Tardigrade test results from may to august 2020

Do you know the legend of how the bistro was invented?

dont know.

Then we will account on that expirience.

2 Likes

LOL!
ā€œBestā€ supportā€¦ after 5 days got this:

Hello Zzzy ŠŃ€Ń‚Ń‘Š¼,
I have been informed that space has been freed up by the Zombie reaper so your usage is now down to about 640GB which should allow you to continue with your file transfers.

Best regards,
Helene Unland
Storj Labs Support

How much unqualified staff work? How am I supposed to upload 4TB of files with a 1TB quota ?! :face_with_symbols_over_mouth:

After such an attitude towards clients, I will never use such a service.
SNO early 2017 - july 2020 :broken_heart:

Do you mean that dissatisfied customers do not need to be served?

Most of us are waiting for real clients with money to appear. And in fact, real fans are running away. So where do customers come from with this approach. Personally, I am as disappointed as possible with the client side. And Iā€™m thinking of dumping all investments in the token.

3 Likes

Iā€™m saying Iā€™m not sure the labels ā€œreal customerā€ or ā€œreal fanā€ have ever applied to this particular case. From what I know serious support requests from paying customers are promptly responded to and resolved. But some people will never be pleased. Like the parabel of the boy who cried wolf, the cries tend to lose their meaning.

And yes, from a SNO perspective Iā€™m definitely looking forward to more paying large scale serious customers. It just doesnā€™t feel like thatā€™s what weā€™re looking at here though.

1 Like

I have been informed that your limit increase request was declined. Please let us know if you have any other requests or if we can help you with anything else.

Real face))))


Unusable project

So with $48 != iā€™m not ā€œreal customerā€?
With this approach to customers, I as a node operator doubt the success of the product.

1 Like

and back to my question, why limit people who want to pay.

Looks like that 50 bucks can be spent with great sense on a backblaze. Competition with traditional clouds looks like mockery now.
Maybe StorjLabs wait clients like NASA, who know.

2 Likes

from my expirience every client who want to pay money is good enouth, more problematic clients are usualy who waht to get all for free.

2 Likes

backblaze realy cool, but my choice is scaleway (2 eur per TiB storage)

looks a lot like my hdd graphsā€¦ and thats not a bandwidth or iops issueā€¦ thats because they are performing like they are suppose to, they finish and am ready for the next task

and maybe something to do with that they are so many working togetherā€¦ after all my pool uses 13 drivesā€¦ so there would be a certain harmony to the system, so everybody isnā€™t just yelling all at onceā€¦


looks like an interesting test, but what exactly it meansā€¦ in regards to the speedā€¦ ill leave the deletion and what not for the programming inclinedā€¦

i would maybe try and 100mbit connection or 10mbit connection and see if it shows the same behavior, because i think itā€™s some sort of IO latency, generally i like to see my system having the spikes, because it means it keeps up when it gets a work taskā€¦ if they start to merge together, then the workload is so excessive that the system doesnā€™t have time to waitā€¦

so from my understanding ā€¦ the spikes are kinda a health sign meaning either the network or the computer you are using is just waiting for more workā€¦

ofc over a network the io latency and such might have some sort of effect, but would be interesting to see if 1gbit simply just gives you a max transfer speed of 368 mbit because of network IO latency

also how do you get 386mbit avg to be 15% of 1000mbitā€¦

the overhead still counts no matter how you want to look at itā€¦ throughput of your network has to deal with it, there is no free lunch

anyways, a few of my views on the topic, wrong or right as they maybeā€¦

386 = real and 140-150 is effective (excl. 2.75 multiplicator)

Spikes from uploading only single segment per moment.

well since your network registers the 2.75 multiplication, then that is obviously done locally before the data is sentā€¦ so yeah that would mean that uploading is reduced to about 1/3 of max internet bandwidthā€¦

seems like a heavy penaltyā€¦ ofc i doubt there is any way around that, since itā€™s like peer to peer traffic when moving data aroundā€¦

a cost of doing erasure coding i guessā€¦

but yeah getting 150mbit out of a 1gbit connection is quite lowā€¦

i would be very interested to see if you get 15mbit avg on a 100mbit capped internet connection alsoā€¦

because then with most regular internet connections people might run into a real issueā€¦ since if we say even a 30-50mbit upload would the in theory if these numbers are not caused by other factorsā€¦

be capped at 5-7mbit uploadā€¦ thus not even reaching 1MByte/s
which could be a big issues for tardigrade usersā€¦ if itā€™s a real effectā€¦ on the lower bandwidths also

Hmā€¦ Nothing interesting, just one mote test (10 aug).
First run - 511MiB file and then some very big file.


1 Like