New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
That's somewhat logical, as people who require IPv6 usually skip those rare hosts that still don't support it.
I mean since you don't support it, it's somewhat understandable that the majority of your client base isn't very interested in it. Those who are interested aren't your clients.
Not complaining (there's a lot of IPv6 ready hosts to chose from), just stating the obvious.
It was more likely xs26, Freenet6 or SixXS. HE wasn’t really that popular among European IRC users, and then they also blocked standard IRC ports (except for users who completed the free IPv6 certification and reached the Sage level).
xs26 was the most popular among early .fi IRCnet/Undernet users. They stopped offering IPv6 tunnels quite some time ago, concluding that their mission was complete since most ISPs now provide IPv6 natively.
So let's digest this bullshit. Funny sure, but just bullcrap.
You can check photos we just posted, first N100 6x3.5" Prototype testing @ lab : https://imgur.com/a/1Xz3KfX
Kerava Datacenter construction work: https://imgur.com/a/gymCbR9
Some sub-component assemblies from the lab: https://imgur.com/a/2PhwRhB
Thanks for catching that!
Not quite that old site, but thanks for noticing we've been in the biz long enough to be called dinosaurs.
Where do we give you service credit for catching this error or shall we just throw some mad 1-2 seedboxes special on this thread?
Fixed the error.
Noted. infact i think we have plenty free on the M10G series. We also just hit 15 years in operation ... we should celebrate it, even if it's a bit late!
IS THIS BETTER;
In photo is SZBOX N100 NAS Motherboard. We bought a couple different variations. But probably CWWK and Topton has similar.
We have one model with direct DC Input on shelf too, excited to see that in action. Also a bunch of various types of N100 minis to test out.
The beauty of our platform and design is that we can just "slap in" whatever, and it scales. There are some quite interesting ideas!
We can easily integrate any 1L MiniPC (Vesa 100x100 mount), mITX (mITX mounting pattern), or make adapters for say 4x4 NUC. That's our plan, eventually we can integrate almost anything and offer wide range of units.
It scales up to 1000W total per platform (in theory), and virtually we also have a 16 node model too (4x4 NUCs) designed. So we will be testing a lot of different hardware configs over time, and offering them for sale.
The best units we will simply make more of, bad units will be taken out of production eventually. One way to bring the best value to our customers! Data will reveal what our customers need, appreciate and is reliable.
Data filters out the bullshit and we'll give customers what survives in production and brings the best value.
One exciting model will be the N100 with 4x NVMe (tho just x1 lane per NVMe so capacity not performance thing). We already have 1 unit in stock for testing.
The AsRock N100 ITX was my personal preferred choice (well known and reliable brand), but the value proposition with them wasn't quite what these offer.
Reliability
Our reliability metrics are a bit different than homegamers, since we deal in bulk. So with some larger dataset we will see how reliable these are in long term, larger number usage.
This is one of the most important metrics, since no one in our staff wants to be running around fixing nodes
For example for HDDs we see the failure rate over like a decade or something like that, it's less strict data, but also maintenance accrued etc.
ASRock based Zen MiniDedis (the old, Zen1 era nodes!) are otherwise reliable, but have random halts/freezes (rare but noticeable!) -- but on the other hand, most of them are still in production and most issues were actually the awful Samsung 4TB SSDs and Mushkin NVMes, not the nodes itself.
Least reliable nodes we've ever had were actually from Dell, or specifically from Dell Datacenter Services. They cheaped out + a firmware bug.
Tyan had specced copper heatsink for chipset, but they used slightly smaller aluminium and less fins (more coarse fins), along with really bad thermal paste.
Second, the dell bios had a firmware bug under reporting chipset heat by 20 or 25C !!
Even worse -> They had used wrong type of plastic for the heatsink clips, so the clips would fail too!
This resulted in random crashes, node would typically first start crawling then crash. Reboot solved it.
We had hundreds of those units at the time. Took us damn long time to isolate the real culprit. We re-pasted every server and installed 40mm fans. They've been chugging along another 6 years or so after that
We don't really care about paper specs or anything like that, data will reveal what is good and what is not.
Yea some skip indeed, but also data from elsewhere about the actual usability of IPv6 in production. Multiple sources, anecdotes and data released.
That being said, when the math works out it will be supported, but right now almost anything else brings more value to our customers, ultimately what matters.
I remember SixXS as well, you are probably right on the tunnel selection, hard to remember something such a long time ago. Just goes to show, not new to IPv6, but business metrics come first, actual real world ROI. Otherwise we would do triple redundant this and that, and Tier this and that, and increase pricing of our services by 10x or 100x too shrugs Our mission is to bring maximal value to our customers, and as long as that case is not supported ... Any feature has a hard time to pass to production.
