All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
3.5 storage VPS comparison (NL, IT, RO) incl host_c
I confess it, I just had to benchmark my @host_c storage VPS, and being at it, why not compare it to the others I have? OK, one is a dedi, so out for reasons of fairness. The others though all are VPS.
- @LiteServer (liked it a lot and still do. it served and still serves me well and reliably)
- @c1vhosting (never really trusted it but it turned out to be quite reliable though slow)
- @Calin /ihostart (halfway trusted it but I didn't have high expectations. Quite reliable too though)
- @host_c (really good provider, I expected a at least decent product, and that's what I got)
Why '3.5' in the title? Because benchmarking a large storage VPS takes a lot of time and unfortunately the c1v VPS's disk is snail slow, so I really only got one full (fresh) benchmark run out of it so far (will update after a couple of days if justified). But hey, all the tested VPSs didn't run more than 10 - 20 runs, about half of which were with the large storage drive and the rest with the boot drive.
Re. the prices I'm not sure anymore about all the VPSs's prices. the host_c VPS is about $45/yr, the Liteserver one is about €90/yr, Calins, iirc was about €35 or €40/yr, and the c1v yet a bit less, maybe about €30/yr.
The storage drive size are
- Liteserver: 1 TB (no separate boot drive)
- c1v: 2 TB (no separate boot drive)
- Calin: 4.5 TB + 20 GB boot drive
- host_c: 5 TB + 20 GB boot drive
which also means that for both the Liteserver VPS and the c1v one it didn't even make sense to test both drives (as they are just partitions on the same large storage drive). For the Calin VPS and the host_c VPS the two have been benchmarked, and are listed separately.
So here you go, as usual processor and memory first:
Liteserver
Machine: amd64, Arch.: amd64, Model: AMD EPYC 7571 32-Core Processor
OS, version: FreeBSD 13.0, Mem.: 1.990 GB
CPU - Cores: 1, Family/Model/Stepping: 23/1/2
Cache: 32K/64K L1d/L1i, 512K L2, 64M 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 sse3 pclmulqdq ssse3 fma cx16 sse4_1
sse4_2 popcnt aes xsave osxsave avx f16c rdrnd hypervisor
Ext. Flags: syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm lahf_lm cmp_legacy svm
cr8_legacy lzcnt sse4a misalignsse 3dnowprefetch osvw perfctr_core
AES? Yes
Nested Virt.? Yes
HW RNG? Yes
ProcMem SC [MB/s]: avg 133.7 - min 44.9 (33.6 %), max 225.9 (169.0 %)
ProcMem MA [MB/s]: avg 221.1 - min 204.9 (92.7 %), max 240.4 (108.7 %)
ProcMem MB [MB/s]: avg 225.0 - min 209.4 (93.1 %), max 239.7 (106.5 %)
ProcMem AES [MB/s]: avg 787.7 - min 681.7 (86.5 %), max 843.9 (107.1 %)
ProcMem RSA [kp/s]: avg 50.8 - min 44.9 (88.4 %), max 55.0 (108.3 %)
Certainly not overwhelming, but then for a backup VPS 1 single vCore of a somewhat older Epyc is certainly good enough. The large spread does suggest though that that node is, um, "loaded".
2 GB memory is nice and the AES and RSA results are OK to mildly decent.
c1v
Machine: amd64, Arch.: amd64, Model: QEMU Virtual CPU version 2.5+
OS, version: FreeBSD 13.2, Mem.: 1.942 GB
CPU - Cores: 1, Family/Model/Stepping: 15/107/1
Cache: 32K/32K L1d/L1i, 2M L2, 16M L3
Std. Flags: fpu de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36
cflsh mmx fxsr sse sse2 sse3 ssse3 cx16 sse4_1 sse4_2 x2apic popcnt
aes hypervisor
Ext. Flags: syscall nx lm lahf_lm
[PM-SC] 315.37 MB/s
[PM-MA] 288.37 MB/s
[PM-MB] 315.08 MB/s
[P-AES] 92.00 MB/s
[P-RSA] 56.82 kp/s
Hmm, at first sight a decent vCPU, much faster than the LS one - but not when looking closely. The AES result is plain ridiculous and the PM-MA results being lower than the PM-SC (single core) result doesn't look trust inspriring.
