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.
Comments
He's a busy man, tickets mark something to do so he can look at it later, I wouldn't expect him to look at it RIGHT NOW, he's got other stuff to handle.
I've had many RamNode's, they've all been wonderful! The network, the I/O and the support are great!
And he does not want to use LET as his support desk, neither does he know who iKeyZ is in his client list.
Currently a ramnode customer, more than satisfied.
I want to compare his specific VPSs to the host node, etc. I'm not sure why you have a problem with a simple request...
Nice i/o numbers - but my DotVPS box is more than adequate for its purpose!
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 1.90909 s, 562 MB/s
Edited ironhide's messages and removed affiliate links
I think it is a given, people cant compete with Ramnode in terms of sequential write IO. Congratz @Nick_A
Out of interest, what is the random read speed as that is generally the real life performance indicator? may take one of the KVM's as a DB box.
@AnthonySmith If I zip 1GB web files data, can it test random io?
I just picked up a Ramnode VPS today to use as a VPN and I have to say I am happy so far. The network speeds seem good and the IO is great.
For random write speed try:
Then read the random data
Probably better ways but that will give an accurate enough result of real world use, I will probably just get a KVM anyway I need something off my own infrastructure for a project.
IOPing was crazy good on my RamNodes when I checked a few months ago.
But really, RamNode's not just a performance host. It's the sum of everything @Nick_A puts into his company that makes it so great to be a customer. If I was a competitor, I'd be benchmarking (business-wise) the hell out of RamNode.
RamNode IO in their NL location.
root@db-host: dd if=/dev/urandom of=test bs=64k count=10000; sync;
10000+0 records in
10000+0 records out
655360000 bytes (655 MB) copied, 54.2678 s, 12.1 MB/s
I've had a number of VPSes with Ramnode for a while, absolutely zero issues with I/O. It's unicorn level stuff.
All I can say is thanks again guys! We are proud of the product we put out and we will continue striving to make it better.
Any plans for SG/HK location?
Not at this time.
I will be interested if they launch the Cloud Based VM.
A little jealous of the above
Results on a SEACKVME5-1
Guess I got a busy node.
That was an amendment of "cant compete" part.
It depends if that is useful or gives value to customers, so the "wish to compete" is more like it.
IMO anything above 300 should do well enough for most applications.
Sure, for a very busy database and low memory (cant cache in memory) a IOPS beast is needed, but, otherwise, I dont see a reason why so much sequential writing is needed apart from editing bluray movies on a beast computer at home.
I hate this test, it increases costs and prices for no really useful purpose. Sure, if you get xx MBs, then that is a SIGN of a possible problem, but 100 MB/s are already enough for almost everything, 300 should really be enough for all usual apps on a VPS. Besides, with 2 GB RAM cache in the controller, it does not really matter what disks are there, the "test" will still show 1+ GB, perhaps even 2+.
Mao - It's not just the sequential write speed of RamNode, its the consistency of service overall that earns my respect. It's the same kind of respect that Prometeus gets from the community. As a provider you have it or you don't. I think you would agree that disk speed alone does not get this kind of positive response, from THIS community.
Ramnode was faster.
Sequential io has NOTHING to do with real world performance though so you can take your xx MB is a sign of a problem and throw it in the bin.
Most people only use less than 1 MB/s on avg in reality.
@AnthonySmith - what about swap? Since most openVZ VPSes I have are very swapy wouldn't I/O matter in respect to swap? Would this not affect the performance of the VPS overall, up to a point?
EDIT: At 30MB/s it might not matter, but at 1MB/s could the same be said?
The result of a cached sequential IO test has almost zero bearing performance of your swappiness, and with OpenVZ you are not truly swapping to disk anyway.
