The question is, can it do a lot better? So for example, it could run an airline, and all the computer systems, but can it do it as effectively and efficiently as it could on other systems? No. That's where you start hitting things like you could store library congress on it 50 times, but does that mean that it's usable? If you want to find a book, do you want to wait a minute to get your results back or would you rather get your results back in two seconds? ZFS can always get the job done and it's reliable but it's just not always fast enough, and that's the main problem with it. ZFS is great for just mass storage, but if you're trying to make fast storage – something like a SAN-type delivery network where you wanted to do any type of RAM disc over the network – it falls flat. ZFS does not do that. It is kind of limiting. For companies or organizations using Kubernetes, multiple Docker instances, or the spin-up machines needed to handle the workload need quick media just for the host itself to be able to run things quickly, but ZFS does not do that. What's important these days is, as you run stuff as a cluster, you want to be able to replicate from one cluster to another, whether that's done in the background or is done in near real-time. ZFS just does not do anything of that nature. You have to use a third-party product or Ceph storage, which we use also. So that's why it's limiting from that standpoint.
Director at a tech services company with 1-10 employees
Real User
2022-02-06T07:18:23Z
Feb 6, 2022
Oracle ZFS does what I've asked it to do, and it has done it very efficiently. The only time I'm running into issues is with Proxmox. If I run ZFS drives, I find my RAM usage is very high. However, I don't have that problem with the TrueNAS system, where I'm running an old N36 Turion with four gigabytes of RAM, and that's running 24/7. There have been no issues with such a low-powered environment there, it works fine, but with Proxmox it seems to go slow. The ability to upsize and downsize an array can be difficult. In a feature release, it would be a benefit to have the ability to add an extra drive onto that pool as a spare. This would allow it to automatically rebuild the array onto the standby drive if there is a failure.
The question is, can it do a lot better? So for example, it could run an airline, and all the computer systems, but can it do it as effectively and efficiently as it could on other systems? No. That's where you start hitting things like you could store library congress on it 50 times, but does that mean that it's usable? If you want to find a book, do you want to wait a minute to get your results back or would you rather get your results back in two seconds? ZFS can always get the job done and it's reliable but it's just not always fast enough, and that's the main problem with it. ZFS is great for just mass storage, but if you're trying to make fast storage – something like a SAN-type delivery network where you wanted to do any type of RAM disc over the network – it falls flat. ZFS does not do that. It is kind of limiting. For companies or organizations using Kubernetes, multiple Docker instances, or the spin-up machines needed to handle the workload need quick media just for the host itself to be able to run things quickly, but ZFS does not do that. What's important these days is, as you run stuff as a cluster, you want to be able to replicate from one cluster to another, whether that's done in the background or is done in near real-time. ZFS just does not do anything of that nature. You have to use a third-party product or Ceph storage, which we use also. So that's why it's limiting from that standpoint.
When retrieving data from the replication of remote sites, it does not give you immediate results. The RPO and RTO rates could be improved.
Oracle ZFS does what I've asked it to do, and it has done it very efficiently. The only time I'm running into issues is with Proxmox. If I run ZFS drives, I find my RAM usage is very high. However, I don't have that problem with the TrueNAS system, where I'm running an old N36 Turion with four gigabytes of RAM, and that's running 24/7. There have been no issues with such a low-powered environment there, it works fine, but with Proxmox it seems to go slow. The ability to upsize and downsize an array can be difficult. In a feature release, it would be a benefit to have the ability to add an extra drive onto that pool as a spare. This would allow it to automatically rebuild the array onto the standby drive if there is a failure.