Packet
The following test was executed on
Packet
Bare Metal Server
Type 1
: 8vCPU(3.4 Ghz) + 16GB(DDR3) + SSDs(2x120GB)
Type 3
: 32vCPU(2.6 Ghz) + 128GB(DDR4) + SSDs(2x120GB) + NVMe(2x800GB)To see test result on IBM SoftLayer Bare Metal Server
1. Time
1.1 Pod startup time
Run a new Pod only takes 376
millisecond (ms) on Type 1
Run a new Pod takes 519
millisecond (ms) on Type 3
The following tests are for "Hyper for KVM" with qboot
1.2 Pod replace time
Replace a running Pod with a new one, only takes 150
ms on Type 1
Replace a running Pod with a new one, takes 174
ms on Type 3
2. Density Test
In the same condition, Hyper can run more Pods than traditional VMs, and there is more available memory for user's applications.
Test method: Create VM/pod ceaselessly, until "Cannot allocate memory" occurs or reach maximum of tap device.
Allocation of resources :
1
vCPU,512
MB MemoryTest Docker image
ubuntu:14.04 - https://registry.hub.docker.com/_/ubuntu/
2.1 Test result on Type 1
Type 1
1) Max traditional VM(KVM) number : 105
QEMU process memory usage(MB): (105 QEMU processes)
memory usage in VM (MB): (105 running VMs)
2) Max Pod(Hyper) number: 252
QEMU process memory usage(MB): (252 QEMU processes)
memory usage in Pod(MB): (252 running Hyper Pods)
2.2 Test result on Type 3
Type 3
1) Max traditional VM(KVM) number : 627
QEMU process memory usage(MB): ( 627 QEMU process )
memory usage in VM (MB): (627 running VMs)
2) Max Pod(Hyper) number on Type 3
: 1023
QEMU process memory usage(MB): ( 1023 QEMU process )
memory usage in container(MB): ( 1023 running Hyper Pods )
3. Memory Utilization
3.1 Minimum startup memory
The minimum startup memory is 28
(MB)
3.2 QEMU process memory usage
When starting a Pod with minimum startup memory, QEMU process uses 50
MB of the Host OS' physical memory.
3.3 Memory usage in Pod
When starting a Pod with minimum startup memory, there will be 6
MB of available memory in a running Pod. Hyper Kernel only takes 13
MB of memory.
4. CPU Performance
Allocation of resources: 1
vCPU, 2048
MB of Memory
The following table is the result of dhrystone
CPU performance test.
In the Target column,
Host
means Host OS,Docker
means Docker container,Hyper
means Hyper Pod
The following table is the result of whetstone
CPU performance test.
5. Memory Performance
Allocation of resources: 1
vCPU, 4096
MB of Memory
The following table is the result of stream memory performance test on Type 1
The following table is the result of stream memory performance test on Type 3
6. Testing environment
Bare Metal Server configuration
Detailed configuration: https://www.packet.net/pricing/
Last updated