New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Fake disk speeds?
Hey guys,
Picked up a $1.50 VPS and ran nench, but the drive speeds seem... really fast and I'm not sure if they're real.
root@test2:~# ./nench.sh ------------------------------------------------- nench.sh v2018.04.14 -- https://git.io/nench.sh benchmark timestamp: 2019-01-28 00:12:09 UTC ------------------------------------------------- Processor: Intel(R) Xeon(R) CPU E3-1240 v3 @ 3.40GHz CPU cores: 1 Frequency: 3399.684 MHz RAM: 1.5G Swap: - Kernel: Linux 2.6.32-042stab134.8 x86_64 Disks: Filesystem Type Size Inodes /dev/simfs simfs 20G 10M CPU: SHA256-hashing 500 MB 2.713 seconds CPU: bzip2-compressing 500 MB 4.650 seconds CPU: AES-encrypting 500 MB 1.345 seconds ioping: seek rate min/avg/max/mdev = 981 ns / 133.0 us / 4.53 ms / 164.1 us ioping: sequential read speed generated 7.49 k requests in 5.00 s, 1.83 GiB, 1.50 k iops, 374.2 MiB/s dd: sequential write speed 1st run: 1239.78 MiB/s 2nd run: 1335.14 MiB/s 3rd run: 1049.04 MiB/s average: 1207.99 MiB/s IPv4 speedtests your IPv4: xxxx Cachefly CDN: 11.81 MiB/s Leaseweb (NL): 10.27 MiB/s Softlayer DAL (US): 11.36 MiB/s Online.net (FR): 9.89 MiB/s OVH BHS (CA): 11.47 MiB/s No IPv6 connectivity detected -------------------------------------------------
What do you guys think?
Seems like a steal for the price.
Comments
Use dd to read/write 10 gigs.
Just done, speeds go to about half of nench:
When doing a 1gb file, it is still 1.2gb/s:
Not bad.
Where did you get that?
It's from one of the shit hosts that use CC, I was very surprised with the performance of the disk...
#disks
Wait a few weeks until the server is running about 200% capacity then try. I wish you luck.
Pls PM where to get it
It simply means you are on an empty node.
Name the host and you can expect your honeymoon to turn into borderline nightmare.
Hence why I have low intentions of sharing the name
No worry. They will create more offer to get as many suckers as they could before disappearing.
HiFormance last offer posted on LEB 1 month then they went down to the toilet.
Damn. Must be a bunch of SSDs running on RAID 0, or it's some sort of RAM disk...
Or NVME?
You're hitting the cache, that's normal.
10GB file doesn't fit in cache => speeds go down.
Could be either Linux caching, hardware RAID cache or ZFS with compression (since a bunch of zeroes compresses really well - but you would see higher speeds in that case, so no).
They will get more sales.
...You see it?
On an uber budget host? Could be but I doubt it.
It is simfs so that is just you hitting the raid card cache rendering 99% of dd <=1GB tests pointless since 2008
I found 10GB to be somewhat useful.
Yeah, it proved the 1GB was just hitting the raid cache, the 10GB hit the raid cache initially giving it a small 10% ish boost and then the result was around 500MB/s which you can hit with 4 x enterprise sas in raid 10 or 2 x desktop SSD's in R1
I wonder: Do people who fake disk speeds also fake orgasms? And - Can one cache several orgasms to produce one big eternal orgasm?
Good point - caching seems most likely.
Does I/O speed matter?
If you can't write to disk fast enough, then you'll overflow your buffer. In that case, I hope you have some tissues or a raincoat handy.
Faster, Harder, Scooter. So the answer is YES, most probably...
Then I recommend a Redundant Array of Independent Dicks to increase uptime.
If ioping is too low, chances are you're hitting a fake one in cache
latency matters, it shouldn't be 0 and shouldn't be too high to be credible
I'd suggest to run some flexible input/output tester, with different block sizes and at different hours, to check if the tested subject is faking or not; also you could learn which size fits better in that liaison
EDIT2:
Yes.
Size matters.
"One (cache) size fits all" probably won't happen.
EDIT2: