Having already written articles comparing various Distributed Storage platforms, and a two-month evaluation of Sia from the renting perspective, I decided to experience being a Storage Provider to evaluate that side of both platforms to be able to give a fair overall evaluation.
In mid-2019 I had deployed a Server to handle my own varied storage requirements. In December 2019, I set about using some of the spare capacity to become a host on both Sia and Storj platforms.
I varied storage capacity (and pricing on Sia) throughout 2020, so I will attempt to take this into account and give an average overall picture.
Disclosure: I am an Ambassador for the 0chain project (which is currently in its betanet phase). I plan to use this very server when the mainnet goes live.
Sia
Running since 2015, Sia was really the first working Distributed Storage platform.
Requirements
Sia requires that Storage Providers commit 'collateral’ in its native SiaCoin to show commitment to their Service Provision. It aims to use spare capacity on Personal Computers and Servers around the world. There is generally a modest hardware (CPU/RAM) requirement, however, the entire Blockchain 'consensus' needs to be downloaded before you can start hosting. This process is CPU intensive, as the consensus now exceeds 20GB and can take hours or days to sync depending on connection speed, drive speed and CPU power.
Storage Capacity and Pricing
Initially I set my Storage prices low, a little below what was calculated to be the average, wanting to experience having storage utilized and gaining a good reputation. I also used the recommended 2x collateral.
My Performance on the network
As you can see by the comment from one of the Sia founders, my host has always been one of the best if not the best performing host on the entire network.
After a few months, the first partition I was using for Sia was getting full, so I added additional spare storage in a couple of other partitions.
A month or two later, these were also filling up and by now I had served thousands of contracts with zero failures. I decided to reduce the sizes of these other storage partitions.
At this point, I started getting several failed contracts. I checked in the Sia discord community, but it seemed that it was acknowledged there were some bugs that were hard to diagnose and since they were small amounts, I didn’t worry about them. However, after a while, some large contracts started failing, and because my collateral ratio was 2x, I was losing double my potential earnings so I had to do something about it.
After consulting the Sia community discord, I decided to stop accepting storage contracts, but stayed online in order to fulfill the remaining contracts, otherwise I would lose the collateral from those too.
As you can see from the chart, my used storage declined until early October. I also had some downtime because Sia had crashed and I did not notice.
I then started from scratch with a brand-new node and wallet. (this time choosing an equal collateral). However the IP address is the same so the stats graphs look like a continuation of service.
I had more downtime later in October. Interestingly I didn’t notice any failed contracts because it was a new node, maybe none expired during this downtime.
This actually illustrates a weakness of the Sia protocol, providers can go offline for the bulk of a contract then come back online and collect the storage fee when the contract ends. They lose potential read fees during this time of course, but in the meantime the client will have had to form new contracts with other hosts to maintain the desired EC ratio.
Within a few weeks, my node was getting full again, so I decided to just increase my prices. I expected a sharp drop in new contracts, but they continued to climb. I increased twice more, even without increasing collateral. Amazingly, even after massively increasing prices I still got some new contracts! However, the rate did decline sharply as contracts expired and decided not to renew.
This illustrates the drastic shortage of decent hosts on the Sia network. My node should not get selected for any new contracts with such a huge cost with such tiny collateral.
Drawing 2020 to a close
In December I got banned from Sia Discord for expressing my views about an unrelated subject. It looks like I am not the only one though, RBZL, a veteran Sia community member and creator of SiaSetup.info was also banned for expressing his opinions about Sia. Read :- https://siasetup.info/concerns-about-sia-and-skynet for more info.
This ended my interest in the Sia project.
Earning Summary for 2020
Average Storage: 5TB
Total Earnings: 30,000 SC ($100) estimated
Average Earning/TB/mo: $1.70
(As previously stated, the earnings could have been much higher if I had priced more aggressively)
Storj
With Storj, the prices are fixed. The determination of which hosts are used is done by centralised entities, called satellites. I do not know the criteria but assume good performing hosts are favoured.
No collateral is required to host, but a portion of rewards are held back in escrow, decreasing with time, (mentioned later).
For most of the year, my Storj host offered over 5TB of storage. It took a few months for this to get filled.
When a host becomes full, the ingress naturally falls off. In June, I took the opportunity to try the Graceful Exit (GE) procedure. (There are other reasons mentioned later).
As you can see, over 80% of my data was on test satellites (Stefan-Benten, Saltlake and Europe-North).
So, I started with the Stefan Benten satellite in Europe as that was a test satellite that would soon be discontinued anyway. The GE only took a few days and was completely successful.
I then GE’d the Asia East and US Central satellites. The Asia worked without issue, but the US Central failed even though it had an audit of 100%. (I have not investigated this because the held amount held is small anyway).
I then decided to GE from Saltlake in early September. This is another test satellite and used by far the most storage. The GE took over 6 weeks but was eventually successful. For November and December, I reduced the capacity to 2.5TB because I knew the two remaining satellites would take a while to fill the space vacated by Saltlake.
Average Storage used: 4TB
Total earnings: $220 (appx 500 STORJ)
Average Earning/TB/mo: $5
This is actually better than I was expecting. Even though the egress is typically only around 10%, since it is 13.3x more profitable than storage, the egress made the lion’s share of the earnings. But before you rush out and buy hard disks, allow me to elaborate.
The first issue..
These figures also include ‘Surge payments’. These are arbitrary amounts that were devised to be paid to hosts. These were awarded in January, March, April and November and equated to around a quarter of the earnings. So, excluding these, the average would have been around $3.75TB/mo.
The big issue..
The vast majority of data on the Storj network still appears to be test data. Although testing is crucial, this level is not sustainable. When I started Gracefully Exiting in June, around 80% of data was on the rest satellites.
Since then, I do not have a true picture on my node because I am only on some of the satellites. However, I believe the ratio to still be around the same, looking at the forums and also a discussion I had on a discord group.
Some additional points
- It takes time to get vetted. You won’t get a lot if data during this period. I am sure this only took me a couple of weeks but the latest I’ve read is that it currently takes a couple of months.
- If it were not for the fact that I Gracefully Exited some of the satellites, there would be significantly more still held back. A percentage is held in escrow. *Correction* The percentage held back starts at 75% and decreases by 25% every 3 months. (Half of this rolled forward amount is returned after 15months). When you Gracefully Exit, you get your held escrow amount back from that satellite. The default minimum period is 15 months before you can Gracefully Exit, it is temporarily reduced to 6 months, so all this was another reason I chose to Gracefully Exit some satellites. Additionally, for each new satellite, your count starts from zero to when you are able to GE.
- One of my Graceful Exits failed. As I stated, I have not investigated as the amount was small. Others in the forums have had to produce all sorts of logs to try to investigate when this happened to them. The general consensus last time I looked was that Graceful Exit is still not a risk-free procedure.
- Storj payments are monthly and via it’s ERC20 Storj token. Three times since mid 2020, Storj have changed the rules regarding disbursements, mainly because of high ETH gas fees. Many people falling below their threshold (including my December payment) are rolled forward indefinitely until the threshold is met. This is particularly detrimental to those who considered committing a relatively small amount of storage.
Summary
Although both platforms offer an opportunity for people with spare storage to earn some money from it, neither platform gives me the confidence that it could turn a profit long-term. However, they still remain viable for spare existing storage.
Most would not argue with this about Sia, as it has always maintained not to invest in new hardware as that would likely never turn a profit.
However, I suspect that some Storj providers may argue that they are happy with the income that their host(s) generate. I would ask them to reconsider in light of the following:-
- As described above, test data rate is completely unsustainable.
- Surge Payments also give a false impression of real income.
- Hard Disks are obviously the most reliable when new. When disks start failing and potentially held escrow funds are forefeit, will it still work out worthwhile? (Storj is very tolerant of downtime but intolerant of Data Integrity Failure.)
Conclusion
These two platforms have failed to realize the full potential of distributed storage. By not having sensible measures to ensure and reward high quality hosts, high EC ratios have been required to offset the host churn and no efficiency has been gained over typical DC triplication of data.
Sia, due to longstanding, unfixed errors and Storj due to a lack of transparency have failed their loyal followers, many of whom run hosts much more for the community than for any hope of ever turning a profit.
Fortunately, there is an alternative.
High quality hosts are welcome to consider joining the 0chain network.
Through true decentralization with high efficiency and advanced protocols, hosts can expect a reward that reflects their true performance on the network.
Poor quality hosts need not apply.
Footnote
Please read my previous article comparing various Distributed Storage platforms if you haven’t already for further key differences between these platforms.
No doubt I will be criticised and accused of writing this solely for benefit of 0chain. I did all this research off my own back and was not paid anything for it.
The reason I conclude that 0chain is better is for reasons described. I believe that 0chain protocols and tokenomics are so much more efficient than either Storj or Sia, and that’s not even mentioning other superior technology like private sharing through Proxy Re-Encryption.
I welcome feedback and will publish corrections if I have stated anything shown to be inaccurate, but please stick to facts and not misplaced loyalty in an inferior technology.
