I'm loving this, I've been using scaleway as a host for a few months and they've provided great value during that time. As a Dutch national it would be great for my latency to see scaleway add some capacity to Amsterdam.
Their low end offering allows me to get up and running quickly with a new machine and experiment, without touching any existing instances or firing up a VM myself.
It's a shame transferring machines to the new region apparently isn't supported, but since I'm running on docker it shouldn't be too hard to recreqte my current instances.
Wait until you get RAM errors, you probably won't notice it until services quietly start to crash and you cannot figure out why because the errors are not logical. So you try to restore a backup - but then the backup won't unpack because the RAM created a corrupt backup. Having a fancy file system like ZFS won't help, but it might give early warnings (if you scrub often).
Combined with a working exploit and when not running on the bare metal server that might actually be a valid point. I know when they were released those attacks looked pretty scary, and ECC was supposed to maybe help. I'm not actually sure it does help completely, and that here aren't other measures already in place by now?
For professional news having data corruption is not something that you'd like, especially if you handle client data (as in backup) or even more importantly if you handle payments.
Overall for pro use lack of ECC is just a catastrophe waiting to happen.
Na. There is a bunch of stuff that can go wrong, but there are also a bunch of safeguards that do not rely on ECC. Backups have checksums. TCP has checksums. I mean, if it is the same price I'd have no problem picking ECC, but I see the danger of it having any impact on what I and 99% people on servers are doing at 0.00000001%.
I'm using Scaleway for some experimental stuff, and it's great value for that.
I did not like that I am bound to their kernel on bare metal. There are setups that require kernel modules they don't provide, like DRBD. I got the modules to compile and work in the end, but it was not fun and will probably break on some future update. This is not how I expected it to be on "bare metal".
Also that prevents you from using the Xen hypervisor, which again I would not expect from "bare metal".
You only get a single IPv4-Adress per system, which not only is a burden to virtualization, but also larger container-setups.
So I found little value in the "bare metal"-part, but it's still great value if you compare it to other VM offerings.
If you treat it like a VM (or use one of their VMs), you need to be aware that disks are not redundant and their content (or at least the changes since the last boot) are gone on a crash or hardware failure. You need to mirror or replicate it yourself, or design around it on a higher level of the stack. There is no option to attach a persistent disk that would survive a hardware failure and could be attached to a replacement host/VM.
So in the end scaleway is a weird mix for me: for rented VMs I expect the platform/provider to take care of some basic reliability and redundancy stuff. For bare metal I expect to be able to setup all of that myself without trouble.
Well, for one I don't care much, I run my hobby stuff on there, not my mission critical services, so cheap is good, even with some really minor drawbacks.
Secondly, I could not find a source if Scaleway uses ECC or not...
I was really excited about Scaleway until shit started going wrong with my test C1 box there. They sent an email asking me to update my "boot script" (kernel version). You can only do this, it seems, when the box is shut down, except there is no power management on the box. The only option they have is to copy the contents of the box to cold storage. Great. It took 4 hours or so.
When I brought the box back online it would stay online for about a day and then lock up for no reason. I spun up a new one, copied the data from the old volume to the new one. Why can't I clone or detach a volume from a running box? It was a read-only data-only volume. The copying between two boxes was at 12Mbps... In the end, the new box had the same lockup problem. Not it's not my software. It had been running without changes for 6 months prior with no downtime. Oh and in the end it locked up again and when it came back online in rescue mode, it was basically a brand new install. My home directory was gone, root password had been reset, all the configuration was not there.
Also no IPv6 for bare metal ARM boxes.
tl;dr: I moved on because of inflexible control panel and random hardware issues. Really wanted to like it.
I was checking out their terms of service (https://www.scaleway.com/terms/), a few clauses in there seem pretty restrictive, such as:
>Users are required to use decent and respectful language. All abusive, violent or hateful sentiments are totally prohibited.
>The propagation of data, images or sounds that may constitute defamation, an insult, denigration or an infringement of privacy, image rights, good morals or public order.
Does anyone who knows this better than I anticipate that this would cause problems if one wanted to host, for example, an internet message board with this service?
Obviously one would comply with DMCAs, delete illegal content, etc, but I can't imagine forcing all users to use "decent and respectful language", for example. Is this something you just generally ignore until a specific problem with the provider surfaces?
