IO Performance of VPS "cloud servers".

Hi ASS_SAZiN,

Unfortunately this is often the case as most providers spin up far too many VPS's clients on a single hypervisor node that often only has a 4 disk SATA RAID10 setup for storage. At times some VM's will use large amounts of disk IOPs lowering the overall speed for everyone on that node.

You could request your provider move your VPS onto another node and hope the disk IO improves.
Click to expand...

That is what I figured,
and the main problem is that they do not publish these contention ratios, so you won't know if you are on a heavily over utilized node and because it was not stated anywhere they can just keep on adding people to a node.

I see that http://isoho.st/ do publish some ratio's which I think should become common place among the SA providers.

On that note most providers will tell you something along the lines of 1 x vCPU, that could mean a million things, I know I would obviously have a virtually allocated processing unit, but what is the speed of said unit, how much performance can I get from it and how many people will have to fight for it at once.

I must also say that my afrihost vps is running much "smoother" as of late, so maybe they optimized something. but like i mentioned; its not like I host facebook.
 
Last edited:
This is from my Web Africa VPS:
root@mail:/turds# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1,1 GB) copied, 8,02494 s, 134 MB/s

What I do know is that they are using a proper SAN for their storage. I know this because when they had that big outage a while back it was due to a LUN going down.
 
There is a big difference in practical performance ( and redundancy ) between a traditional VPS and a proper virtual server. Traditional VPS is typically a shared option where the resources of a physical machine are shared amongst all the virtual instances on that machine without consideration for resource control. A proper virtual server uses resource control to pre-allocate and guarantee resources to an instance, and to make sure that the usage of one virtual instance will not impact the usage of another. A proper virtual system also supports resource sharing or borrowing while maintaining performance guarantees. In SA terms the pricing differences are quite big with the traditional VPS starting at a few rand to R150, and a proper virtual server starting at the R400 mark to about R1000. And so using pricing you can normally guess which is which.

Performance can also vary greatly from one provider to another. Whether they are using cached clustered SANs, SAN-backed cluster file-systems or simple software RAID solutions on commodity hardware can define an experience as being great or bad. Entry-level solutions will generate around 100-250MB/s of throughput and 5000 IOPs, while an enterprise-level solution will generate up to 2GB/s of throughput and 50000+ IOPs. How many virtual instances are loaded per physical box can also make a big difference - ISPs are know to sweat their hardware assets to the max.

While there is nothing necessarily wrong with a std VPS, you have to understand that your instance may be impacted by usage on other instances. If you want guaranteed performance, you need to go the proper virtual server route ( ala Xen, KVM, VMware, etc. ). It may cost more but you get what you pay for.

The next rung up the ladder are dedicated servers which means that the entire physical hardware is yours. Be careful here as some of the entry-level options in this category for local hosting providers have software RAID.
 
lol so something is broken...

dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 770.911 s, 1.4 MB/s
 
rpedrica said:
There is a big difference in practical performance ( and redundancy ) between a traditional VPS and a proper virtual server.
Click to expand...
A vps is a vps. Are you talking about the difference between technologies like OpenVZ and Xen? OpenVZ is shared resources that you can oversell and Xen is shared resources you can allocate. Im trying to understand the relevance of what you are saying to this discussion.

The next step up is not necessarily hardware. It would depend on their needs and budget. Hardware based servers fail all the time and can cause great headache when they go down. Ask me I have too many to count. A cloud based VPS can scale to any size and does not have multiple single points of failure like your own hardware would have. So for many people that makes more sense than dedicated hardware.
 
Last edited:
ghoti said:
A vps is a vps. Are you talking about the difference between technologies like OpenVZ and Xen? OpenVZ is shared resources that you can oversell and Xen is shared resources you can allocate. Im trying to understand the relevance of what you are saying to this discussion.

The next step up is not necessarily hardware. It would depend on their needs and budget. Hardware based servers fail all the time and can cause great headache when they go down. Ask me I have too many to count. A cloud based VPS can scale to any size and does not have multiple single points of failure like your own hardware would have. So for many people that makes more sense than dedicated hardware.
Click to expand...

He probably meant "cloud server" instead of "virtual server" since cloud servers are independent of hardware, whereas VPS is one server somewhere
 
hmmmm, Macbook Pro retina (the one with SSD's) running a Ubuntu Server in a Hyper-V vm:

dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 1,02241 s, 1,1 GB/s

dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 0,86021 s, 1,2 GB/s

dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 1,31148 s, 819 MB/s

that doesn't look right? Am I doing it wrong?
 
Ipwn 4 said:
hmmmm, Macbook Pro retina (the one with SSD's) running a Ubuntu Server in a Hyper-V vm:

dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 1,02241 s, 1,1 GB/s

dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 0,86021 s, 1,2 GB/s

dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 1,31148 s, 819 MB/s

that doesn't look right? Am I doing it wrong?
Click to expand...

Cool, and how many clients do you have that uses your laptop as their server?
 
AcidRaZor said:
Cool, and how many clients do you have that uses your laptop as their server?
Click to expand...

none, I just don't believe the results...
it maintains 1GB+ even while conducting the tests while copying data to the machine to place some more I/O load.
 
Ipwn 4 said:
none, I just don't believe the results...
it maintains 1GB+ even while conducting the tests while copying data to the machine to place some more I/O load.
Click to expand...

What don't you believe? Test this under a real VPS loaded environment and you'll see the same sort of stats. note that most are normal drives in RAID, not SSD's since capacity is still rather expensive
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X