All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
[Review] Netcup Root-Server M SSD v6a1 12M
This is a now discontinued product from Netcup (https://www.netcup.eu/bestellen/produkt.php?produkt=1121). I am terminating the contract with them but I have to say that it was a quite good experience overall.
Posting here freepvs' bench.sh results (two runs separated by few mins) and then nench.sh to get some CPU benchmarks.
Server was idle during all these tests (I checked with htop/iotop for few minutes) as I stopped almost everything (except SSH of course).
Other relevant information: server has always been used for compilation, so quite CPU intensive (100%). Disk is full at 88% and it has fluctuated also above 90% many times in the past during regular use.
Disk is not reported as SSD by Linux:
$ lsblk -d -o name,rota /dev/vda NAME ROTA vda 1
What do you think about this root server?
The only downside is that I could not get the dedicated cores, so the 2 virtual cores always felt a bit slow, but you get what you pay. Also, is this a real/good SSD in your opinion?
Edit: added ServerScope benchmark: https://serverscope.io/trials/GVNr
Benchmark started on Fri Apr 27 12:56:02 CEST 2018 Full benchmark log: /root/bench.log System Info ----------- Processor : Westmere E56xx/L56xx/X56xx (Nehalem-C) CPU Cores : 2 Frequency : 2493.748 MHz Memory : 5961 MB Swap : MB Uptime : 11 days, 22:16, OS : Ubuntu 16.04.4 LTS Arch : x86_64 (64 Bit) Kernel : 4.13.0-38-generic Hostname : ubuntu Speedtest (IPv4 only) --------------------- Your public IPv4 is - Location Provider Speed CDN Cachefly 72.2MB/s Atlanta, GA, US Coloat 13.9MB/s Dallas, TX, US Softlayer 11.1MB/s Seattle, WA, US Softlayer 10.6MB/s San Jose, CA, US Softlayer 9.41MB/s Washington, DC, US Softlayer 18.6MB/s Tokyo, Japan Linode 8.18MB/s Singapore Softlayer 8.30MB/s Rotterdam, Netherlands id3.net 36.5MB/s Haarlem, Netherlands Leaseweb 63.5MB/s Disk Speed ---------- I/O (1st run) : 219 MB/s I/O (2nd run) : 265 MB/s I/O (3rd run) : 333 MB/s Average I/O : 272.333 MB/s
Benchmark started on Fri Apr 27 13:03:59 CEST 2018 Full benchmark log: /root/bench.log System Info ----------- Processor : Westmere E56xx/L56xx/X56xx (Nehalem-C) CPU Cores : 2 Frequency : 2493.748 MHz Memory : 5961 MB Swap : MB Uptime : 11 days, 22:24, OS : Ubuntu 16.04.4 LTS Arch : x86_64 (64 Bit) Kernel : 4.13.0-38-generic Hostname : ubuntu Speedtest (IPv4 only) --------------------- Your public IPv4 is - Location Provider Speed CDN Cachefly 73.8MB/s Atlanta, GA, US Coloat 12.1MB/s Dallas, TX, US Softlayer 12.3MB/s Seattle, WA, US Softlayer 9.95MB/s San Jose, CA, US Softlayer 9.82MB/s Washington, DC, US Softlayer 16.2MB/s Tokyo, Japan Linode 7.72MB/s Singapore Softlayer 8.02MB/s Rotterdam, Netherlands id3.net 32.0MB/s Haarlem, Netherlands Leaseweb 61.0MB/s Disk Speed ---------- I/O (1st run) : 341 MB/s I/O (2nd run) : 246 MB/s I/O (3rd run) : 279 MB/s Average I/O : 288.667 MB/s
------------------------------------------------- nench.sh v2018.04.14 -- https://git.io/nench.sh benchmark timestamp: 2018-04-27 11:07:42 UTC ------------------------------------------------- Processor: Westmere E56xx/L56xx/X56xx (Nehalem-C) CPU cores: 2 Frequency: 2493.748 MHz RAM: 5.8G Swap: 6.0G Kernel: Linux 4.13.0-38-generic x86_64 Disks: loop0 512M HDD vda 117G HDD CPU: SHA256-hashing 500 MB 4.185 seconds CPU: bzip2-compressing 500 MB 7.847 seconds CPU: AES-encrypting 500 MB 2.370 seconds ioping: seek rate min/avg/max/mdev = 80.8 us / 135.1 us / 31.9 ms / 311.1 us ioping: sequential read speed generated 13.8 k requests in 5.00 s, 3.38 GiB, 2.77 k iops, 691.3 MiB/s dd: sequential write speed 1st run: 288.01 MiB/s 2nd run: 196.46 MiB/s 3rd run: 222.21 MiB/s average: 235.56 MiB/s IPv4 speedtests your IPv4: xxx.xxx.xxx.xxx Cachefly CDN: 46.26 MiB/s Leaseweb (NL): 34.94 MiB/s Softlayer DAL (US): 1.73 MiB/s Online.net (FR): 18.96 MiB/s OVH BHS (CA): 9.87 MiB/s No IPv6 connectivity detected -------------------------------------------------
------------------------------------------------- nench.sh v2018.04.14 -- https://git.io/nench.sh benchmark timestamp: 2018-04-27 11:09:51 UTC ------------------------------------------------- Processor: Westmere E56xx/L56xx/X56xx (Nehalem-C) CPU cores: 2 Frequency: 2493.748 MHz RAM: 5.8G Swap: 6.0G Kernel: Linux 4.13.0-38-generic x86_64 Disks: loop0 512M HDD vda 117G HDD CPU: SHA256-hashing 500 MB 3.982 seconds CPU: bzip2-compressing 500 MB 7.579 seconds CPU: AES-encrypting 500 MB 2.458 seconds ioping: seek rate min/avg/max/mdev = 76.4 us / 121.4 us / 19.3 ms / 169.6 us ioping: sequential read speed generated 15.5 k requests in 5.00 s, 3.78 GiB, 3.09 k iops, 773.6 MiB/s dd: sequential write speed 1st run: 276.57 MiB/s 2nd run: 310.90 MiB/s 3rd run: 238.42 MiB/s average: 275.29 MiB/s IPv4 speedtests your IPv4: xxx.xxx.xxx.xxx Cachefly CDN: 74.94 MiB/s Leaseweb (NL): 10.65 MiB/s Softlayer DAL (US): 1.06 MiB/s Online.net (FR): 21.84 MiB/s OVH BHS (CA): 6.21 MiB/s No IPv6 connectivity detected -------------------------------------------------
Comments
IMO the HDD version is better.
seek rate isn't too bad compared to the ssd server, sequential write speed is nearly doubled, and you get much more space.
And if you need really fast accessing disks, you better go with nvme, right?
How can the SSD version I have be worse than the HDD one...I don't think it's truly SSD. Maybe it's
SSD-enhanced HDD.
So those measurements you reported are from Netcup's HDD boxes? Thanks for the information
or oversell and/or bad configured raid
7 euro for a server with 120gb of actual ssd is not unbelieveable but it is pretty darn cheap. Cached HDD is a plausible explanation.
Yes, I find the performance very close to other SSD-enhanced HDD. I am just pissed that it was sold as SSD. It never felt SSD (I have previously tried consumer and server SSDs), but never investigated...
Another possibility could be that after months of use as a build server they somehow penalised my box; but this is still a miss on their side, building software is a perfectly legit use (you do not rent VPSes/root servers to keep them idle).
it still is a vserver that shares ressources esp. such as disk/ssd with other clients. so the numbers you provide are fine for a SSD. afaik their disks either if SSD or SAS are setup in ZFS arrays and bound to the compute nodes. so of course everything might depend heavily on the node your are on and the neighbours you are with... also other providers might use other setups, only thing to be sure of - anyone will involve caching strategies. :-)
I doubt they need to lie about it being SSD, the few results to be seen are reasonable enough esp. for the price and probably at least stable even over the years.
that said: if you're unhappy with the results it is the right thing to move on and choose another provider that might be a better fit. remember to actively cancel 31 days upfront to not be billed for another year!
