All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Short netcup root server review & some thoughts
I just got my hands on a netcup root server (KVM), the small "1000 G8" model with 2 dedicated cores and 8 GB dedicated RAM for a bit less than 7€/month.
In fact, there are two particular reasons why I write this short review besides providing some benchmark results. The first one is netcups reputation as being very picky and bureaucratic and the second one is some thoughts about a possibly quite interesting technicality.
But first the numbers, as usual from vpsbench:
Machine: amd64, Arch.: amd64, Model: Intel(R) **Xeon(R) Gold 6140** CPU @ 2.30GHz
OS, version: Linux 4.19.0, **Mem.: 7.985 GB**
CPU - Cores: 2, Family/Model/Stepping: 6/85/4
Cache: 32K/32K L1d/L1i, 2M L2, 16M L3
Std. Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat
pse36 cflsh mmx fxsr sse sse2 ss htt sse3 pclmulqdq ssse3 fma cx16
pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline aes xsave osxsave
avx f16c rdrnd hypervisor
Ext. Flags: fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm mpx pat
pse36 rdseed adx smap clflushopt clwb sha umip pku syscall nx pdpe1gb
rdtscp lm lahf_lm lzcnt
--- proc/mem/performance test single core ---
64 rounds~ 1.00 GB -> 197.16 MB/s
--- proc/mem/performance test multi-core ---
4 times 64 rounds ~ 4.00 GB -> 499.21 MB/s
--- disk test ---
Sequential writing 257.69 MB/s
Random writing 168.39 MB/s
Sequential reading 803.42 MB/s
Random reading 821.53 MB/s
--- network test - target 100KB 1MB 10MB -> 64 MB ---
http://speedtest.mel01.softlayer.com/downloads/test100.zip AU,MEL:
661.9 Kb/s 1.9 Mb/s 6.4 Mb/s -> 15.0 Mb/s
http://speedtest.che01.softlayer.com/downloads/test100.zip IN,CHE:
1.3 Mb/s 3.5 Mb/s 11.3 Mb/s -> 31.0 Mb/s
http://speedtest.lon02.softlayer.com/downloads/test100.zip UK,LON:
13.4 Mb/s 35.7 Mb/s 104.6 Mb/s -> 266.1 Mb/s
http://speedtest.fra02.softlayer.com/downloads/test100.zip DE,FRA:
48.3 Mb/s 92.3 Mb/s 234.8 Mb/s -> 503.2 Mb/s
http://speedtest.mil01.softlayer.com/downloads/test100.zip IT,MIL:
15.4 Mb/s 39.4 Mb/s 125.7 Mb/s -> 208.0 Mb/s
http://speedtest.par01.softlayer.com/downloads/test100.zip FR,PAR:
14.8 Mb/s 40.5 Mb/s 120.2 Mb/s -> 269.9 Mb/s
http://speedtest.ftp.otenet.gr/files/test100Mb.db GR,UNK:
2.2 Mb/s 9.3 Mb/s 34.9 Mb/s -> 92.5 Mb/s
http://93.95.100.190/test100mb.bin RU,MOS:
3.8 Mb/s 9.5 Mb/s 11.8 Mb/s -> 11.7 Mb/s
http://speedtest.dal05.softlayer.com/downloads/test100.zip US,DAL:
1.6 Mb/s 4.4 Mb/s 14.3 Mb/s -> 37.5 Mb/s
http://speedtest.sjc01.softlayer.com/downloads/test100.zip US,SJC:
1.3 Mb/s 3.5 Mb/s 11.9 Mb/s -> 32.1 Mb/s
http://speedtest.wdc01.softlayer.com/downloads/test100.zip US,WDC:
3.1 Mb/s 7.3 Mb/s 20.3 Mb/s -> 51.4 Mb/s
http://speedtest.sao01.softlayer.com/downloads/test100.zip BR,SAO:
1.0 Mb/s 2.8 Mb/s 9.3 Mb/s -> 24.2 Mb/s
http://speedtest.tokyo.linode.com/100MB-tokyo.bin JP,TOK:
1.0 Mb/s 2.5 Mb/s 7.4 Mb/s -> 18.9 Mb/s
http://speedtest.osl01.softlayer.com/downloads/test100.zip NO,OSL:
5.7 Mb/s 15.6 Mb/s 47.1 Mb/s -> 103.3 Mb/s
(Unless you are interested in technical details, the last number (after the "->") of the network tests is the one you want to care about)
Note: If you don't like texts with more than a couple of lines, just skip to the TL;DR at the end.
