New on LowEndTalk? Please Register and read our Community Rules.
AMD Ryzen dedicated within EUR 20-25
Hello, I'm looking for a dedicated server in anywhere in Europe with Ryzen 6 cores (or more :P) with at least 500 GB Usable NVMe (RAID or No RAID) within the price range of 20-25 eur.
Max is 25 eur that's why I'm posting, found Mevspace and Hetzner with my requirement but both are slightly over budget.
Any suggestions are greatly appreciated, thanks a lot!
There are none.
name says it all
I can make it till 28 EUR?
Nope. Minimum you'll find is mevspace and hetzner. Nothing lower.
Consider getting a Ryzen VPS/VDS instead, if you don't need the physical hardware?
buy mevspace for 1 year, they give 2 months free. Makes it around 29 euro. Best you'll get
I already have a Contabo VPS with 6 cores, I'm having issues with a single eCommerce site with 60-70 realtime visitors (it is a heavy site btw).
Planning to upgrade to dedicated resources, what are the differences in VDS and physical hardware?
VDS has dedicated cpu, still shared disk and network. Dedicated server has dedicated disk performance, also dedicated network port (does not mean dedicated network speed)
I see, do you think upgrading to this will help in my case?
Kindly recommend some providers with good reputation that match my requirements, thanks.
Also, VDS: 6 cores, usually means virtual cores not physical cores. So 3 physical cores. With the least ryzen on the hetzner AX series, about double performance of what you have already. (depending on what CPU your contabo node has)
Also what you have with contabo is not VDS, it is vps. The cores are shared.
https://www.netcup.de/vserver/ - the best you can get; 10 dedicated cores for 29 a month. Not Ryzen though, EPYC. This is more powerful than a 3600 dedicated server.
Netcup is one of the best EU VDS/VPS providers I would suggest. However, you should get used to German contract policies. The service won't automatically cancel. You have to place the cancel request before a certain deadline.
Any reputable LE VPS is better than a Contabo VPS.
consider your next VPS to be quad core ryzen/epyc with DDR4 ECC ram and NVMe RAID10. that will fix everything.
for 500GB, good luck finding one.
I want a Dual EPYC server with 512GB of RAM and 4*4TB of PCIe SSD. My buget is $10.
Seriously, unless you go with something like co-location, there's no way you can get Ryzen Dedis at that price.
IMO you'd be better off getting a high-end VPS or VDS rather than a low-end dedicated server. Specs are going to be better in that case. Many providers on this forum would give you a pretty beefy VPS/VDS for 25 Euro per month.
This is mentioned in the terms that you're supposed to read when signing up so it shouldn't come as a surprise to anyone...
Not 10 dedicated cores, 10 dedicated virtual cores... Meaning 5 actual dedicated cores. And they might not truly be dedicated (i.e. the CPU affinity may not be configured to always run your VM on the same 5 physical cores, meaning that you'll still have more context switching overhead than you would with actual dedicated cores). Still a great deal.
(also, no, a vCore isn't exactly half of a real core. HyperThreading only increases CPU throughput by around 25-30% max, not 100%)
But actually, there are a lot of people who don't read TOS before purchasing. That's why we saw same complaints about Netcup nearly every year...
Sorry if my post offended any of you, I noticed deals as low as EUR 32 for similar configuration few days ago, so thought I'd ask.
Thanks for all the suggestions and comments.
60-70 realtime visitors is nothing. You can easily push hundreds on single core with REALLY heavy website. In case of eCommerce if you optimize it slightly (nginx long&short cache strategy is often enough) then even a thousand is okay on single core.
How did you came up to conlusion that CPU is a problem in your case? Why are you requesting NVMe for just couple of reads/writes per second?
Seems like you just want specs that look good but you didnt test why you have problems. It can be packet loss, it can be network speed, it can be routing, it can be IO limit, it can be stupid error in application, it can be problem with server config... etc...