Sure 1MB/s could indicate an issue, but the absolute truth is most people dont even need that.
here is a snapshot sorted by top writers on a p/second basis on a busy node:
45 domains, 45 active, 5 running, 40 sleeping, 0 paused, 0 inactive D:0 O:0 X:0 CPU: 35.4% Mem: 112927 MB (111624 MB by guests) ID S RDBY WRBY RDRQ WRRQ DOMAIN DEVICE 562 S 0 20K 0 5 vm722 xvda1 1055 S 0 24K 0 2 vm933 sda1 1165 S 0 24K 0 2 vm967 sda1 4 S 0 0 0 0 vm638 sda1 1115 S 0 0 0 0 vm641 sda1 793 S 0 0 0 0 vm642 sda1 556 S 0 0 0 0 vm648 sda1 57 S 0 0 0 0 vm650 sda1 1072 R 0 0 0 0 vm660 sda1 784 S 0 0 0 0 vm667 sda1 1005 S 0 0 0 0 vm670 sda1 991 R 0 0 0 0 vm672 sda1 833 S 0 0 0 0 vm676 sda1 1077 R 0 0 0 0 vm683 sda1 209 S 0 0 0 0 vm684 sda1 1012 S 0 0 0 0 vm689 sda1 758 S 0 0 0 0 vm690 sda1 293 S 0 0 0 0 vm692 sda1 516 S 0 0 0 0 vm726 sda1 513 S 0 0 0 0 vm727 sda1 514 S 0 0 0 0 vm728 sda1 1166 S 0 0 0 0 vm729 sda1 1065 R 0 0 0 0 vm752 sda1 1026 S 0 0 0 0 vm799 sda1 1023 S 0 0 0 vm807 sda1 1113 S 0 0 0 0 vm905 sda1 1008 S 0 0 0 0 vm912 sda1 1112 S 0 0 0 0 vm931 sda1 1114 S 0 0 0 0 vm958 sda1 1125 S 0 0 0 0 vm963 sda1 1173 S 0 0 0 0 vm969 sda1 4 S 0 0 0 0 vm638 sda2 1115 S 0 0 0 0 vm641 sda2 793 S 0 0 0 0 vm642 sda2 556 S 0 0 0 0 vm648 sda2 57 S 0 0 0 0 vm650 sda2 1072 R 0 0 0 0 vm660 sda2 784 S 0 0 0 0 vm667 sda2 1005 S 0 0 0 0 vm670 sda2 991 R 0 0 0 0 vm672 sda2 833 S 0 0 0 0 vm676 sda2 1077 R 0 0 0 0 vm683 sda2 209 S 0 0 0 0 vm684 sda2 1012 S 0 0 0 0 vm689 sda2 758 S 0 0 0 0 vm690 sda2 293 S 0 0 0 0 vm692 sda2 516 S 0 0 0 0 vm726 sda2 513 S 0 0 0 0 vm727 sda2I am saying this from experience, what kills performance more than ANYTHING on a VPS node is volume of I/O R+W requests p/second, you can have someone with a terrible bloated wordpress install on a VPS that is perhaps only hitting a constant 5 - 10 MB/s while actually generation over 6000 disk I/O R+W requests p/second and that is what kills performance for everyone else.
@AnthonySmith - Thank you for taking the time to answer my question in such an detailed way. Very informative.
so far I've purchased 5 KVM with their coupon in two days. very impressive performance and fast & helpful customer support. @Nick_A good job.
I hate testing too, but sometimes have to do some tests to advise people to buy. It's like I told my friends who hadn't tried prometeus, look! new offers from prometeus, oh look, SSD with 4 cores. After showed them the bench and I/O, they bought.
That number is actually closer to what I'd expect on SSD-cached, so I think the others are just doing it at the right time of day
We shoot for around 400-500MB/s for SSD-cached.
May I ask what raid and how many disks you use behind the SSD-cache?
Ask Erawan, we had 1.9, nobody beat that yet. Of course, that is irrelevant, no disk will do that in reality, it is only the cache being hit.
Yes and no, if it is way too low something is wrong on the node. I would expect decent constant performance from something over 100, 50 might work, of course, but I would take it as a sign something is/could be (in the future) wrong.
Frankly, most processes hitting 10 MB/s constantly are KVMs using swap as ram. It is rare that some syslog goes nuts at similar speeds, even less frequent for other processes. We had someone doing 200 MB/s on a KVM on SSD array sustained, that is almost certainly swap.