Interesting! Maybe this will be a way to expose storage activity metrics up through to our Grafana dashboards (without pulling those number direct from OS)?
No, this is a gimmick that reads from /proc/diskstats on Linux. This does not need to be in the storagenode software. I wish they stopped wasting company resources on gimmicks.
I could see a Select operator: who always wanted to make sure his ‘real’ customers had plenty of disk IO… wanting a node to reject uploads if it detected a disk has been persistently busy. Or maybe having compaction events always look at load first too?
Basically make very sure Storj is only using idle/leftover iops…
I’m talking about regular Select: where the SOC2 datacenters trade custom payouts for the removal of some restrictions (like no /24 block, so they fill faster). They can reclaim space at any time for regular customers paying more for it (like they rented a dedi / vps / whatever).
I think what you’re talking about (where Storj rents servers themselves) was what they called Surge nodes? I’m don’t think we ever heard how many there were (or if they even still exist). Cool idea though!
“Data sets with specific requirements stored on nodes in this program are not subject to the /24 restriction, allowing them to fill their nodes more rapidly”
Ah you may have found the nuance: Select has their own customer pool… but having nodes that are “used” may not mean they’re fully-paid-for.
I guess it could even mean the contract covers minimum payouts? Like if they offered 1PB then at least 10% may be paid for (even before it’s filled)? Above my pay grade
They definately could be steering regular/global uploads to Select SNOs with lots of free space: as a way to always be using the lowest-payout storage as possible.
Then when a normal Select customer comes along: willing to pay the higher price: they give them the SOC2 space and simply repair that global-data-in-select-nodes back to the global network.
It would actually be really clever if they’re doing that: great idea
A third-party owns the datacenter, and the servers, and the network gear. They maintain their own NOC staff, and SOC2 audit reports, and minimum capacity online. They own everything independent of Storj: I think it’s reasonable to call them Select SNOs.
Like… my nodes run the vanilla update logic: so Storj decides what version of the node I’m on, and even controls when they restart for those upgrades. Storj Satellites decide what data gets sent to the node, and when it’s deleted. I could see an argument that the “nodes are operated by Storj”… so maybe I shouldn’t be called a SNO?
And yet… ownership of the environment matters. Which is why everyone here still calls themselves a SNO. So I think Select SNOs exist