You just wasting money right now. You dont need dedi for 60-70 visitors. It should run fine on 1vcore+1gb ram. Install monitoring software, check was is bottlenecking. If you wont find anything then debug app (maybe some stupid long query every visit?) And in the end cache your app. Even 1s-5s cache in nginx can do wonders if done correctly.
was it a Magento e-commerce?
Most likely, it is a woocommerce (wordpress) store with tens if not hundreds of plugins. Maybe his setup is only using 1 core and not utilizing all cores. Maybe there's no static caching or object cache. Maybe this, maybe that.
The simplest solution that won't require any configurations or research that is budget-friendly is to get a vps with fast single core performance. But if 500 GB of nvme is a requirement that you absolutely need, then you probably are better off getting a dedi. You rarely get that much disk for this budget (regardless of ssd or nvme). You can get that much from netcup but those are epycs and what you want are probably ryzens.
Other solutions are to forgo woocommerce or at least fix your setup.
No matter what he has he can just make cache with Nginx - cache all pages until there's a cookie (something in cart or user logged in). Pages will be sent instantly to all new visitors and only people that are doing something are making any load to server. Its like 90% decrease of server usage in most cases.
This is very "dumb" setup (no cache invalidation etc. so cache cant be long), but even that is enough for 60-70 visitors. If its WooCommerce then ofc he should look into dedicated cache solutions or LiteSpeed, maybe that will work for his website.
But... Caching shouldn't be needed at all with that amount of people and 6 cores. And also NVMe is not needed at all, even HDD is plenty fast for such traffic lol
He should monitor what exactly is wrong now. Maybe he is in very bad node and has high CPU steal - that creates latency, because you need to wait in queue for calculations. In this case it will show him that CPU usage is 100%, but that doesn't mean that he needs more cores or dedicated server.
Unless the “1GB RAM” is from providers like CaC which took like 2 and a half hours to install LEMP
My site's performance is good when nobody is on the site (late night).
When I check it during peak hours, my load avg crosses 18 on a 6 core server.
I'm using LiteSpeed Cache on Cyberpanel + OpenLiteSpeed along with Redis object caching.
60-70 visitors is nothing unless your server hasn’t been optimised so even OP server specs will be not enough if person unable to manage the VPS or server
is your website built on MySQL/MariaDB? if so, have you checked MySQL logs, which keeps records of slow queries.
system load is an important indicator, but not inferring too much information. high load may results from running out of memory, too many i/o ops, extremely large amount of active processors/threads, or a too crowded server where your vps is located.
apart from MySQL logs, you should take a look at top command as well.
I don't think people telling you that "60-70 visitors is nothing" know anything about eCommerce performance. Magento sites as well as WooCommerce sites often have a lot of plugins and do heavy database queries.
Caching with Nginx as someone suggested... it's great, but it doesn't work like that.
You cannot cache stuff like ajax requests, checkout pages, admin area requests, most logged in users, etc... so that cache only affects portions of the app, not others.
For 70 users on WooCommerce, I would not use less than 8 cpu cores, regardless of any cache. As long as there are ajax requests you are also hitting PHP and using CPU resources.
That being said, while you need a fast CPU (because PHP is single threaded) it doesn't need to be Ryzen, just fast enough.
For MySQL there is a noticeable difference for MySQL between SSD and NVMe even with a large innoDB cache, so you are right on NVMe regardless of what anyone else says.
If you are on a tight budget, you can try 2 servers on the same private network.
You could buy a smaller NVMe server for MySQL and another for the web server, which would not require NVMe, or could use even some sort of object storage for static assets.
500 Gb NVMe at that price is probably not going to happen, or if it does, it will come with a poor CPU with Sata 2... or a slow network like 100 Mbps.
I configured servers for big eCommerce sites in Poland. Milion of visitors.
60-70 visitors is nothing. Nothing. Null. Zero.
Last year I configured server which was serving 1000requests per second(!!!) on Black Friday. It was OVH 4core server. 1000 REQUESTS!
CPU usage? Around 40-60% maximum.