Even the strongest IPv6 advocates still fail to bring any kind of business case for it. I don't think i've seen a strong business case to date. There's a lot of crazy spending in IT for stuff that don't bring end user value.
Example; typical minimum for some is dual AC feed power + UPS ... in area with historical 99.99+% grid availability data ... or using UPS units with historical data of commonly failing on power outage or even a brownout, and always wasting 20-30% of power on that same power grid facepalm
I know of a premium datacenter where something like 95% of downtime is caused by their fairly expensive ÚPS units. Ouch!
Anyways, just noise alone is not enough for us to bring IPv6, but once we see concrete data it brings more value to our customers as whole, not just the "0.1%" interested.
it wont even cover filling the 72TB drives, let alone downloading anything from the box
good for you?
Jesus christ, can we please have a little bit more hostility I got plenty of popcorn.
What is the refund policy after 30 day moneyback period has ended? I saw on both Discord and Reddit where customers were being charged with their account already closed. No refunds were made and no accounts to credit back, what happens to the customers' money?
sure. but to be honest I've never used anywhere near 330TB in a month
same, i often prefer hosts with >1Gbps ports offering <1Gbps monthly traffic than a host offering 1Gbps unmetered. its more useful for my bursty on-demand usage pattern
Our refund policy: https://wiki.pulsedmedia.com/index.php/Pulsed_Media_Refunds_Policy
We never charge anyone, ever. Customers always control their own finances. We do not have access to their payment account(s), nor should we ever have such access.
Service renewals are not refundable. For seedbox/storage services we offer 14day initial moneyback guarantee so you can safely test the service.
If subscription accidentally continues after cancellation, the payment is added as service credit automatically. This is completely automatic, and credits are automatically deducted on new invoice generation.
Sometimes we can cancel a paypal subscription if they have made one. The new paypal API afaik allows autocancelling the subscription from our end, but the old paypal API (which is the original, called "Paypal Basic") does not allow this, and we send multiple warnings to customers (knowledge base, and cancellation confirmation email) to make sure they do not have automatic payments remaining.
With Paypal Basic atleast i have noticed that we do not even have Subscription ID on file in the billing details, therefore we have no access to even cancel it. This is suprising common, but somehow WHMCS still finds the correct account to credit. We use WHMCS as billing solution, it is the most common webhosting billing solution as far as i know.
You can always open a billing ticket to check if we can cancel the subscription for you. Sometimes we do this proactively if we can and happen to notice the user has lots of credit or incorrect subscription setup (ie. package changed or something).
You can see amount of credit under notifications in the billing portal, or again ask via a ticket.
Downstream fully unmetered & dedicated for all practical purposes. Our traffic patters are roughly 8:1 Upstream to Downstream, so it's the one resource we have more than we know what to do with.
As stated before, it is a softcap, not a hard cap. Soft cap means a human will check and evaluate if any action should be taken, and if so, what.
It is very rare tho, we have quite a bit of over allocation typically. For these units it's a 10Gig link for 2x 2.5G nodes, meaning the first link will never saturate, so only full rack saturation / congestion may be the limiting factor.
With our current MD units we have never had to intervene, not even once. ZEN MiniDedis we had some abusers within the past year tho, but i do not remember any congestion event with MD units. and even that was a very clear abuse case: all their servers mostly maxed out 24/7, didn't look like organic traffic. I remember more than a decade ago (12-13years ago), a well known seedbox provider used our servers and did a wget loop for external test files, just because "i can", and technically we had sold them 100Mbps unmetered on each unit, so technically it was not abuse in his eyes.
That's the main metric; Congestion. If there is congestion, then a human is alerted to check which port(s), which unit(s) and are they at sensible, or is there enough uplink capacity.
The 50TB mentioned is already plenty and safe number, to best of my knowledge that is way more than others in this budget class.
Exactly. Softcap allows bursty usage freely, even over the scale of months. There might be genuine reason for a user needing 2Gbps for multiple days, but not much otherwise.
We basicly intervene only when people abuse.
Some quick lab/testing photos: https://imgur.com/a/dRm5hBd
Just received a 71kg pallet of ABS filament. Currently iterating various designs for 3.5" HDD mounts used in these N100 units and all future similar storage builds.
Goal is to find the highest longevity, ease of maintenance option.
Got solid engineering ideas or insights? Drop A Comment below! 💪
Would be interested in what you come up with in the end, basically going through a similar thing with mounting both 2.5" and 3.5" drives next to a N100 MB via 3d printing designs.
