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.
Black Friday 2021 - NVMe and Storage deals - Deploy in 16 global locations (APAC/EU/US)
This discussion has been closed.
Comments
@Mrali
From the message by Hosthatch above, it became clear that - most likely - your VPS is under their observation. This is what I would interpret.
If this assumption is true, your relation with them is coming to an end. I suggest you find a different provider and use your current VPS for testing purposes instead of transferring it to someone with restriction imposed upon your VPS by them.
no, this I would NOT consider normal.
the case above was a different one though and not about cpu bench but IO. speaking of, your IO benches are also on the lower end of what can be expected, I won't question that. still it seems good enough and from what I know @hosthatch balances the nodes by limiting IO eventually. that way you are at least able to get to those 200MB/s throughput and not have it drop down further.
I am a user myself and do the very same benches a lot. that's exactly why I write about expectation management. would you really believe that, just because you book something with a label like NVMe you'll get those ressouces:
do the math, it's simple throughput equals IO multiplied with blocksize. so 40-50k at 64k bs is exactly 2.5-3.2GB/s. so you'Re expectation with 1-3 is already off for the lower number.
and even if you want to see 1GB/s constantly on a cheap VM, what do you think how many user can have that in parallel? do note that doesn't have to do anything with defending providers or not. that's a simple technical limitation.
$30 a year makes it not even $3 a month (and the original post I commented on was even a smaller/cheaper one). so it's obvious that you'll easily need to have 50-100 users in parallel at the very least. so the node would need to be capable of how much GB/s on those NVMes to allow each user each time achieve great number?
again, nothing to do with taking sides. just technical limitations and expectation management.
apart from all that: I did not deny that the numbers are not neccessarily perfect or what one could wish for. yet I stand by it they are not entirely bad either. 200MB/s for $30/y VPS is reasonable (while that geekbench is not!).
TL;DR; if you don't like it, don't buy it. just don't expect a provider to fix things that aren't broken for $2 a month.
thank you
I don't want to transfer it to others. I just can't stand the meaningless ridicule of others.
I have used hosthatch's service for more than 6 years. I never thought they would treat old customers like this.
@Falzo
There are other aspects to the story. One makes a contract with a provider - I do not mean Hosthatch - for one year. That provider does not declare the read/write timings. They simply say that the resources are 10% guaranteed and thereafter burtstable.
What if these bunch of providers oversell the resources and then impose too much of throttling? The affected user is compelled to sit down with a VPS for the whole year.
Thereafter a new argument is launch that it happens to be on shared resources.
So it is not always just the technical side that one has to look upon and calculate the maths, like you did above. There are other reasons too, which cannot be neglected.
In advertisements, why should one declare - again I do not mean Hosthatch here - NVMe, where that provider has an intention to offer timings of normal SSDs, or even less (due to overselling)?
Do you get my point?
NB: Again, here I discuss only something very general on a technical quality against advertisements based on malicious intentions and this has nothing to do with the problem described by the user above or Hosthatch.
You may be right, but I don't agree.
You can look at the other owners, and it is unfair to me.
https://www.lowendtalk.com/post/quote/175492/Comment_3340173
I did not asking for more than I should have. I emphasize that.
sure I see your point. that's why I warned back in the day already that the term NVMe is rather meaningless in the world of shared ressources. it only helps the provider to eventually put even more people in parallel on the same node without degrading performance too much. if you are lucky you might be able to burst better as a client, but that's about it. and of course it's a marketing thing, because people fall for it.
that's actually a very good example. I agree with you that this is not about math at all then, but still, math and knowing technical possibilities help to set your own expectations and not be disappointed later. like rather expect the worst case instead of wishful thinking ;-) providers need to make a living, so yes, they'll most likely overpromise.
If you really didn't have the means, after backuping the data you may have a try to reinstall your server system, which is not centos 7 but ubuntu 18+ or debian 10+.
yeah as said before I don't deny that your bench does not look good compared to others.
about it being unfair, it's rather a lottery and for now you lost. however if you give them a chance to sort it, maybe they will do so? obviously that might take time. they ran a sale and have to sort through offers, weed out abusers and such.
the node you are on seems to be heavily crowded, and again, your geekbench score is not reasonable anymore, so I also would ask about that via support ticket.
chances are they raked in a lot of torrenters and sh*t that are now abusing. I'd try asking for being moved to another node but still expect that to take a few days.
agreed. however defining 'what I should have' is not always easy and could well be translated into 'what I expected' ...
