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
Spin it up via the API. They normally keep extra capacity even when the UI shows no stock.
Just spin up ARM. They are much less oversold, have amazingly fast hardware AES, and not affected by any of the recent Meltdown/Spectre bugs.
@Xel Right now there's C2M, C2S, VC1M and VC1S available in AMS. You must check it frequently, because it's enough that someone will stop VPS, and it will be available again to order for someone else..
Last time I stopped mine instance for few hours, and when I decided to start it again I had to wait about 2 hours to got it working, because there was no free node. FIrst, I thought there's some problem with control panel, because it stuck on "provisioning" but I wrote to support, and they told me that there's out of stock, and node will be assigned when somebody free it
.
My apologies, I have several thousand instances in Scaleway running for a temporary project.
I recommend ARM .
Lol. Damn you @hrz
Thanks for all the valuable info guys. You are all so pro.
I wonder if Scaleway browse this forum cause I couldn't find anything and even support was like it'll be restocked someday to me weeks ago...or a coincidence but I wouldn't be surprised. Gonna pick some stuff up now anyhow. Thanks all.
You just need to check every so often (throughout the day); people do destroy their instances and you can quickly grab one.
Can someone explain what the heck is ARM? Its virtualization method like kvm?
No, it‘s a cpu architecture. The Intel and Amd ones use x86 architecture
Unlike intel's x86 CISC architecture, ARM are risc processors. Giving a higher compute per watt and also provinding thumb mode makes ARM especially interesting. Since armv8 they are able to do virt on arm, so you share a server just like KVM, only different cpu
This is literally the major problem I had with Scaleway, their stock availability is horrendous. Even though I payed for their developer program subscription. It took a freaking ticket to get them to upgrade my limits I PURCHASED FROM THEM. They need to get their shit together, at the end I just decided to just stick to the other 3 cloud providers(DO/Linode/Vultr).
It's the CPU commonly used in smartphones and cheap Android tablets.
ARM can be a headache, you'd better have a "let's learn something new today" attitude if you want to use it in any serious capacity - though support is far better today than it was a few years ago
Out of curiosity, could you show us how fast AES is on these ARM babies? (e.g. openssl speed [-evp] aes-256-cbc)
(btw, Scaleway is and remains a f*ckin' nightmare still in the infant stage of its alpha phase, let's hope they sober up and ditch the whole project)
https://www.lowendtalk.com/discussion/128315/upgrade-your-openssl-on-scaleway-arm-5x-performance-gain/p1
Ah, I see. Hmm, it's only for the newest ARMs (v8?) I suppose... and with OpenSSL 1.1 only... And considering how much Scaleway sucks... I pass.
(btw, reading the aforementioned thread: there's been no such thing as a "price hike" wrt OVH VPS-SSD...
)
Rm how significant is the difference between ARM vs non-ARM with AES/OpenSSL performance at Scaleway. Did you test that?
I did not test OpenSSL speed on their x86, or did not save the results. In my experience their x86 is badly oversold CPU-wise, you can get as much as 50-60% CPU "steal" in top, which means more than half of your vCore allocation is being taken away due to heavy usage by other customers. Have never seen any "steal" whatsoever on the ARM counterpart.
VPS "nodes" for x86 are only 8-core Atom C2750, and it's anyone's guess how many customers they try to fit on one. Certainly way more than 4 (assuming 2-core allocation each). On the other hand the ARM ThunderX servers are 48-core, and they can be dual-CPU (unclear if they use that), which would fit 96 cores in a server.
For the price, it can't be beat. Unmetered 250+mbit, 50gb SSD, 2 GB ram, # of cores, availability, etc.
I did: https://paste.ee/p/HAN5g#s=1
The latter two (labelled AMD64 and ARM64) are Scaleway's €3 offers, while the former two are just a couple of my own computers.
Isn't this indicating the non-ARM offering performs better for single threaded but not multi threaded? When does OpenSSL use single vs multi threaded in the real world? Are two concurrent ssl connections an example of when OpenSSL is multi threaded?
It makes sense for the ARM offering to have better multi-threaded performance but worse single-threaded performance: the ARM is offered with 4 cores, the Intel with only 2 cores, but the Intel cores are generally faster. But I found it interesting that a few values are better with the ARM anyway, even in the single-threaded test.
Certainly anything serving multiple SSL connections concurrently would benefit from threads, unless it's programmed in a bit of a dumb way.
Did you read https://www.lowendtalk.com/discussion/128315/upgrade-your-openssl-on-scaleway-arm-5x-performance-gain/p1 ? Did you compare it with your results on ARM? Did that raise any questions for you? Do you still think your AMD64 vs ARM comparison is not entirely useless?
Are you "still" beating your wife? I've never said anything about what I thought the usefulness of this benchmark was. You were talking about it...
This was certainly not a triggered response.
I guess neither you, nor the admin who "warned" me, knows that the "have you stopped beating your wife?" thing is the well-known prototypical example of a loaded question. That particular question is covered in detail on this article for example, and Wiktionary has an entry about "wife-beating question". A simple web search on these keywords would also easily show it's a very clearly acknowledged as a loaded question example.
Under that light, I would say that if anything is "triggered" (where, I guess, "triggered" is misdefined as some kind of knee-jerk reaction when some research would be warranted instead), it is your response as well as the warning.
If the next "triggered" response if going to be a ban just because I dared offer a rebuttal, then it's okay, I won't mind: I am new here and I won't lose much, and since examples of form over substance abound on the internet, then given the trend, I guess I can only learn to get used to it.
You see, unlike my single response that made a specific point, @rm_ has rather randomly snarked at me for posting a benchmark that had been requested and that he said he didn't have... but they did it "civilly" (and so did I, but that isn't obvious unless one stopped to think and/or google for a second, which rarely happens on the internet), so that wasn't seen as an issue.
Nice alt, @bsdguy. Sorry, I'm not going to get into your headspace; mine is bad enough.
TL;DR.
@WSS Indeed, randomly accusing someone of being someone else just to jeopardize their credibility is an example of the sort of "deeply harmful statements" that aren't literally "uncivil", which is the whole point of the article I linked in the part of my comment you called a "TL;DR" and as such, presumably, Didn't Read.
Sorry, I don't care for fallacious argumentum (sin populum). You just look spergy.
wow