Netcup Special R2000 Plus SSD G8

lemon

@ehab said:
if only this had cpu pass through, would have been super.

for just 8€/month extra the support enables it for you.

jvnadr

willie said: Hmm interesting, they must be overallocating iops in that case, which is probably ok to an extent. Anyway I've never felt constrained by raw disk i/o (especially sequential) on these servers. It's always been about cpu and network speed. There are some applications where I'd want more IOPS, but that really wants an SSD rather than faster HDD's.

Talking from my experience in a production system and not with dd testing, their HDD performance is very good, they use raid50 IIRC. I am very pleased with the performance for the price, and for being a semi-managed service (managed at server level, e.g. monitoring node performance, disk integrity etc.).

tr1cky

@Neoon said:
Unixbench returned a result of 3066 benchmark points, I actually expected a bit more for 4 dedicated cores, around somewhat 4k.

Decent with 24GB memory, could allocate everything, I/O is fast, network is fast...

The current i3 at Kimsufi reaches 4k, so yea.

As long Kimsufi has no deals, this may be a better option atm.

If they had decent network, yeah:

jvnadr

tr1cky said: If they had decent network, yeah:

Are you referring to OVH or Netcup? If OVH, then, yes, their network is less or more, mediocre: not the worst but not the best. But on Netcup, I get some of the fastest speeds almost always in Europe than from many of my servers/vps. And never noticed any serious hickups.

vimalware

Argh. netcup billing is confusing.

I tried to add a domain to my cart but it shows the 'monthly' (divided by 12) price in the right hand column. How am I ever going to take advantage of this 30 Eur coupon when I can't get the total above 30?

I give up for tonight.

FWIW I even tried adding in a VPS to the mix; coupon isn't allowed for two products it seems.

Neoon

You need to add 3x. de domains + servery.

Could be that they limited it, after we abused it and just paid 2EUR.

Skyrider

Seeing the offer still stands and I happen to bought the R2000 offer, here's the ServerScore benchmark with SSD rather than SAS.

https://serverscope.io/trials/e19P

Also, with Azenot's SAS benchmark, I ran the same under SSD:

• https://pastebin.com/ZkNFpyft (SAS Benchmark)
• https://pastebin.com/0Bew5dLL (SSD Benchmark)

angstrom

@Skyrider said:
Seeing the offer still stands and I happen to bought the R2000 offer, here's the benchmark with SSD rather than SAS.

I had also noticed that the SSD version was still available, but the SAS version isn't.

Falzo

@Skyrider said:

could you run this fio command on it?

fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --filename=test --bs=4k --iodepth=32 --size=4G --numjobs=4 --readwrite=randrw --rwmixread=50

if it happens to estimate a long time for running simply ctrl-c it after a few minutes or so ;-)

Skyrider

@Falzo said:

@Skyrider said:

could you run this fio command on it?

fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --filename=test --bs=4k --iodepth=32 --size=4G --numjobs=4 --readwrite=randrw --rwmixread=50

if it happens to estimate a long time for running simply ctrl-c it after a few minutes or so ;-)

Saw you asked this before on page 1 but sure, here are the results:

https://pastebin.com/Sy5kGe6n

Getting tired of LET putting all my text into 1 single line, so posted it into a pastebin instead.

Falzo

@Skyrider said:

@Falzo said:

@Skyrider said:

could you run this fio command on it?

fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --filename=test --bs=4k --iodepth=32 --size=4G --numjobs=4 --readwrite=randrw --rwmixread=50

if it happens to estimate a long time for running simply ctrl-c it after a few minutes or so ;-)

Saw you asked this before on page 1 but sure, here are the results:

https://pastebin.com/Sy5kGe6n

Getting tired of LET putting all my text into 1 single line, so posted it into a pastebin instead.

thanks, that was fast! pastebin is totally fine...

for being the same as on the first page, only kind of - the result there is from the SAS box. so here I was really curios for the same run on SSD, which differs quite a lot I'd say.

about 10k/10k iops in a 50/50 random rw test with 4k blocksize vs 360/360 on their SAS thing is a remarkable difference I'd say ;-)

while this clearly is a big plus on the SSD line, 80GB is a bit stingy for that box though.

vimalware

@Falzo said:

about 10k/10k iops in a 50/50 random rw test with 4k blocksize vs 360/360 on their SAS thing is a remarkable difference I'd say ;-)

while this clearly is a big plus on the SSD line, 80GB is a bit stingy for that box though.

I don't know about raw compute power, but ram+ssd-wise, it works out to about a 3rd of the 39euro i7 dedis(2x240G consumer ssd edition).

But I eventually decided to add a single Datacenter SSD to my hetzner auction dedi instead ( LVM-thin pool).

Skyrider

I edited my second above posts to make it more readable (added pastebin links with results).

Also, with Azenot's SAS benchmark, I ran the same under SSD:

https://pastebin.com/ZkNFpyft (SAS Benchmark)
https://pastebin.com/0Bew5dLL (SSD Benchmark)

The only thing i noticed that the single CPU benchmark is relatively low as seen on: https://serverscope.io/trials/e19P

• UnixBench (all CPUs): 2713.7
• UnixBench (one CPU): 965.9

neik

@Skyrider, would you mind running Geekbench 4?

Skyrider

@neik said:
@Skyrider, would you mind running Geekbench 4?

https://browser.geekbench.com/v4/cpu/9386211

If you'd like any other tests, let me know!

neik

Thanks, mate!