I'm really pleased with those results. While the network results aren't smashingly impressive I'm quite pleased anyway. Reason: The official "dedicated" bandwidth for the VPS is iirc 200Mb/s and that (and some more) is indeed achieved with the major european targets. More impressioning however are the "odd" and far away destinations like Tokyo, Melbourne, and even Chennai; the results for those targets are much better than what I usually see. The grain of salt is the US targets; around 30 - 50 Mb/s to those is nothing to brag about.
The processor and memory results are what one would probably expect from a Xeon Gold processor running at 2.3 GHz. Respectable but not earth-shattering.
But what I'm really excited about is the disks. netcup offers their root servers with either SSDs or SAS (hw. Raid 10) drives and the SAS option is about 8 times the size of the SSD. That's a point that got me thinking because it's an interesting trade-off: you can go for higher speed or for more storage.
Obviously, most would probably think, SSD is always better because it's faster and faster disks ~ faster web applications.
Well, not always. disks are just one factor, an important one for many use cases but still just one among others. Having a whopping 8 GB RAM - in a cheap 7€/month VPS - opens a door to some quite interesting options ...
Looking closer at typical scenarios, we find that, unlike with a desktop, boot time isn't that important; after all, servers usually run for months or even years. Looking at the disks we find that only a couple of hundred MB or maybe some GB are "hot". Whether a database starts up in 1 s or in 5 s isn't important; even the database files themselves aren't that hot because in most scenarios INSERTs and UPDATEs (writing) are by far less frequent than SELECTs (reading) plus it's the indices that are hot and not the data.
Or let's look at a web site like this one: PHP, a DB and a web server. What are the choke points? Can we make that site really much faster by using SSD instead of spindles? Only mildly. If we have plenty of RAM though we can use a lot of caching which equates to the difference between "low hundreds of requests per second" vs. "thousands of req/s - easily".
Reason: smart caching makes the DB much faster and smart caching also takes away quite some of the pain of having each and every request run through a slow interpreter (PHP, Python, Ruby are by far the worst cycle hogs on a typical web server).
I have some quite active web sites that run on 1 GB VPSs incl. DB. It works but it's certainly not fast. Adding just 1 GB to those would allow for some caching that would speed up those sites very considerably.
Well, that netcup root server has 8 GB RAM. Which, if properly used, translates to very significant speed increases.
But there is another point: SSD life time. SSDs are fast - but they are also "sensitive" and have only so much life time. Looking at what is written most frequently on a typical server we find that the typical culprits are /tmp (e.g. http session data) and some /var sub directories. None of those are large and the contents are usually temporary anyway. Translation: ideal RAM disk candidates - if you have the RAM ...
And there's 320 GB on those spindles. I like that a lot. One application I have in mind is a temporary backup stage where I collect, say, a weeks worth of backups from different VPSs and then pack them up and push them to 1fichier or C14. And btw. they are actually quite fast for spindles; I've seen a lot worse.
The other point: "netcup is picky and bureaucratic".
