All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
How could bandwagonhost get such a good result in IO test?
hiphiphip0
Member
http://www.hostingwizard.net/
Order by iops and the tops are all BandWagonHost and Xvmlabs.
ioping -c 10 .
4 KiB from . (simfs /dev/simfs): request=1 time=11 us
4 KiB from . (simfs /dev/simfs): request=2 time=4 us
4 KiB from . (simfs /dev/simfs): request=3 time=113 us
4 KiB from . (simfs /dev/simfs): request=4 time=3 us
4 KiB from . (simfs /dev/simfs): request=5 time=3 us
4 KiB from . (simfs /dev/simfs): request=6 time=2 us
4 KiB from . (simfs /dev/simfs): request=7 time=93 us
4 KiB from . (simfs /dev/simfs): request=8 time=3 us
4 KiB from . (simfs /dev/simfs): request=9 time=3 us
4 KiB from . (simfs /dev/simfs): request=10 time=2 us
--- . (simfs /dev/simfs) ioping statistics ---
10 requests completed in 9.08 s, 42.2 k iops, 164.8 MiB/s
min/avg/max/mdev = 2 us / 23 us / 113 us / 39 us
I have never seen "2 us" on other servers. Is there something like linuxatemyram?
Comments
Getting high IOPS is a combination of the following:
You can get high IOPS by using high quality SSDs in a RAID 10, a RAID card which is SSD aware, large-local RAID cache, write-back enabled.
Then you can also tweak the host OS and guest for additional disk I/O performance.
Hopefully, this helps. It can grow from here and get even better when using newer and faster types of disks and supported protocols.
There is another possibility which might overshadow everything else:
Very true.
If you have some bwh vps you know that performance don't decline, I have three vps since two years ago and they are always e same.
That is good to hear.
Just want to be sure it is communicated that disk I/O is a finite resource as is RAM, disk space, etc. So, if you have a few folks running applications which are consuming a large amount of disk I/O, the other folks on the node will suffer.
Sounds like BWH does a good job of balancing nodes, I/O limiting, or something similar. Which is good.
Interesting points, but you miss the major step in tech - NVMe and PCIe AHCI SSDs delivering 3000MB/s in each direction with 150k IOPS+ (eg. Samsung SM951, SM961).
High quality is pointless, Samsung SSDs cost more - yes, but they deliver less IOPS and MB/s per dollar than by now comparable Sandisk and especially the ultra cheap Intel 600 NVMe drives with 3D NAND. More SSDs and a larger array ALWAYS makes more sense than less higher quality SSDs.