Processor : Westmere E56xx/L56xx/X56xx (Nehalem-C)
Created January 7, 2010
Could be the sata ports is only 300Mb/s (SATAg2), and it feels way slower, or it's just a bad(ly configured) raid card.
I don't think they have cached HDDs, because performance would be worse.
Alright, I posted here to know more about this. I didn't know performance could vary so wildly but I believe it for the reasons you mentioned. My concern was more about "what makes it special as a root server vs a shared VPS?"
Yes, I already did that even before posting here on LET. As I said in OP, it was a good experience for these few years, they have been professional all the time.
I bought a cheap root server, I thought it would have been better than a KVM VPS but I have to conclude that performance-wise it was not.
I thought a root server just meant a server where you can login as root. I.e. basically any VPS. It differentiates the product from shared hosting, shell accounts, ftp storage, and other servers where you don't have root. It doesn't mean dedicated hardware. Maybe I'm wrong about this though.
The term is used mostly by German providers, it seems to me, so I figure it's customary in the hosting industry there.
As far as I can tell, your Root-Server belongs to the old series of Root-Servers at netcup, because the new (=present) series of Root-Servers have dedicated cores. The old series of Root-Servers are more like the new (= present) series of vServers. (Makes sense, right? :-) )
Edit: I imagine that @lemon has a new Root-Server. If so, then the comparison may not be accurate.
In Germany, usually "root servers" means a server that some resources are dedicated, mostly cpu/vcpu. Netcup is definitely this: their root servers do have dedicated cpus, while vservers share the cpu among other users of the node.
But there are some Greman providers that with the term "root server" they mean KVM and vservers, openvz.
Before you jump to the conclusion (and get pissed!) that it's definitely not an SSD, be aware that since you're in a virtualized environment, there's probably no sure way of determining whether the disk is HDD/SAS or SSD from within Linux. In particular, your test above with
lsblk
is not reliable in a virtualized environment.As @Falzo said, there are a number of factors influencing disk performance in a shared environment, and it seems unlikely that netcup would claim to sell you an SSD if it's not in fact an SSD.
Just curious, what result do you get from running the following script? I find that running it as root (
su -
) yields more meaningful results.(See https://serverfault.com/questions/551453/how-do-i-verify-that-my-hosting-provider-gave-me-ssds/551495#551495 for the source of this script.)
On my laptop, which has a HDD/SATA (7200 rpm), I get (running it as
su -
):On my Root-Server at netcup (purchased at Easter), which has a HDD/SAS, I get (again, running it as
su -
):Naturally, the values vary with each of run of the script, but this difference between my laptop HDD/SATA and my Root-Server HDD/SAS makes sense. I'll also say that the values vary more radically on the Root-Server, which also makes sense, given that it's in a shared environment.
To return to my question, what values do you get (running the script as
su -
)?I was thinking about Netcup Root-Server is there any better alternatives.
Except that the OP's "root server" (at netcup) doesn't have dedicated CPUs. (See my comment before yours.)
They list root servers and vServers, I rented a root server with virtual cores.
I think you nailed it. It was labelled root server but I find it hard to see any difference with a vServer.
I think you missed my reply to @Falzo, where I said that I believe it is like he said e.g. hard to determine for various reasons.
I know. I reported it because it adds to the uncertainty.
Think of it: there is no way to determine if it is SSD or not, then why selling it as SSD? I measure a read speed of 5 Mb/s. You can call it vynil for what I care, there is deception in the intent and implementation As I said, I am generally happy of the service they provided over the years, but at the end of it I am drawing some conclusions about some aspects of it.