Yes and no. When that VPS was ordered something wasn't to netcups liking, something didn't match well enough. So we called their hot line. First big plus: No long waiting loop; we had a support person on the line within 20 or 30 seconds. Funnily the simple fact that we called them (from the number we had provided when ordering) already seemed to solve some of their worries. Next they asked some questions and we answered those. Done. We had told them that the payment would be transferred as soon as we were cleared and they activated the service. Simple and painless. And, big plus: They were brutally honest ("Yes, we are picky. To offer cheap services that are also good means that one can't tolerate abusers. And we also very seriously follow the law and protect our customers") but they were friendly, really friendly.
TL;DR
If you are a normal TOS abiding user netcup is one of the best places I know to get a really nice VPS. Yes, they check you seriously but once you are a customer they offer great stuff and great service at very attractive prices and their support reacts quickly. What more could one ask for?
Highly recommended.
Edit: Another small detail: an additional IP is available for 1€/mo (plus 5€ one time setup). That's quite a good price and well below what many other providers demand.
Comments
Can I install VirtualBox on their root server?
Nice review
Why not QEMU/KVM?
I don't know, I didn't try - or even think about that. Kindly note that the netcup "root servers" are not dedis but KVM VPSs, albeit ones with dedicated Cores and RAM.
I don't think it makes a lot of sense to install a VM within a VM.
If I recall correctly, they will charge a fee for enabling nested virt.
This is what I've heard, but I had a root server with vmx enabled. I tried it once and ran a nameserver via internal NAT. It wasn't all that worth it.
I think they allow it nowadays on the Root Servers, but not fully sure.
Regardless, @jsg , check out this Hetzner Auction server: 955605 (just paste into the search box on the Auction page)
Xeon (9k Passmark), 32GB ECC, 2x3TB Ent. HDD, Unmetered 1GBit (Intel NIC) at 31.93 Euro.
Thanks, but why should I be interested in this context? I know that Hetzner has (sometimes very) attractive dedi offers but this thread is about a VPS, albeit one with dedicated cores and memory.
Which basically means it's a dedi carved up into pieces with dedicated resources. So, why not just get a dedi with ~68% more CPU, same RAM, ~120% more disk space, and unmetered BW?
If you need just one VPS, then I can understand; but this is LET - multiple idlers are the norm.
Because there is a significant difference between 8 vs. 2 cores, 32 vs. 8 GB memory, and 32 vs 7 Euro/month. And because for quite many use cases that would mean to waste about 20 €/mo. For example I would see that only very few web sites really need more than 2 - 4 cores, more than 8 GB memory, and 100 GB disk space.
Something similar goes for traffic. I guess that 95+% of web sites do not use even 50 Mb/s and 98+% don't use the 200 Mb/s that that netcup root server offers.
Of course looking at numbers and providing offers somewhat drives us to want ever more and to think that a VPS with 500 Mb/s is more desirable than one with "only" 100 Mb/s - but hardly anyone really needs more than 100 Mb/s - so why should we pay for it? Just to get horny over numbers?
Btw. I do have a dedi with 16 cores and 32 GB memory (and relatively decent disk space) and pretty much unlimited traffic (and enough IPs). But I also want some VPSs. The way I see it I can get 2 really decent VPSs (like the netcup one) plus 4or 5 cheap small ones (1C, 512 MB, 10 or 15 GB disk) for the same price of that hetzner dedi - and that is more attractive for many use cases.
Whatever, this thread is about the netcup root server 1000 G8 VPS which is a really nice and solid one. I'm enchanted and that doesn't happen often to me wrt a VPS.
"short"
The review is short.
It's the thoughts around it that are not.
Finally a nice word about netcup from someone.
I have their Easter special, which is a Root-Server comparable to RS 1000 but with E5-2670V3 and 400 GB SAS. Also €6,99/month. Also a great deal.
Absolutely. I don't care that much about having the most up to date CPU and a 2670v3 is a very decent processor. What I do like - and notice with that root VPS - though is the dedicated core.
All in all that root server, although it has spindles and not a SSD, feels more responsive than most KVM VPS (incl. those with SSD) I have used. It pretty much feels like an Atom dedi but with a much faster processor.
