[fChk]

I was planning to do a meta-review of Virmach VPS-9 offerings -especially CPU wise- (a review of reviews), thus when I was going through the available threads on the subject, it appeared that @ikk157 was the only one who did run a Unixbench test :-)

Not only that, he did it in double-test mode!.. Kudos! and +1Rep for your courage :-)

Did you know that you most probably broke Virmarsh's ACCEPTABLE USAGE POLICY while doing it ? Both the high CPU and high load clauses:

High CPU: Customer’s Service cannot burst to 95-100% usage for more than five (5) minutes and cannot average higher than 50% usage within any two (2) hour period. Packages advertised to include dedicated CPU, Services with the high CPU option, and any customized Service plans that include high CPU option may burst to 100% at all times.

High Load: Customer’s Service cannot have a 15-minute load average higher than the number of full logical cores assigned and cannot have a 1-day load average higher than 70% of the number of full logical cores assigned.

I knew that for a fact because I was planning to run both the simple and double tests my self on the Phoenix-based VPS, but ended up doing the simple test only as it was the only one who can finish within Virmarch's rules, ie a load of 6.18 (1') 3.65 (5') 1.88 (15') and a CPU usage of 56% for most of the test duration and ending at 95.83% .

For the meta-review thing I did cancel it as there is not enough data in the published reviews for KVM-VPS-9 while others are still missing!

I also wanted to go with Geekbench to perform a benchmark. However, Geekbench isn’t free so that’s not an option. So I went on with ....

You can still use it in its trial mode-like everybody I would guess and it's more than enough. By the way, it doesn't abuse the CPU the way UnixBench does.

I haven’t tried VPS9 back when it was on OVZ. So I can’t make the comparison between that and KVM. However, I don’t like the nature of OVZ, I absolutely HATE how the resources are shared and how your “noisy neighbors” could affect the performance of your VPS, which also resulted into many companies utilizing strict resource policies. However, with KVM, you get your own dedicated resources. Which solves all of the issues that OVZ has.

The “noisy neighbours” effect can still be felt within KVM's VMs; it shows up CPU-wise in the steal time statistic when querying your CPU usage.

When the steal time gets above 10%, your vCPU's performance drop as it's struggling to get its fair share of the real CPU from the hypervisor.

Container's VM (like OVZ) vCPU performance suffers most from the IOWait time statistic which can jump to levels above 30% making everything crawl to a halt.

It's all relative, and the main culprit in both cases is usually the overselling of servers' resources.