>Does anyone who knows this better than I anticipate that this would cause problems if one wanted to host, for example, an internet message board with this service?
Probably anything illegal in France or Netherlands. The sentence sum the most common thing you could do that could get you in legal troubles if you cross some lines.
I actually love Online.net and Scaleway. The problem is just that I personally prefer to have one strong core and not 8 slow cores since single thread performance is still the most relevant thing. Even though you get 2GB RAM for 3 EUR they are still shitty CPU cores. I mean you can't expect a lot for that price but I'd prefer a shared burstable E3 or E5 core over Avaton cores. Sicne Scaleway only offers these slow cores it's not useable for me even though I love them.
We design and manufacture our own server hardware so we can optimize our cost. We select each component of our servers to get high quality servers at the best price and remove all unneeded components. The result is that you get a server with the best price-performance ratio.
Whereas the competition enjoys paying extra, and carefully tries to find new unneeded components? Yeah, didn't think so...
I guess the truth is a mixture of "larger willingness to cut corners" and "aiming for a lower margin". That's not a bad thing. It's just that in a market as transparent and competitive as this, it's really hard to be better along every single dimension.
Since we design our own hardware, we put exactly what we want and need into them.
We share some components like power, network, and casing.
This means for example :
- one integrated network switch/router for 18 servers; no need to buy an external cisco/juniper/whatever
- one dual power supply for the whole gang instead of 19 (18 servers + switch)
- less metal/plastic parts
We try to reuse hardware designs and software from one iteration to another when possible.
In short, we do have less "unneeded components". Equivalent designs (thinking at moonshot for example) are expensive, so to achieve same density it's more expensive to buy existing solutions than what we design & build.
EUR/USD went way down those last years so Europe is becoming a low cost area.
As well, I don't know about DigitalOcean but with Online there is some basic services they don't provided.
For example it drove me mad that you can't just move a server to an other account or change the account owner with online. I've never been stuck with the dozens of other hosting company I've been working with to move a server.
I would imagine that intrinsically the costs of buying/renting an X number of CPU cores and Y gigs of RAM just has a certain cost and therefore the price of VPS's just have to be at a certain minimum level. I guess Scaleway is just closer to that edge, but being about twice as cheap as DO (which is quite big and popular) just makes me question the long-term viability of these prices.
I've been a Scaleway user for a very long time (I cancelled recently in an effort to cut frivolous personal expenses). Their cheapest offerings are very slow, but they are perfect if you're looking for something extremely cheap, dedicated (with your own IP), and with unlimited traffic (!!!). Also, as people noticed here, their web console is pretty great. They even have preloaded images, I had a personal torrent server up in 2 minutes.
Is there a reason for not supporting C1 instances in Amsterdam? I was an early adopter before Scaleway started offering virtual instances, and liked the idea of running bare-metal for such a low price, but it seems it is not available in Amsterdam.
The ARM cores are slower - there was a thread a while back: https://hackernews.hn/item?id=9309459 the X86 servers are not very fast either: https://hackernews.hn/item?id=11251671 but if you look at the numbers in that thread it's likely that DO overbooks their machines and if you are unlucky you can end up with slower performance than on Scaleway.
But for me it doesn't matter. I'm not doing heavy computational lifting and for nginx + some database it's fine and 2GB memory is pretty nice to have.
A minor annoyance is that you can't choose the kernel version and you are stuck on 4.5.7 at the moment - I guess if you care about local security priv escalation bugs it's a no go.
But docker seems to work and you can select a kernel flavour in the admin menu.
Another annoyance is only a /128 IPv6 address for the VPSs - but IIRC the dedicated boxes get a /48
But for 3€/month I'm more than happy - you usually only get OpenVZ VMs with ancient 2.6.36 kernel for that price if at all.
Here [0] are some nice comparisons from a year ago. If these are still more or less correct, then I'm happy to be paying the extra few USD/month on Digital Ocean. If they've made performance improvements since then, I would consider switching. Would be nice if someone re-ran these benchmarks.
Their low end offering allows me to get up and running quickly with a new machine and experiment, without touching any existing instances or firing up a VM myself.
It's a shame transferring machines to the new region apparently isn't supported, but since I'm running on docker it shouldn't be too hard to recreqte my current instances.