rootnerds [DE] VPS review & benchmark


After offering to do a benchmark and a small review to them rootnerds quickly reacted and sent me the data and access info needed for one of their VPS. Unfortunately they did not react to my request to provide the product name and indeed the VPS I tested seems to not match any of their VPS offers which I find regrettable. The provided VPS had 4 vCores and 2 GB memory (see vpsbench part below for details).

I am not and have never been a customer of rootnerds and approached this test totally unbiased.

Version 1.02b, (c) 2018+ jsg/moerm (->
Machine: i386, Arch.: i686, Model: Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz
OS, version: Linux 2.6.32, Mem.: 2.0 GB
CPU - Cores: 4, Family/Model/Stepping: 6/79/1
Cache: 32K/32K L1d/L1i, 256K L2, 20M L3
Std. Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat
pse36 cflsh ds acpi mmx fxsr sse sse2 ss htt tm pbe sse3 pclmulqdq
dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm
pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline aes xsave
osxsave avx f16c rdrnd
Ext. Flags: fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm pqm
fpcsds pqe rdseed adx smap intel_pt nx pdpe1gb rdtscp lm lahf_lm lzcnt

--- proc/mem/performance test single core ---
64 rounds~ 1.00 GB -> 178.92 MB/s
--- proc/mem/performance test multi-core ---
8 times 64 rounds ~ 8.00 GB -> 89.64 MB/s
--- disk test ---
Sequential writing 974.54 MB/s
Random writing 1.367 GB/s
Sequential reading 3.079 GB/s
Random reading 2.829 GB/s
--- network test - target 100KB 1MB 10MB -> 64 MB --- AU,MEL:
308.1 Kb/s 427.5 Kb/s 469.9 Kb/s -> 494.0 Kb/s IN,CHE:
515.4 Kb/s 426.1 Kb/s 532.6 Kb/s -> 567.8 Kb/s UK,LON:
643.5 Kb/s 546.4 Kb/s 649.1 Kb/s -> 788.6 Kb/s DE,FRA:
506.4 Kb/s 443.5 Kb/s 798.0 Kb/s -> 698.5 Kb/s IT,MIL:
548.6 Kb/s 673.5 Kb/s 1.0 Mb/s -> 786.2 Kb/s FR,PAR:
357.3 Kb/s 469.7 Kb/s 783.6 Kb/s -> 851.4 Kb/s RO,BUC:
362.8 Kb/s 388.1 Kb/s 1.2 Mb/s -> 854.8 Kb/s GR,UNK:
355.9 Kb/s 307.5 Kb/s 260.8 Kb/s -> 316.3 Kb/s RU,MOS:
174.6 Kb/s 215.6 Kb/s 253.2 Kb/s -> 293.7 Kb/s US,DAL:
803.1 Kb/s 503.2 Kb/s 902.3 Kb/s -> 826.0 Kb/s US,SJC:
492.0 Kb/s 392.1 Kb/s 602.9 Kb/s -> 711.5 Kb/s US,WDC:
355.4 Kb/s 552.6 Kb/s 666.8 Kb/s BR,SAO:
1.0 Mb/s 962.8 Kb/s 839.7 Kb/s -> 1.3 Mb/s JP,TOK:
269.1 Kb/s 501.4 Kb/s 1.0 Mb/s -> 1.2 Mb/s NO,OSL:
545.5 Kb/s 658.5 Kb/s 1.3 Mb/s -> 1.4 Mb/s


The VPS is based on a relatively recent 8 core processor and provides 4 i686 vCores, 2 GB memory, and a 20 GB disk. The VPS is obviously OpenVZ based as the kernel version suggests and the processor test confirms (more see below).

The processor and memory performance is theoretically/potentially quite OK but practically subpar and closer to an Atom than to a Xeon. The disk performance is very good and the VPS almost certainly either has a rather good SSD or some good SSD cache. The network however is VERY poor.


As the VPS does not seem to match any of rootnerds standard VPS products and has a spec that is not low-end (4 vCores, 2 GB memory) it seems reasonable to assume that rootnerds provided a VPS on a node with some kind of (maybe LET) specials and/or the VPS under test being a somewhat upgraded VPS for testing purposes.

One interesting detail is shown in the proc/mem test. In single core mode the results are quite normal for that kind of processor but in multicore mode the VPS brutally fails. Rather than showing roughly 3 times the single core performance (as would be expected) it actually gets slower! This clearly a result of OpenVZ and especially of rather excessive overbooking.

Well noted, vpsbench has been made expressly for VPSs and always leaves quite some breathing room for other clients on the tested node. This also means that vpsbench doesn't push the OS multithreading in multi core tests. If the multi core test shows such a sad result that's a clear indication for brutal overload be it by abusers on the node or by crude overselling which is the case in this test.

Regarding the disk there isn't much to say beyond "really great". A very nice result.

Regarding the network however things get really ugly again. The node seems to be in accelerate's DC in Frankfurt,DE yet even the result for the Fra,DE target shows but a measly ~ 700 Kb/s! Funnily some of the "exotic" targets like Brazil and Japan show better results. Also for some weird reason the test doesn't even complete for the Washington, US target.

Also the "ladder" looks strange. Normally there is a clear increase from 100 KB to 1 MB to 10 MB and a more limited increase after 10 MB. In this case however the rise is quite slow which indicates some serious problems either with very poor connectivity or highly congested bandwidth. Second or third rate "filler" traffic via cheapest route and with high latency come to mind.


Stay away!

The node hardware seems to be OK and the disks are even really nice but the network is seriously poor and so is rootnerds (over)booking. Looking at the pricing on their web site it pretty much comes down to "VERY bad performance for a quite high price".

Top News