@Falzo
Man, finally you got my point rather perfect. Clever of you!
haha, I'd say I got that from the beginning and didn't disagree after all. you were the one putting me into the providers circle ;-)
as said before I am a user myself and obviously also would love to always near RAM performance as disk IO on every service. yet I am simply trying to get away from that expectation for everything I buy from lowend providers. it's called lowend for a reason ;-)
php friends
8 GB RAM
2 vCPU cores (200% dedicated)
80 GB NVMe
$60 per year
This one has less ram, but cpu cores are fully dedicated. And nested virtualization is enabled as well.
Unless if said user has been severely limited because he has been using what the provider considers too much resources for a while, if this goes according to the plan he subscribed to, of course. In this case, if that's the plan with 100% cpu (1 full core) - seems like it's that one if we take into account other specs - it seems too low. On their 12.5% CPU this kind of limitation/low performance could make sense.
That being said, you get what you pay for, and this kind of deal, while being great bang for the buck is probably not the right place to run CPU intensive tasks for a long time (which could be what brought the limitation, if we consider @hosthatch's latest message).
There are two different issues here. They are not the same issue at all.
One user's CPU is limited to their dedicated limit, which is 50%, as defined in their plan. The node is not overloaded. We place a limit when someone is using constantly more (much more!) than their dedicated limit, for a very long time (at least 48 hours consistently to even show up on our notification system....)
The most likely reason for the low CPU performance (and IOPS too) is because the server is likely in use. Most people run their YABS on an idle server. So this isn't an apples-to-apples comparison. Even if there was no CPU limit, the performance would only be slightly higher......since the CPU is in use by whatever is running on the server too.
I am truly at a loss of how this is being used to represent the performance of the server. Because a 64 core EPYC, which is 99% in use, might show the same result when running the same benchmark.
The second user is complaining about not getting very high IOPS. We've asked 4 times so far in the ticket to explain how this is negatively affecting whatever they are doing on their server, since the 4K IOPS is fine (which matters far more). "My friend has a better benchmark" does not qualify as a valid reason for us to spend time on troubleshooting this unfortunately.
Please use your server instead of running YABS on it 24/7. If there is any real issue, we will be happy to look into it for you.
@Mrali
Let me translate that for you. Type "top" in your server.
@hosthatch why not put the issue to rest by running yabs from a livecd?
yabs addiction is like cancer
OK, I guess I should have said "providers with servers in the USA", since that's what I was looking for. The German providers are in a league of their own 😛 PHP-Friends in particular has pretty bad download speeds to me in the USA - tested with the speed test files on their looking glass.
@hosthatch
I respectfully disagree with a couple of points you mentioned in your above message. Mind you, I an none of the two users you mentioned.
Firstly, when someone says, the limit was reached and therefore throttling was justified, the first thing would be a change in the contractual situation. Either you should have this in the offer, that you are authorized to restrict resources, or in your terms and conditions. It would not be justified, if it is not there. That would be an abuse of provider discretion.
Secondly, things should be made transparent. If, and when, you impose one sided restriction, you need to declare that to the affected users, that such a penalty is imposed.
Most obviously you did not declare the imposition of penalty and let the affected run around like a chicken with a cut head. You should have also offered a possibility to terminate the contract. You cannot just continue the contract upon a throttling resources. If the resources were provided earlier during the contractual term, then the affected user should have an equal chance to terminate the contract.
The reason why I debate this is because I am a very strong opponent of abuse of discretion by certain providers. I was victimized by some unknown party and Vultr terminated the contract with me with a 30 days notice. Later, when I exploded against their childish game, they found out that the abuse report of spamming against me was based on an IP that did not belong to Vultr at all. Even after they clarified their mistake themselves, I refused to continue with them. I learned how they had abused their discretion.
Regardless of IOPS, which you claim to be OK, I find it not OK, when websites were running with a speed - for e.g. - of 8 seconds, when with such a throttling suddenly comes down to 15 seconds. The affected user may have a problem, I do not know.
It is not the IOPS that counts at the end of the day when loading complex php environment but the final throughput. I can readily imagine that user may have a website with a complex php environment that needs a few hundred MB/s. So he is complaining.