But I was also impressed by and liked a lot their quick and friendly support.
Thanks for the nice, detailed review!
I have one question about connectivity that members might be able to answer. I've heard and read that Netcup colocates its hardware in a Hetzner-operated datacenter. Does this mean they use Hetzner for their uplinks (with the same transit/peering arrangements as Hetzner)? Or does Netcup operate their own network?
From what (little) I know so far, netcup's servers are in Nuremberg and not "near" Nuremberg (Falkenstein) like Hetzner. While Nuremberg isn't exactly a major point on the european network map it does have some not insignificant colocations/DCs and it's not too far away from Frankfurt (ca. 250 km I guess) and shouldn't add more than a couple ms latency.
If nobody else has more information I can do a little testing tomorrow and hopefully tell you a bit more if needed.
They have their own network. You can compare them here:
Edit: here they also say they use their own datacenter.
Short answer is no. Even if you can install virtualbox without enabling vt-x capabilities with some tweaking (disable hardware virtualization), the performance will be sooo poor...
Have in mind that their root servers are in fact KVM instances with dedicated cpu cores and memory and shared SSD hard drive. They offer nested capabilities, but you will be charged with a monthly fee, extra from their normal price. And IMO, there will always a performance penalty and not an unnoticed one when you do KVM over KVM.
On the other hand, you can install container virtualization like OpenVZ, LXC and docker that will work like a charm. I have installed Proxmox in two root servers from netcup dividing it to several LXC containers and all are working more than fine.
@sgheghele
Hmmm, NTT, Telia, and Anexia explain what I saw in my tests, in particular the nice connection to South and East Asia.
And that's (Anexia) also where netcup seems to have their Nuremberg DC area.
All in all not top class but certainly really decent. I'm quite a happy customer.
Wait, so Hetzner gives you a crazy high 20TB of traffic with their cheapest VPS, and here comes netcup and doubles it to 40TB? With a VPS that costs less than 3 Euros? I wonder what happens if you actually use it.
??? 3 Euros? The cheapest netcup root server is the one I reviewed here and it costs a bit less than 7€/mo (with annual contract).
20 TB, 40 TB ... whatever. I guess that most users don't even really use 1 or 2 TB. I guess those TB numbers are largely marketing or "racing" competition numbers. Largely pure theory.
Similarly I presume those 200 Mb/s per VPS to be mainly a barrier to keep the kind of users away who would, say, with a porn site, bring the node to a lame duck crawl for all the other VPSs on it.
My understanding is that netcup mainly targets small companies, freelancers and bigger hobbyists (running e.g. well grown communites) with those root servers. For us who are within that pattern reliability, consistent performance, reasonable quality at an attractive price is important, not "3 GHz would be better!" or "hell, I need 2 Gb/s lines" and it doesn't matter at all whether netcup provides 5 TB, 20 TB, or 50 TB with those VMs.
For perspective: A friend of mine runs a medium size community with PHP and MySql (Percona iirc) roughly the size of LET I guess on a 2 cores, 2 GB, 3 or 5 TB traffic KVM. Typically his load is about 0.5 and 1.5, used mem is in the mid hundreds of MB and his traffic is pretty much always below 1 TB (but there they have very few images and no videos, so I guess LETs traffic might be in the mid 1 digit TB).
I mention him/his site because he often mentioned that his VPS is "like the ocean. Sometimes steady, sometimes big waves", meaning his VPS is pretty much defined by what the other VPSs on the node currently do - and that's one of the points with netcups root server I like: No (or very low) "waves" and very steady sailing. Exactly what companies, professional freelancers, etc. need. How many tens of TB of traffic is absolutely not a relevant issue in I guess 98+% of cases..
@jsg: This guy @jgillich is either trolling or not paying attention.
Because I want to use windows, and I am more familiar with windows, and because I am more familiar with VirtualBox.