Edit: for the records (and to avoid flames) I acknowledge it is SSD, just not a good one and/or not a good setup/performance for me
I made a quite comprehensive test in OP and also here: https://serverscope.io/trials/GVNr
Would that be enough to judge the quality of the medium?
Well, if it is SSD, then why not sell it as SSD? (Likewise, if it is SAS, why not sell it as SAS, and if it is NVMe, why not sell it as NVMe?)
I mean, it's not netcup's (or any provider's) fault if one can't completely reliably test for HDD vs SSD in Linux in a virtualized environment.
Listen, you're doubting that it's SSD. Fine. I was just trying to make a constructive suggestion of a further test. You're not interested. Fine. You're free to believe what you want.
@angstrom @deer76 calm down mates, there is no wrong here ;-)
this. the serverscope shows that everything outside caching does not look ssd worthy, however this most likely is just the underlying setup and crowd on that node.
Thanks @Falzo but I am soon to be banned anyway I feel like it's a bit toxic here
@deer76 what do you mean by new?
The benchmark I provided is the Eierkanone (Easter Special), but I had another one over 1 year ago, the speed was the same.
lol. you're either really new here and just unlucky in choosing your words. otherwise that special one implies that it just depends which ban you are evading? :-P
Here's a wild thought, which if correct, would support your suspicion that it's not SSD. (!) If you look closely at the description of the hard disk (Festplatte) in that old offer, you'll see that it says "120 GB SSD Performance". The addition of "Performance" now makes me suspicious: why not simply "120 GB SSD"? This makes me suspect that it may not be real SSD but is rather SAS.
For comparison, if you look at their current Root-Server offers, e.g., https://www.netcup.eu/bestellen/produkt.php?produkt=1769 , you'll see that the word "Performance" isn't added.
If "120 GB SSD Performance" was secretly SAS and wasn't equivalent to "120 GB SSD", then I would definitely say that that was a misleading description. (But perhaps my wild thought is incorrect.)
Don't be a little girl, you were given a mild warning for having a pop at someone completely out of context on an old thread you decided to necro, it could not have served ANY purpose but to create drama... that is toxic.
Put on your big boy pants.
Sorry, I overreacted -- I felt like you were acting a bit stubbornly.
Ironically, I'm beginning to think that you may be right that it isn't really SSD! (See my previous comment.)
this is a good find. the alternative to 'SSD performance' was only SATA back then. so the SSD performance thing simply might have been SSD cached something and no real SSD at all.
german providers used to trick around with terms all the time, 'rootserver' only being the most common example. also netcup did so with their 4x 123 GB RAID 10 before, which could lead to the idea you'd get 4 virtual devices to work with instead of just a single one on a raid10...
Indeed. Again, looking at that old offer, https://www.netcup.eu/bestellen/produkt.php?produkt=1121 , for the hard drive there was a choice between "120 GB SSD Performance" (the default, which the OP chose) and "240 GB SATA" (click on the box for the value of "Festplatte"), so "SSD Performance" very probably means something like "SSD caching but not real SSD", as you suggest (above, I suggested SAS, but SSD caching is more plausible).
Yeah, not a great practice, and rather disappointing, though nowadays, it's probably harder to get away with those naming games.
I definitely sympathize with anyone (e.g., the OP) who thought (a couple of years ago) that "120 GB SSD Performance" meant real SSD.
So I think that @willie got it right (see above):
Wouldn't be easier than the detective work, if someone (like OP for example) just opened a ticket and asked them if it is pure SSD or SSD cached setup?
Anyway, I think @angstrom finding is really showing what is going on here: the word "performance" is their way to tell "caching", just like they use their own terminology on "root" vs "vserver". Probably it's marketing term because I think 120GB pure ssd is very difficult to cost than 7€, even more if this is combined with 6GB of memory...