Calin
Machine: amd64, Arch.: amd64, Model: Intel(R) Xeon(R) CPU E5-2630 v4 @ 2.20GHz
OS, version: FreeBSD 13.1, Mem.: 1.442 GB
CPU - Cores: 1, Family/Model/Stepping: 6/79/1
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 sse3 pclmulqdq vmx ssse3 fma cx16
pdcm 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 rdseed
adx smap umip syscall nx pdpe1gb rdtscp lm lahf_lm lzcnt
AES? Yes
Nested Virt.? Yes
HW RNG? Yes
ProcMem SC [MB/s]: avg 135.9 - min 48.8 (35.9 %), max 224.6 (165.3 %)
ProcMem MA [MB/s]: avg 212.6 - min 204.7 (96.3 %), max 220.2 (103.6 %)
ProcMem MB [MB/s]: avg 213.6 - min 209.4 (98.0 %), max 220.9 (103.4 %)
ProcMem AES [MB/s]: avg 494.9 - min 488.3 (98.7 %), max 499.5 (100.9 %)
ProcMem RSA [kp/s]: avg 54.3 - min 48.8 (89.9 %), max 58.2 (107.2 %)
Funny. That E5 v4. CPU delivers results quite similar to that of the LS Epyc, just the AES result is significantly lower (but still much better than on the c1v VPS. Even the spread is similar *g
But it's the VPS with the lowest amount of memory so far (1.5 GB vs 2.0 GB). Never the less plenty good enough for a backup server.
host_c
Machine: amd64, Arch.: amd64, Model: Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
OS, version: FreeBSD 14.0, Mem.: 1.991 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 vmx ssse3 fma
cx16 pdcm 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 ospke syscall nx
pdpe1gb rdtscp lm lahf_lm lzcnt
AES? Yes
Nested Virt.?Yes
HW RNG? Yes
ProcMem SC [MB/s]: avg 194.3 - min 77.5 (39.9 %), max 314.4 (161.8 %)
ProcMem MA [MB/s]: avg 567.7 - min 540.9 (95.3 %), max 589.5 (103.9 %)
ProcMem MB [MB/s]: avg 576.3 - min 545.9 (94.7 %), max 602.4 (104.5 %)
ProcMem AES [MB/s]: avg 1086.0 - min 1069.5 (98.5 %), max 1094.1 (100.7 %)
ProcMem RSA [kp/s]: avg 80.7 - min 77.5 (96.1 %), max 83.2 (103.2 %)
Yay, now we are talking! Even when ignoring the 2nd vCore (because all others only have 1). Just look at those AES and RSA numbers! Now, that's a machine to which you can effortlessly send encrypted backups within a SSL session and/or decrypt files on the fly. Besides, the (single) vCore I count is significantly faster than all others, too.
The winner in this category, again even ignoring that the host_c VPS is dual vCore, clearly is host_c, followed by Liteserver and Calin (not far behind) and finally c1v.
Comments
A somewhat sad preamble: Although I tested each and every target from my desktop, quite some of them do not work at all, or only via some paths/ASs when used from a server. My guess is that a mix of good (or not) routing and of IP "reputation" (and be it just "coming from residential or from host") is responsible.
Now, one might think "just use iperf3 or similar". But sadly that's not a real alternative for me because I want to see and to show what you can expect from your VPS and not some marketing optimized BS shitshow.
So, sorry, some targets just are not reachable, be it because a VPS usually is a host or be it that a hosting providers routes aren't really good.
Anyway, here you go, geographically ordered. Europe - Asia - Ozzyland - Americas - Africa
Liteserver
So, all of Europe solidly > 100 Mb/s and the "big" targets LON, FRA, PAR >= 500 Mb/s and quite many not much slower. Even RU MOS is > 200 Mb/s and only RU SIB isn't bad with > 50 Mb/s ... but then, the provider is in Europa (NL) so that's not a surprise.
India, Mumbay just a bit under 50 Mb/s isn't bad, China, oh well ..., Tokyo a bit over 30 Mb/s is OK I guess, and Ozzyland also over 30 Mb/s is OK too I guess.
Massachusets < 70 Mb/s and LAX < 50 Mb/s is not OK though in my books. Even Sao Paolo, BR, isn't much worse.
Kenia and South Africa in bteween 40 and 50 Mb/s isn't great but I've seen worse.
TL;DR I would expect better from a reputable provider in NL. Much better. After all, there is a world outside of Europe.
c1v
First a remark: That VPS reacts so slowly and shitty that multiple attempts to download the test results failed. So, these numbers are based on a single run at the end of 2023 ...
Not much to say. Even the fastest download speed achieved not even within IT but to DE FRA is well under 100 Mb/s. But then I got that VPS only because it was dirt cheap and only a secondary backup anyway. Plus I count it only as half a test (-> "3.5 VPS test") ...
Calin
Europe not utterly poor results, a few even halfway decent (sorry, >500 Mb/s within RO doesn't mean a lot) but I see nothing to write home about.