Ah this is easy, we already have 6x2.5" hotswap bays -- these are common in China apparently, very cheap. These take "1 motherboard slot." each.
BUT! That being said, if there is a 3.5" drive, there is enough space on top of it for 2.5" too ... That means a sled like this could have 12x3.5" + 12x2.5".
That would be curious engineering challenge to solve. Will give this some thought later on.
These are already quite densities.
what drives are you planning on using
7200rpm SATA.
We have various models and makes.
all CMR tho right?
We do not use SMR drives, so yes they would be CMR. Performance matters on seedboxes.
Seagate made it hard to distinguish SMR from CMR (or Perpendicular Magnetic Recording) at one point, had quite an argument with vendor for them to accept the return due to false marketing (ST8000DM00 vs ST8000DM000 if i recall right)
Here's some more photos out of the lab today: https://imgur.com/a/rdqamkd
The Engineering challenges!
These are from old Zen MiniDedi, to be upgraded for the mPlate platform;
Heatsinks we used are way too high, about 12mm too high or so!
Got to find proper heatsink solution for those.
Quite an size difference here, ZEN MiniDedis were so damn bulky!
Already got 2 potential solutions. I guess we'll get some AM4 to G34 heatsink adapters fabricated at a local metal shop. If anyone needs these in bulk, let us know (shipping anything under 50 sets is not worth our time). Guesstimating it's going to be like 5-7.5€ VAT 0% per set.
We got hundreds of those heatsinks in picture sitting in warehouse from very old opteron servers. If i had to guess we got something like 250 total of those copper heatsinks.
Dang it! 5mm too high!
No worries, this engineering challenge has already been solved, just need to test it with fresh batch of NVMe drives
I'm not a customer and will not as I need ipv6. (for dedis)
But have to say that I have always loved the way you experiment
Reminds me of OVH in their early days
Thanks!
Now if i remembered to do the whole post before posting (just edited more info on above) ..
Experimenting with HW like this is actually quite fun. Wish i had more time to do it, instead of getting constantly bogged down by bureaucracy. Thankfully we just hired more staff
I actually consider this stuff relaxing sunday activity too, so spent the afternoon fiddling with hardware configs, and getting prototype #1 drive sleds glued for the presale N100s
-Aleksi
(yea the founder of Pulsed Media / MCX)
Oops! A slight mistake in the positioning of the motherboards.
Thankfully, these are using printed adapters to Vesa 100x100 mount temporarily. Accidentally had wrong hole sizing on the DXF sent to metalshop (or accompanying schematic wrong, human hands make human errors eh), so couldn't tap M3 or #6 threads into the mITX pattern, hence the grounding lead too.
These plates are actually surprisingly expensive as we are not manufacturing these on a robot yet (need like 100+ per batch).
Glued up the first 4x2 set of HDD bays for the first prototype.
These jigs are no good, but got the first set done.
Got to make proper metal frame jigs when needing to make more than a few (10 total units == just 5 of these for now ...)
Then put on the giga for additional heat to get to test out sooner;
and YES using old 3.5" HDDs as weights (added more after the photo). Using terabytes of weight!
(No worries, no production HDD was harmed here!
These are just old 3TB models each with a few bad sectors in the counters, still fully functional, but worth nada for anyone)
This is what the jigs etc. look like currently, and printer tolerances come here into play. The white ones are tighter tolerance, higher precision ... but they are loose! The black ones are tighter as the machine which produced them is not as precise and sloppier. Bed Slinger @ Max Accel/Speed VS CoreXY at leisure speed.
These was perfect example of tight vs. loose tolerances. The funny thing is, loose tolerance (black) parts were better as jig as the more precise parts which were within in dimensions.
Regardless, since this is too tolerance tight etc. this design doesn't fit our needs. Got much better metal frame jig in mind when we start producing more of these.
forgot to ask, what is such a IPv6 use case for you that tunnel will not work?
Tunnel works for my use case
I tend to think it would be less reliable and stable than native support. As many hosts support it natively, I don't want to bother with that
But I understand the cost of adding it to your existing infra.
I was just pointing out that even if we're silent, there are people for whom it's important.
Will you be offering any with less HDD, maybe for lower price (say 4)? And even less memory (say 8GB)?
Kudos for offering something unique in this area of hosting. If I was still buying servers for seeding I would absolutely hop on this but alas I have brought everything home.
We have a VPN provider as a customer and they seemed to be happy with tunnel reliability. Not sure who they used etc. tho.
It would be interesting to see some reliability statistics for tunnels.
Sorry no, not anytime soon. 6x3.5" is the platform we got.
There might be a say 6x3TB + 8G Offer, but makes no sense until we exhaust our supply of larger drives.