So you mean like 90%+ of pageviews? Admin area requests as something that will slowdown 6 core are you for real right now? XDDDDDD
Its like 2 ajax requests per second at most. People don't hit buttons every second. They look at static page and MAYBE they do action.
There's zero difference with several read/writes per second. Normal SSD can do thousands EASILY.
If "I know nothing about eCommerce" then lets bet $100 maybe?
Give me your heavy eCommerce site (replace database items with random numbers for security) and I'll show you how to serve 100 clients per second on 1 core.
I have done similar work for the past 10 years, but your experience and my experience are obviously on a a very different level if you have never seen a 6 core cpu struggle with just backend requests from a 10 person team, on a ecommerce store.
If you were hired to configure it, then in 4 words, static content is cheap.
If you can be lucky enough to have a site built that can be fully cached, or if you were paid to go in and do advanced work like caching queries at the template level... then of course, you can do 1000 req/sec or even 10,000 req/sec on cheap low end servers if you like (hope you didn't hit the network limit before that).
But that is all irrelevant for this.
What most ecommerce site owners do, is to install some cache plugin and add Litespeed Cache, or Nginx for GET requests and call it a day.
Also, the OP is probably not an engineer and neither he wants to pay for one to come and fix the performance. And the fact remains, that he is struggling with performance for his site, so he needs more power.
It doesn't matter if you can do 1000 req/sec on a $5 server, you can do 10 Million too on a $10 server (google it), because your site and his site are built very differently.
I have seen ecommerce sites with 10+ ajax requests and I've seen those ajax request looping on thousands of mysql rows doing table scans, with RAND() queries or ORDER BY queries, that believe it or not, take at least a few seconds to run and block 1 cpu thread while they are still running.
It all depends on how the site was built, so you cannot generalize your experience, for a site with millions of views that probably had the budget to micro optimize at the query level and cache level, if not even adding stuff like Elastic Cache, Solr or similar.
That is completely different from the typical site owner, that has a very limited budget and thinks he can run the next Amazon on WordPress, WooCommerce with WPML, multi currencies, BuddyPress, and 100k products on the database, all filtered nicely on the sidebar and grouped by counts, price, categories, etc, using nothing else other than PHP and MySQL. And of course, running on the cheapest server possible.
Of course SSD can do thousands of read/writes per second, but NVMe can do more and if you benchmark, you can see the difference. The speed at which queries execute, is not just about the number or reads/writes.
Take a large enough query that needs to do a table scan on a large table for example. If you do a RAND() query, or a ORDER BY title ASC on a joint table, between the posts and postsmeta tables on WordPress for example, a NVMe disk is going to be much faster than the SSD simply because the temporary table has to be written to disk.
This is very obvious on large ecommerce stores with complex queries... so not trying to offend you, but I don't know what sort of sites you been working with, or perhaps you never cared to benchmark things.
That being said, sorry if you took it personal, but that wasn't my intention.
I understand you came basically to say you stand correct... but sorry, the more you say, the more lack of experience you show.
Optimizing a few dozen stores that happen to be well built doesn't give you authority to generalize that all sites are the same, and that everything is very easy to achieve with caching.
That is not the furthest from the truth. Cache is useful for scalability, but there will still be random slow response times when it expires, if you do not solve any slow queries and underlying problems.
If the Poland site you configured was already fast and had very lightweight ajax requests, or none at all, of course it's trivial to add nginx cache and then it's basically a static site, so it's the same as serving static images and this has nothing to do with the OP needs.
In the end, if he cannot find the server he needs and cannot pay someone to fix the performance at the code level, he will have to either put more money for bigger servers, or do as I said and split the database and web into 2 cheaper servers.
NVMe for the database is ideal, but for the scripts, SSD is sufficient.
Now please don't come and tell me that you can also use an HDD...
HDD can also do thousands of queries, and that's what we were using 10 years ago in RAID 10 when SSD were too expensive, but tech evolves for a reason, and SSD > NVMe also exists because there is a difference.