Asia and Ozzyland not exactly decent either, Americas the same with one exception, LAX is better than Liteservers result. Africa halfway decent.
TL;DR Sorry, Calin, your connectivity really needs to get better.
host_c
Europe OK results but the top results are lower than with Liteserver. Overall decent though.
Asia and Ozzyland results also OK'ish and both chinese targets could actually be reached with one even showing a not bad result.
The Americas not great, just like the other providers but at least no target was a complete failure. And the Africa results are even halfway decent.
TL;DR Although there are no good "peaks" and only Liteserver showed those, I find the result set reasonably balanced as in: you wont see records but neither will you see failures. Kind of well balanced.
The winner is split. Liteserver if you want a few peaks but also some failures or weak results -or- host_c if you value a decent result everywhere. The second place goes to the one you didn't chose for No 1. The third place goes to Calin, not because his connectivity is good (it isn't) but because at least it's not as crappy as c1v.
This probably is the one you care most about. After all, we're looking at storage servers here.
Keep in mind that only Calin and host_c offer separate boot and storage drives. That also happens to fit the order that I chose well.
So, here you go:
Liteserver
large storage drive
It seems reasonable to focus on reading and writing mostly 64KB or 1 MB blocks
since chances are that one doesn't have gazillions of small files on a storage server, alone for not wanting to waste space.
In that regard I'm not exalted by what I see but neither am I disappointed, except for one point: 1 MB reads are slower than 64 KB reads.
Also the very significant difference between reading 64 KB sequentially vs. randonly indicates that something isn't well balanced and configured.
What I do like though is that the spread everywhere is relatively low, meaning that users don't step on other users feet in terms of disk usage.
c1v (again, only a single run)
Well, I guess it's clear now why that snail machine isn't exactly a pleasure to use ...
Nuff said.
Calin
boot drive
It's a bit like the Liteserver va Calin network situation, but now reversed. Reading over 1 GB/s in 64 KB and 1 MB blocks? Nice!
But let's be clear, this isn't a speed daemon, and it doesn't need to be. After all a (more or less) faster small boot disk is nice to have but at the end of the day just a gadget.
Let's see how the large storage drive performs. That's what counts on that kind of VPS.
large storage drive
Hmm, I has expected boring rusty spindles, and quite likely they are rusty spindles, but with decent RAID and OS caching it seems.
Writing is slow, though. Writing out 1 MB blocks at less than 10 MB/s isn't something to write home about. But OK, it's not utter crap either.
But hell, if even c1v outperforms you, you seriously have some work to do...
host_c
boot drive
Yep, I like what I see here. 3 digit MB/s when writing (buffered) 1 MB blocks and still 2 digits in the upper range with 64 KB.
Not that I actually care a lot about the boot drive speed of a storage server, because those beasts are rarely rebooted, neither does one do lots of compiling or the like but it's still nice to see that a provider really tries to please their customers by offering something really decent.
large storage drive
Yessir, me like a lot! To achieve that kind of performance isn't hard - but to achieve it at low cost definitely is.
To keep it simple: A storage server should be able to comfortably fill a disk from a network pipe, meaning, what good is it when a storage server has a fast network connection but a slow disk that can't digest what comes in?
Or in other words: You need a well balanced system to backup to and from a storage server and at the core is a good drive.
host_c's large storage drive can keep up with a 500+ Mb/s pipe. That's what I want to see with such a VPS.
An explicit ranking isn't needed anymore, I guess.
Again, I'm a current customer of all 4 of the providers. I actually have and use those VPS, some for primary backup, some for secondary or backup-backup.
I expected that some are "kind of crappy but cheap" and considering the price for my 2 primary backup systems (one a dedi) it seemed not unreasonable to get a few of those cheap VPS which together are cheaper than any of my primary backup systems.
At the end of the day one shouldn't only look at performance but also at reliability, support (in particular how fast and how reasonable and knowledgable they react), and of course price.
With Liteserver my experience was very positive. Yes, the VPS was expensive, but as much as I'm willing to have a look at a super-cheap and possibly new provider, with my primary backup and storage servers I'm not willing to take any risk. I'm not demanding (or expecting) 5 nines, but a high level of availability and reliability - and that usually isn't cheap.
c1v was one of those wild bets. Maybe I'd experience a clusterfuck, but risk (as in $$) was low as well as the risk of losing important data (because I anyway had some high quality primary systems in place). My personal resumee isn't as bad as that of some others; and frankly, if the system performs reasonably well and is reasonably reliable, I don't care whether it's in someones cave or garage - the problem is, c1v's VPS all in all is a disppointment and the results here confirm that. I will not renew that box.
