About a year ago, I moved one of my main bare metal servers from a datacenter in Finland to a datacenter of another provider in France. Like the old server, the current server has two 1 TB SSDs. Now that a year has passed, I was wondering how much data has been written to the SSDs by the applications running on it. While the number itself is interesting, the second interesting thing is how much life the SSD has left if my usage pattern remained the same. So let’s have a closer look.
Write Statistics
Like the title suggests, the overall write activity across the two SSDs as reported by ‘smartctl -x‘ was 43 TB over 13 months. I’m using ZFS across the two drives in a slightly strange fashion (for details see here), so it’s not a surprise the writes are not evenly split. One drive with a smaller part assigned to the single ZFS pool has seen 17.6 TB of writes while the second SSD I exclusively use for ZFS has seen 26 TB of data written.
So What’s Running On the Server?
That sounds a like a lot of data written, and indeed it is, but I have a lot of stuff running on that server:
- This blog
- Two Nextcloud instances
- Two OnlyOffice instances
- A CommaFeed RSS news aggregator
- A prosody XMPP server
- A BBB video conference server
- An Etherpad instance
- A database application
- A video processing project
- A Notify push server
- A number of internal services.
Yes, it’s all running on one physical server with 2 TB of SSD storage and 32 GB RAM in virtual machines and containers, so the server is not just idling around. According to the SSD specs, each drive can handle 400 TB written. At a rate of 26 TB per year on that second drive, that would be 15 years. Fortunately, I had a smartctl log from the beginning which showed that the same amount of data was already previously written to it. So in other words, I’m still good with those SSDs for another 13 years. Smartctl also shows a wear counters for both drives which have jumped from 1% to 4% for the first drive and from 3% to 14% for the second drive. That roughly reflects my numbers above as well. And personally, I doubt that I will use the server for another 12 years. After all, the server itself is already well over 10 years old.
Read Access
Equally interesting is how much data was read from the drives during the past 13 months. From the first drive 6.8 TB was read, while from the second drive, 5.6 TB was read, i.e. a total of 12 TB was read. This means, twice as much data was written than read. I don’t have a full explanation for this. Could it be that the logging done everywhere on the server and saved to files but rarely read could be the reason for this? That would be quite hefty. Still, no problem, I’m still good to go on those SSDs for at least another 12 years.
Summary
So much for today from my main server. I have a number of other servers that run fewer services and I will have a look at those next. But I can already say that it will not be the old age of the SSDs memory cells that will make me migrate to something newer eventually. Whenever that might be.