Another ECC question

[charlespick]

I've read alot of posts about whether you need ECC or not before but I'm having trouble piecing it together for my own scenario. 

 

I run 2 Dell Optiplex's with ESXi and a Synology with Btrfs. VMs go on the nas over iscsi and there's some smb fileshares on there too. I store all my design work, including that for clients, past and current work on that nas. But there is no ECC in my datacenter at all. Should I be concerned? I heard that different filesystems needed it more. Is that true? I get the idea of ECC but like, I don't have it on my workstation or MacBook. Isn't it just as possible that my work could get corrupted on my workstation as it could while being processed in my homelab? I have Veeam backups of the vms and file shares to the same nas and to the cloud but I don't open every file every 90 days, which is my retention period, so I guess it's possible for a file to get corrupted and go undetected until the good copy in backup has been aged out? 

 

What are your thoughts on if I should invest in ECC workstations, hosts, storage, or other solutions and please let me know why. I'm curious. 

 

Charles

[CWALD]

For most average people ECC is not a concern. It will also not affect things that are already on the heard drive unless you are moving it without a backup. ECC is primarily used in servers where a a fliped bit can crash an application and cause downtime in a production environment. Especially if it is going to take the admin 20-30 minutes to restart it. If you have a server where uptime is a priority then yes you should have ECC memory. If you have a home server that serves movies and will simply have to wait for you to get home after work to restart, it probably is not a big deal. ECC should also not be a concern until you have taken care of lower hanging fruit to provide uptime such as having a UPS and figuring out how you want to manage updates.

[charlespick]

7 hours ago, CWALD said:

For most average people ECC is not a concern. It will also not affect things that are already on the heard drive unless you are moving it without a backup. ECC is primarily used in servers where a a fliped bit can crash an application and cause downtime in a production environment. Especially if it is going to take the admin 20-30 minutes to restart it. If you have a server where uptime is a priority then yes you should have ECC memory. If you have a home server that serves movies and will simply have to wait for you to get home after work to restart, it probably is not a big deal. ECC should also not be a concern until you have taken care of lower hanging fruit to provide uptime such as having a UPS and figuring out how you want to manage updates.

Thanks for the info. So it sounds like ECC is a lot more concerned with application availability than data integrity? I store a little more than movies on my home server, but downtime from a simple crash isn't a big deal. Data loss is however. This is why I store 90 days of backups in immutable cloud storage, another copy locally, a monthly offline SSD backup offsite. I also have UPS, RAID, and I think decent security practices. Again, I'm concerned with files being corrupted as I work on them I suppose. I should be good as is right?

[CWALD]

10 hours ago, charlespick said:

Thanks for the info. So it sounds like ECC is a lot more concerned with application availability than data integrity? I store a little more than movies on my home server, but downtime from a simple crash isn't a big deal. Data loss is however. This is why I store 90 days of backups in immutable cloud storage, another copy locally, a monthly offline SSD backup offsite. I also have UPS, RAID, and I think decent security practices. Again, I'm concerned with files being corrupted as I work on them I suppose. I should be good as is right?

I think you should be fine. The UPS means that things have time to save properly even if the power goes out and the offsite/cloud backups mean that even if something goes corrupt or your machine dies entirely, you still have a copy. Sounds like you already have what a lot of us strive for in backups and redundancy, well done sir.

 

That being said, since you already have the basics done. Making the application more stable with ECC memory would reduce crashes, and reduced crashes would prevent corrupting files while working on them. The exact numbers on this very with application and use case but it could take you from 99.9% file integrity to 99.99%. Again this is a difference that maters more in enterprise environments and normally would not mater to the individual user but it is there if you want it.

[LIGISTX]

22 minutes ago, CWALD said:

I think you should be fine. The UPS means that things have time to save properly even if the power goes out and the offsite/cloud backups mean that even if something goes corrupt or your machine dies entirely, you still have a copy. Sounds like you already have what a lot of us strive for in backups and redundancy, well done sir.

 

That being said, since you already have the basics done. Making the application more stable with ECC memory would reduce crashes, and reduced crashes would prevent corrupting files while working on them. The exact numbers on this very with application and use case but it could take you from 99.9% file integrity to 99.99%. Again this is a difference that maters more in enterprise environments and normally would not mater to the individual user but it is there if you want it.

So…. ECC is useful to protect data against corruption. Is corruption common? Not really? Is it worth mitigating in enterprise due to the downtime or issues it can cause… yes. Many workstations will use ECC to prevent this, most large corporations buy workstations for their employees with ECC in them for example. Just about any PC with a Xeon will have ECC, and that’s all large companies buy from suppliers like Dell or HP, and it’s specifically because spending a few extra dollars on workstations is worth it to prevent downtime or corruption. In my homelab I run ECC simply because it’s worth it to me, and the cost is not high (you can buy 4x16GB sticks of ddr4 2133 ECC on eBay for 50 bucks…), and I figure the point of my homelab and NAS that lives as a VM on it is to store my data indefinitely, without corruption. So it was worth while for me to use server gear to be able to run ECC (mobo and CPU need to support it). 
 

ECC protects data in flight between CPU and storage, or while data is being processed and stored in RAM, which at one point or another is basically all data on a PC. 

[charlespick]

1 hour ago, LIGISTX said:

So…. ECC is useful to protect data against corruption. Is corruption common? Not really? Is it worth mitigating in enterprise due to the downtime or issues it can cause… yes. Many workstations will use ECC to prevent this, most large corporations buy workstations for their employees with ECC in them for example. Just about any PC with a Xeon will have ECC, and that’s all large companies buy from suppliers like Dell or HP, and it’s specifically because spending a few extra dollars on workstations is worth it to prevent downtime or corruption. In my homelab I run ECC simply because it’s worth it to me, and the cost is not high (you can buy 4x16GB sticks of ddr4 2133 ECC on eBay for 50 bucks…), and I figure the point of my homelab and NAS that lives as a VM on it is to store my data indefinitely, without corruption. So it was worth while for me to use server gear to be able to run ECC (mobo and CPU need to support it). 
 

ECC protects data in flight between CPU and storage, or while data is being processed and stored in RAM, which at one point or another is basically all data on a PC. 

I see. What solution are you using for your nas? I currently have a Synology but am probably not getting another ever. $900 for a Celeron was not a great investment in hindsight. ECC Synology with 10g and certified drives is like $3k min. Can I put ECC in my desktop with a 13900K?

[LIGISTX]

1 minute ago, charlespick said:

I see. What solution are you using for your nas? I currently have a Synology but am probably not getting another ever. $900 for a Celeron was not a great investment in hindsight. ECC Synology with 10g and certified drives is like $3k min. Can I put ECC in my desktop with a 13900K?

My homelab parts are in my forum signature. It’s all used enterprise gear I got on eBay. Without the cost of drives included, it was about 600 bucks. CPU was ~70, 128GB of RAM was ~200, mobo was 200, brand new 980 boot drive was 90, nvme card I needed was 40, HBA and sas expander were probably 100, new noctua heatsink was 90, and 10GB card with the fiber and transceivers was probably 100. 
 

I didn’t buy it all at once as I used to use a different mobo/cpu/ram, but homelab is an always evolving hobby. I run proxmox as my hypervisor and have truenas virtualized which works as my nas. 
 

Intel 13th gen supports DDR5, which has partial ECC as default. It doesn’t have any error checking between CPU and RAM, but has error checking for data while it is in RAM. But no you can’t use true ECC on Intel consumer stuff (you can, but I think only on i3 and maybe i5…? But would need a mobo that supports it as well). Basically, no. But you probably don’t need to worry about it anyways.