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
Friendly advice: set virtio at provisioning, if it aint working, let the customer call, i think it is a better way now since very few OSes dont support virtio out of the box still.
so now is around 150MB/s
is it a normal speed for harddisk KVM vps?
yes, most things will work. In the past, 50+ was enough, now i think 100 should do for almost anything but some kinds of DBs.
But keep in mind this only tells a part of the story. IMO DD is relevant only in the low side.
is unixbench a better option to test VPS?
unixbench mostly tests the CPU, just test it with your app, if it is behaving normally, it passed, if not, go and search for clues.
But you will be reasonably sure it is the disk when you see high iowait.
Neither.
For my case I usually find IOPS numbers more important than the dd results.
Unless you're in the business of streaming porn to a single user, the sustained 1GB dd test is useless.
So as we have witnessed here, "that useless dd test" has just helped quickly detect and solve a real performance issue with a VPS configuration.
That makes no sense.... if you purchase a brand new pc with a highend processor and what not and you don't know shit about computers and they rip you off and put an old intel pentium in there and you just browse facebook and don't really experience the pc as being "slow" would you accept that?
Like hell you would, you get what you pay for or money back, no one should expect nothing less than that.
Most people will (not saying it's good, but that's just how it is)
dd is LET's religion. It shouldn't be disturbed.
lol that got out of control fast
So apparently the car salesman lied. It's just a printing error on the brochure I guess.
You're part of it.
It seems like every host who said that sequential I/O throughput speed result test is not best measure of performance (which I agree with) didn't forget to add that THEY infact "sell" service with higher dd results to distance themself from results they defend in thread ;-)
Dear @rm_ - are you consider this sugar-coating http://lowendtalk.com/discussion/comment/446795/#Comment_446795
LEB hosts and your marketing should be blamed for "dd religion" not clients who run those tests, impressed with high numbers without real understanding of them.
Actually, I didn't just say something and not test. I've spun up a KVM VM with write throughput limited to 10MB/sec and left the IOPS unlimited. The yum -y update took 6 minutes to complete on it. So yeah, it seems 10MB/sec is really bad. Usable, but bad.
No, go blame serverbear. That's who started this. Providers just saw people use that test to measure if a VPS is good or bad and that's how it started.
but to be honest, this VPS is not as fast as the VPS I use before, and I am not satisfied. I will check with it. This VPS is stable but a bit slow. I give up my previous VPS because it randomly down. Now I don't have this problem ... just slow.
I just came to say LOL
What is slow exactly?
the virtual desktop is laggy, this is not happened in my old VPS
That's bollocks mate, if you live by the sword, you die by it. If you don't want customers complaining about the results of DD tests, don't advertise your services with DD tests.
That comment has fucked me off so much, if I ever manage to work out if I have any services with your numerous brands, I'm cancelling them.
Business runners check what their target customers want and act accordingly. If they want more dd porn, you give it to them. That's it. If it's too harsh for you, I can't do anything.
And they shouldn't complain about the results of the DD tests. They should complain if they are not getting the service they expected.
That's not correct. He made and popularized whole benchmark script when little dd test "line" was already established marketing gimmick at LET. Serverbear site isn't even so old and you can maybe blame him for jumping on bandwagon with complete benchmarking test, but not for simple "dd test" religion which started before that by hosts advertisement as something great, not just born with serverbear script some year and half ago.
What's the use of your message which says in this thread something like: "sequential I/O throughput speed result test is not best measure of performance, but don't worry, I offer only high sequential I/O throughput speed result with my service" if not supporting dd religion as salesman even if you know as tech person that it's nonsense. And the other host in this thread did the same...
The first sentence makes sense, what comes after is gibberish. Don't keep digging.
I agree, and this is how any successful business will operate: adapt to your customers. The Pareto Principle essentially states that 80% of your sales will come from 20% of your clients. It's much cheaper to keep a customer than to find a new one.
If DD shows an abhorrent result on a properly configured VPS, it may be a sign that performance will be problematic in the future. A DD test alone, though, remains relatively useless in assessing overall performance in real-world environments.
I honestly, started to notice these many dd threads and posts after serverbear became popular.
I said that to make it clear that I'm not defending that a low end VPS should get 10MB/sec of dd result. If you check my first post of this thread you'll see that I asked, yes it's 10MB/sec but what feels slow to the OP. What did he actually do to see if the VPS is slow to him.
I said that to point out 10MB/sec dd is not what I target for our customers.
Running a dd test and then opening a thread about how slow it is is what I argued about. If you check his actual answer to my question: "the virtual desktop is laggy, this is not happened in my old VPS"
The dd is good now, it's 154MB/sec, yet the guy still says VPS is slow. See? The problem was not even the sequential speed at the first place. It's probably the CPU load on the node that makes his VPS slow if not network latency.
VPS's IO speed cannot be measured purely on dd, and almost everyone on this thread already knows this. And now after I ran tests, I agree that that kind of low dd is actually a sign of slowness. Previously, I thought that 10MB/sec is not the best performance a VPS can get but it should be still pretty usable for most of the things. Now, I'm suspicious about it. I still need to run website tests and stuff but yeah, it's not exactly what I thought.
Well, if you don't know, one of the few reasons why providers use hardware controllers on their SSD arrays is to get the dd and ioping results higher. 8 x Intel 530 240GB SSD does 450MB/sec dd on mdadm (software) RAID10, and 1.5GB/sec on LSI 9271 (hardware) RAID10. Yet, in real world, there is no difference in performance.
I understand all that and I completely agree with. I said several times in last few years at LET that sequential I/O throughput speed result test is not best measure of performance and I am quite pleased using some hosts which don't give more than 30MB/sec with those tests. But I responded in this thread to "dd religion" comment which you're part of.
You state something as a tech person and then negate it as a sale person. That's approach where this "dd test results religion" atleast partially comes from.
(clarification: I am not aiming you personally here, and I can easily use some other host example from this same thread)
This is like the RAM and price race going on on LET. Both providers and customers are part of it. It's like a wicked loop now. You cannot win.
You post an offer for $5, 2 days later some other guy posts the same specs for $4. Customer goes asks for $3. Some guy offers it for $3. Now another offer gets posted as $3. And another guy posts it for $2 and so on.
Some decrease the prices, some increase the specs.
A thread gets open, people compare their VPS's dd results. One sees the other VPS company gives more dd. Then starts to ask for that kind of dd from his provider or switch to the better dd giving provider. And provider sees this and starts to advertise high dd results. And it goes to the loop again.
I don't know the solution for this.
What does that have to do with my issue?