Close

Performance Evaluation: Virtual Machine vs Physical Box

Share tips and tricks related to Sighthound Video or your full security setup.

Moderator: Staff

no avatar
Jmtyra
 
Posts: 16
Joined: Tue Jan 19, 2016 1:05 pm

Performance Evaluation: Virtual Machine vs Physical Box

by Jmtyra » Fri Aug 03, 2018 2:44 pm

Sharing this experience in case anyone else is also considering a virtual machine versus a physical box.

The SH installation was recently migrated from a virtual machine (8GB RAM, 6 cores, dedicated NAS) to a physical box. WOW! What a difference! The mobile app live view loads very quickly; 1-2 seconds on wifi including app launch time. If the app is already launched, it's almost instant. Over a cellular connection it's roughly 3 seconds for the first camera to 8 seconds for the last camera. Individual cameras load quickly and clips are very fast too, roughly ~5 seconds or so depending on the camera and clip. I'm very happy with the performance.

This setup has 10 cameras at 2MP. The physical box is a quad-core Xeon with 32GB ECC RAM with an SSD for the OS and a single 4TB 7k HDD for the video. The 'tmp' video folder has been relocated to the HDD (see this thread) so that all video I/O is sent to the single HDD. The ESXi server running the VM was using a RAID 10 NAS for storage which could easily push more I/O than the single HDD and yet the physical box still greatly outperformed the virtual machine.

I hope this is helpful for others. :)

no avatar
Edith12
 
Posts: 1
Joined: Fri Sep 14, 2018 1:36 am
Location: United States of America

Re: Performance Evaluation: Virtual Machine vs Physical Box

by Edith12 » Fri Sep 14, 2018 1:49 am

This information is helpful.

User avatar
Semper Vaporo
 
Posts: 293
Joined: Sat Oct 09, 2010 11:51 am

Re: Performance Evaluation: Virtual Machine vs Physical Box

by Semper Vaporo » Fri Sep 14, 2018 8:54 am

Remember, a Virtual Box is a REAL computer running a program that is pretending to be another computer, that might be a computer of a different type or might be the self same computer that the REAL computer is... so the REAL computer is busy converting all the Input and Output from the REAL world to the pretend world of the Virtual Box. This includes data coming from the video cameras as well as data being read from and written to the hard drive, the mouse, keyboard, the display screen and anything else connected to the computer that the pretend computer is using. That can be a tremendous amount of work that has to be accomplished in the REAL world so the pretend computer can do what it is doing.

A "Virtual box" is like you driving a car by telling someone else how to drive the car. You are the pretend driver and you have to wait for the real driver to tell you what it sees and then you tell the real driver how fast to go, when to brake and when to turn the wheel... the real driver does nothing until you tell it what to do, and it won't do it, if it decides that the request is not a good idea.

The advantage of a "Virtual Box" is that the real computer can detect if the pretend computer is about to do something that might "wreck the car" and decide to not do it, in order to preserve itself (the car). A wonderful thing if what the pretend computer is doing might cause the loss of data or even destruction of the hardware, but downright silly if what the pretend computer is doing needs the speed of direct control of the hardware and not have to wait for the real computer to decide if what is going on is a good thing.

"You want me to park on the railroad tracks? No way! You must be malware! I will stop processing your instructions now!"
Semper Vaporo,

no avatar
sspeed
 
Posts: 1
Joined: Mon Oct 29, 2018 5:44 pm

Re: Performance Evaluation: Virtual Machine vs Physical Box

by sspeed » Mon Oct 29, 2018 5:49 pm

I appreciate your experience, but you are also comparing apples to oranges.

Was the CPU the same on the VM host as this physical box, or was it an entirely different slower CPU?

How many vCPUs did you have assigned to the VM vs how many are on the physical box?

What class RAM was in the VM host compared to the RAM in this new physical box?

Have you done drive performance tests with the NAS and with the new machine? Local storage will always be faster than network attached storage unless you have a crappy hard drive and an enterprise class SAN.

I have no doubt you saw an increase in performance, but the VM also needs to be set up correctly on an appropriate host.

no avatar
piter85
 
Posts: 14
Joined: Fri Sep 06, 2019 9:39 pm

Re: Performance Evaluation: Virtual Machine vs Physical Box

by piter85 » Fri Sep 06, 2019 10:40 pm

The selection of a contact or non contact all kinds of sensor depends on your application. Although non contact technology seems advantages it is not possible to use non contact Nagano keiki Pressure sensor for each and every application. Some non contact abs speed sensor like laser and ultrasonic sensors may be dangerous. The advantage of non contact sensors are that there is no mechanical contact so there is no friction and the mazda 6 suction control valve does not wear out. Hence operating cycles are more. Non contact suction control valve are more costly compared to contact Speed Sensor. Operating temperature and frequency response of a contact Pressure Sensor is high. Some non contact sensors like ultrasonic sensors may have a blind range.


Return to General Discussion