sounds like it really should matter then who I pick to be my remote backup target...was planning on going with one of the VPS with RAID-6, because it conveniently has large amounts of storage as well, so that sounds like a good choice.
Given the existence of NVME, does RAID-0 for performance boost really make much sense anyway? I can't really see why you'd use that as opposed to just having a single SATA SSD, which should outperform two spinning drives in RAID-0 anyway, right?
RAID 0 is popular with SSD and NVME providers as it allows them to buy storage at up to half the cost per GB compared to a redundant array whilst adding performance.
This is a bad practice and usually RAID 1 is more than fast enough for NVME. Especially in this low end market.
I don't think this is true at all. Only the providers you see giving away 8GB RAM for $30/y are going to be potentially running RAID-0, and that's far and few. RAID-0 is definitely no popular by any means lmao. NVMe costs are low compared to overall hardware costs for VPS companies.
ExtraVM
Selling a dedicated server with 1 disk should be crime.
Amadex • Hosting Forums • Wie ist meine IP-Adresse? • AS215325
Forum for System Administrators: sysadminforum.com
Maybe popular is the wrong word but it’s definitely a tactic used by certain providers selling large amounts of NVME disk space for very little money.
i would prefer choosing provider that explicitly explain on what drive configuration are they using (i.e. raid1/5/6 .etc), it's just something nice to have but not a must.. well as long as it's cheap
Fuck this 24/7 internet spew of trivia and celebrity bullshit.