Finally, ~275MB/s average i/o for a vps costing 7€ per month, is not bad at all for the big majority of the potential users. In fact, anything over 120-150MB/s would be really good in this price range for such a product.
Yes, but 2-3 years ago when I picked them, I might have chosen the VPS labelled "root server with SSD performance" instead of another one (and I probably did), this was my point. I think it was a good VPS for the cost/quality ratio, just not so much on the marketing perhaps.> @lemon said:
OK. I used "new" as in this comment, they are all vServers but new root servers have dedicated cores (from what I have read on their offer and reviews).
And for the records, here are two benchmarks from my servers. The first is a vserver and the second a root server. Both servers do have SAS and their performance seem better than the SSD bench OP posted:
+++++++++++++++++++++++++++++
https://www.netcup.de/bestellen/produkt.php?produkt=1754
System Info
Processor : QEMU Virtual CPU version 2.5+
CPU Cores : 4
Frequency : 2493.744 MHz
Memory : 3965 MB
Swap : 3899 MB
Uptime : 105 days, 20:28,
OS : Debian GNU/Linux 8
Arch : x86_64 (64 Bit)
Kernel : 3.16.0-4-amd64
Hostname : xxxxxxxxx
Speedtest (IPv4 only)
Your public IPv4 is xxxxxxxx
Location Provider Speed
CDN Cachefly 63.0MB/s
Atlanta, GA, US Coloat 13.5MB/s
Dallas, TX, US Softlayer 11.3MB/s
Seattle, WA, US Softlayer 10.2MB/s
San Jose, CA, US Softlayer 9.84MB/s
Washington, DC, US Softlayer 2.75MB/s
Tokyo, Japan Linode 8.85MB/s
Singapore Softlayer 8.23MB/s
Rotterdam, Netherlands id3.net 58.6MB/s
Haarlem, Netherlands Leaseweb 81.3MB/s
Disk Speed
I/O (1st run) : 307 MB/s
I/O (2nd run) : 373 MB/s
I/O (3rd run) : 313 MB/s
Average I/O : 331 MB/s
+++++++++++++++++++++++++++++
https://www.netcup.de/bestellen/produkt.php?produkt=1762
System Info
Processor : Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GHz
CPU Cores : 2
Frequency : 2399.996 MHz
Memory : 5967 MB
Swap : 3899 MB
Uptime : 99 days, 18:10,
OS : Debian GNU/Linux 8
Arch : x86_64 (64 Bit)
Kernel : 4.4.98-3-pve
Hostname : xxxxxxxxxxxxxxxxxxxx
Speedtest (IPv4 only)
Your public IPv4 is xxxxxxxxxxxxxxxxxxxxxx
Location Provider Speed
CDN Cachefly 39.4MB/s
Atlanta, GA, US Coloat 2.48MB/s
Dallas, TX, US Softlayer 11.4MB/s
Seattle, WA, US Softlayer 10.2MB/s
San Jose, CA, US Softlayer 9.22MB/s
Washington, DC, US Softlayer 8.74MB/s
Tokyo, Japan Linode 8.46MB/s
Singapore Softlayer 6.79MB/s
Rotterdam, Netherlands id3.net 51.7MB/s
Haarlem, Netherlands Leaseweb 107MB/s
Disk Speed
I/O (1st run) : 99.5 MB/s
I/O (2nd run) : 233 MB/s
I/O (3rd run) : 353 MB/s
Average I/O : 228.5 MB/s
Wouldn't that be way too easy?! ;-) (You're right, of course.)
Okay, but the difference is that "caching" is a technical term (so "SSD caching" is something that technical people generally understand the meaning of), whereas "performance" is simply an everyday word which can mean different things in different contexts.
But, you're right, if we want to be kind to netcup, we could say that they understood "performance" to mean "caching", and they also thought that everyone would understand "performance" in this way. (But I'm skeptical about the second part of the previous sentence!)
Here, I completely agree. It was/is a good price for a good product.