Tl;dr, no kvm, yes openvz
You can use Windows natively on these VPS, no need for double virt "tricks".
With that said, VirtualBox's performance sucks ass.
Windows'sTrojan's performance as well.Great review & thoughts! As you said as long as you have enough ram storage speed isn't what matters the most for many uses.
Never used them yet, but they seem worth trying!
Wrong, always.
SSD > always than any production mass-market SAS.
If you got unloaded empty node, it does not make SAS better then SSD.
Wrong.
For what needs? For what usage?
My all HDDs died, mostly all bought in different time, for different purposes. Bad sectors, just bad drivers, bad hardware, etc, can continue counting reasons why they died.
Not yet any of SSD since 2009 died in my experience. No one.
For example, right now I have 300TBW.
I was worried a lot about this TBW.
Even on my PC where I have a lot of videos (4k) and backups - I consume daily around 60-80GB of data. And its' really a lot of transfers here and there.
70GB * 365 = 25.5TB / y
300 / 25.5TB = ~11Years.
Last time when I bought a new SSD for another purpose = was around 2-3 years ago.
I switch or buying more disk space every 2-3 years.
Every 2-3 years I see a huge price drop or for storage space, or for quality, or for a prolonged period because of progress around SSD / NVMe.
========================
And this is on my real PC where I do a lot of work.
What about a server? For example web-server.
It's only very very specific cases when you will consume a lot of your SSD resource.
@desperand
You bring up an interesting point and I know that there are different "schools of thought". Here is why I came to my conclusion:
One can't put a cheap low-end spindle drive next to an expensive enterprise drive. Similarly - and even more so - one can't put a cheap TLC SSD next to an enterprise SLC SSD.
And there is reality. Will a hoster spend considerably more money for a better SSD? In many cases not, in many cases price is the major criterion, read, you'll find cheap TLC (and soon QLC) SSDs rather than SLC SSDs.
This also increases the consideration/pressure from above. "SSD" for most hosters isn't a technical factor but one of marketing and costs that is, one the one hand he (often) needs to have products with SSD but on the other hand he needs to keep costs low. The result more often than not is cheap TLC SSDs.
But the technical environment is more important and critical and there the probably most decisive factor in terms of life expentancy is temperature. SSDs are considerably less heat tolerant than spindles but the temperatures in a system in a DC are higher. One shouldn't be fooled by all the climate controlling and max 22°C and the like in a DC. The decisive point is inside a system and the fact that usually multiple disks are relatively densly packed. For spindles that's usually no problem at all, for SSDs however that's a problem.
Another very important point, particularly with VPSs, is that in a hosting environment many customers, often with quite different work loads, use storage which puts additional stress on drives (with SSDs being more sensible, among other reasons due to their internal overhead and structure).
I recently came across an offer from @Prometeus that I found interesting, attractive, and indicative of something we might see far more often in the near future: A combo, a VPS with some SSD space (like 20 GB) plus some spindle based capacity (like 500 GB). They probably have good reasons (and years of experience) to have come to that conclusion as it addresses both, the (perceived or real) need of customers for speed as well as the (usually real) need for large storage capacity - for a tolerable or even attractive price.
But you are right. The life time of SSDs has indeed increased and so did reliability, in particular at the mid and higher end (SLCs and, somewhat less, MLCs). And I think that in the long run SSDs will take over completely and spindles will become all but extinct (Side note: unless a miracle happens spindles, which are already at the outer edge of physics and engineering, are about to hit the wall while SSDs still have lots of room, e.g. smaller gate size, cheaper production processes, etc).
Last but not least - and that's why I like and emphasized the 8 GB ("lots of") RAM in the netcup root servers: On a server I don't care about boot up time; I do however care a lot about e.g. a DB, session stores, web caching, etc, and there RAM is the optimum, easily outperforming even SSDs (well, if allocated and used wisely).
So I'll stick to my view and what I've written in the review - but in a probably quite near future I'll highly likely move to your view.