Calin is considered something in between an evil scammer and a clueless (very) young man. Based on my personal experience with him I see no basis for such accusations. But I also see no basis to recommend him or to even consider him for a real (primary) backup/storage server. With a secondary one I'm still content. Hey, I get 4.some TB of, OK, not exactly great quality for a very low price. And so far my VPS has an acceptable up-time/availability. I guess I'll keep that VPS.
Finally, host_c is the clear winner in my eyes. Not only because the performance of that VPS is similar, or in some regards even better, than my LS VPS, and also not only because his offers/products are really cheap, but also because I'm under the impression that he is bloody serious about his business and quality and a true professional (sorry, as a techie I'll always tend to relate to other techies over say, a beancounter or sales type). I had some conversations with him and I didn't feel doubts even once.
So, the two I'll keep are host_c and Calin, and the two I'll let go are c1v and Liteserver. Why?
Recommendations? OK, here you go:
TL;DR Go and get a VPS from host_c!
P.S. I not only had full permission from host_c to benchmark the hell out of my VPS but he also disabled to IO limit for my VPS. Although frankly, I didn't see a significant difference to some earlier benchmarks I had run (without permission * blush).
From all other I had not, and did not ask for, permission. Sue me if you feel like it *g
For me, @host_c, @hosthatch and @rsk are premium storage vps providers.
You should highlight this. The post is wayyyy too long, even with the results hidden.
FreeBSD!
Yes, I pretty much always use FreeBSD.
Would love to hear why you chose to use FreeBSD over Linux or other BSDs
what about Greencloud?
That's another story for another time.
I like them, generally speaking, but I only reviewed storage VPSs I happen to actually have. Never had one from Greencloud.
Please share your story here: https://lowendtalk.com/discussion/197572/let-bsd-thread#latest
Hello!
Thanks for your review, you should try our Cloud VPS Linux SSD
Thanks for feedback!Very apreciate!
--
I am pretty sure we have different definition of word 'reliable' then.
You can't even download test results text files from it, I can't imagine how you are supposed to get your actual data off.
regards
Disregarded
regardless
Miss the great storage VPS from @labze
Price/quality it's unattainable
C1vhosting is the worst server I've ever used.
@jsg
They say that a picture represents 1000 words, so here's ours:
I think a storage deal represent more than 1000 words, do you think so?
Can’t sell what you don’t have….unless you’re evoshosting
mom when I grow up I want to be like @host_c
I wish to add the following, regarding IO limits on our Storage VPS products.
Promo 5 and 10 TB Services have:
4000 IO and a max of 125 MB/sec.
This translates to ~8-12 MB @ 4K and 125-135 MB/SEC @ above 1MB
List priced products on our website have the IO limits stated on the product page. The more expensive the product, the more you get.
https://www.host-c.com/store/vps-storage
For example the 20 TB Nebula Storage VPS has the following:
9000 IO and a max of 300 MB/sec
This Translates to ~15-25 MB @ 4K and 300 MB/sec @ above 1MB
As we work with hardware RAID, RAID6 and RAID60, which has its physical limits regarding IO/Write/Read, this was necessary to ensure smooth operation of the VPS for all customers.
Even so, we don't think this will be an issue, as both 4K and 125MB/sec can easily saturate the 1G bandwidth allocated to the services. For fast boot times, we provide NVMe storage.
Have a wonderful Weekend, HOST-C
GB5 score for c1v VPS
When it comes to low benchmark scores, @c1vhosting is unbeatable.
Not to mention, just two choices under OS install templates.
I have a "please, pretty please" to the usual gang: Please do not abuse this thread for the usual anti-Calin games!
Btw it's a lot of work to do such a benchmark and comparison. I've invested lots of patience (well, storage servers tend to not be the fastest ...), keep an eye on the little herd, etc. All in all I spent the better part of multiple days and hours upon hours of work.
Thank you
Reliability != performance. And this was a comparison and related to LET, where @c1vhosting is "known" to be a shitshow and questionable. My "quite reliable" boils down to my experience and the fact that that box wasn't operating then down then operating again for a while etc, as one might think after having read LET. Nope, after an initial stutter phase it was running "quite reliably" for a dirt cheap system from an ill-reputed provider. Current uptime e.g. is over a month and before that it was a couple of months. For a secondary and dirt cheap box that's something one can live with.
Also, kindly note that I clearly said that I won't renew that box. Putting it brutally I get way, way, better service, reliability, and performance out for (relatively) less money from @Calin who, on top of it, seems to improve his services albeit in small steps.
EDIT:
@dev_vps -
When restock packet 5 TB sir..?
Getting text files out of server != performance.