I agree with you that the CPU power may not be used with a total burst all the time.
On my server network, there are hundreds of bad and nasty bots blasting resources.
There are thousand of nasty idiots, who constantly trigger downloads by scripts.
While I have hardened many areas, I am not able to avoid such situations.
So it is very wrong to simply blame a user blindly (not knowing what both the mentioned users do) that they uses too much power. It could well be, that there are other factors that may be in place not in control of the affected users.
I can confirm that MJ12bot visits my server network and queries thousand of pages based on cgi scripts that existed between 1999 - 2005. In this 15 damn years, the idiots of this bots have not managed to write a script that updates the error code 404 or 444. When these bots visit, the CPU power goes very high up to 60%. This happens ones in few hours.
Beyond that, offering such a miserable read/write timings of just 50 MB/s by throttling is really, really pathetic. I, and everyone else, would be affected in the same way like the above with such values.
So you are trying to shill each and every provider... based on your contributions to non racknerd threads...
PMS eh?
noice!!!
I never used host hatch before and I just ordered my first deal from this thread.
My usage is very minimal that I am confident that I will never trigger any alarms that they set, but how they handled the situation of not replying for 2 weeks to the guy when they almost instantly replied to all my questions when I asked them on the site, this makes me nervous that they can just ignore someone and don't care, that's just gaslighting someone into questioning themselves.
This made me tempted to cancel and get a refund only to find out in the main page that refunds are not allowed.
Fingers crossed that the VPS will be in a decent shape and good enough for minimal usage for the 2 years of contract.
If anyone used them for a long period and can vouch for them, please do.
I would, and lots here would too.
Their "Top Provider" tag was earned, by community votes and not self-awarded/proclaimed.
@TODO
I have also heard that Hosthatch offers good quality of service. So you do not have to worry, until you enter into such a situation like the other two users did.
Been using them over last 3 Black Fridays... they lay out the land in their post most of the time.. slips up may have happened. But for the most part the provider delivers as promised.
, is it really that hard to read the whole offer before buying something? Have you has any issues with your services? You have said no yourself. You have got great support.
Just dont mind bad apples making a noise for something they messed up with...
Oh and it's fun. Disks fail. Keep a backup for yourself and do not rely on anyone else!
Why would you need a few hundred MB/s for a PHP site?
Disk speed is really irrelevant for a lot of sites - The code will be loaded off disk once, and then it'll be in RAM for all future requests. The code isn't loaded off disk for every single request - That would be prohibitively slow since the parsing + compilation to opcodes takes a non-trivial amount of time. Even static files will likely be in the Linux disk cache (in RAM). In
htop, the orange part of theMembar is cache.Me too. I (kinda for my own project / company) spend either high 3 figures or low 4 figures annually on a bunch of storage servers and NVMe servers for development. Couldn't be happier.
Sorry to be so blunt, but I stopped trying to understand your messages a while ago, when you started sending similarly long tickets, made posts and sent PMs to refund an order on Black Friday which clearly said “all orders are final”…..because you wanted it in a different location.
If you think Emil or me have time to read your similar length private messages over the moralities of not providing refunds (when we said we wouldn’t….) for a service that costs $20/year…..you are deeply mistaken.
We are pretty straightforward with our policies. It’s really not that hard to read what you’re buying. Most users appreciate that, because it leads to a better service for everyone else.
I haven't been with them that long, just since 2018 and yes I've had my fair share of frustrations with certain things, trust me, I've escalated to Abdullah a couple of times but I'm also willing to work with my providers via the standard support process.
Yes, sometimes tickets are slow, I agree, but that's just how it is in this industry.
Fortune/Global 500 companies, paying millions for Managed IT Services, also gets support and things done via tickets
can we still have a bundle? is it too late?
What bundle do you have in mind?
@hosthatch
In my ticket as well as to Emil, I had clearly mentioned about my mistake and confusion. For a $20/year, I can afford a Giveaway to your company, if you cannot issue a refund. These are mere pre-orders on which you want to impose stringent rules.
I had already forgotten about the earlier issue and was freely debating on this matter of such ridiculous read/write timings.
What has my earlier confusion to do with the problem discussed here about IOPS? If you do not like someone, you trigger user specific attack, correct? I was simply discussing something that I believed, and further believe, on principle and this has nothing to do with anything.