RAM Disks performance tests for a home developped soft

Started by infopat

Hi,

I am a software Engineer and I am required to test our home developed software in a new virtual data center. We want to rehost the software into a Virtual Machine on Windows 2008 Server Entreprise R2 x64 based on a VM Ware ESX. Unfortunately, we are experiencing some big performance issues.

Our software is doing a lot of disk access (read/write) and that's why we are testing your RAM Disk software. We can allocate a lot of RAM to this VM so we allocated 26 GB and tested a 20 GB RAM Disk so that the system has at least 6 GB. We have tested a lot of configuration using your software but it doesn't improve the performance. We also checked some key performance indicators (CPU, RAM swap, I/O Disks, I/O Network, using the windows 2008 performance monitor) but nothing seems to be weird.

That's why we have some questions :
- Is SoftPerfect RAM Disk compatible with WM Ware ? ESX ?
- Are there some known issues with the OS we are using or the Virtualization ?
- Is mounting a 'RAM Disk attached to an SVI Image already created' less efficient than a 'simple mounted RAM Disk' ?
- Is the "save contents to Image" option makes the RAM Disk less efficient ?
- Is the "Mount as Removable" option makes the RAM Disk more efficient ?

At least have you some advices or "best practices" in using your software ?

If it works we'll certainly purchase more than 20 commercial use licenses.

Regards,

Patrick Ladouceur
Sofrecom - Part of the Orange Group
Paris, France
SoftPerfect Support forum - Andrew avatar image

Re: RAM Disks performance tests for a home developped soft   20 March 2014, 23:44

Hi Patrick,

It's barely technically possible that a conventional hard disk drive and a RAM disk have the same performance. When you create a RAM disk, make sure you create an actual RAM disk, and not an image file associated with a drive letter. The latter will not show any performance improvements.

I guess a simple and good test you can perform, is to create a RAM disk and then run a disk benchmark against that disk (and then another one against a physical hard disk). The results should be at least 10 times higher when using a RAM disk. If possible, please run the Crystal Disk Mark software on a regular hard disk and on a RAM disk and post here a couple of screenshots of the test results?

Finally, to answer your questions
  • We haven't tested it in ESX, but it works without a hitch in VirtualBox and should do so in any virtual machine. Windows 2008 is also okay.
  • Mounting a RAM disk with an SVI image apparently takes time to read from the file upon the RAM disk's creation, and if you opted for saving changes, more time for saving them.
  • Saving the contents to an image file, obviously takes time. If it's only saved on shutdown, that doesn't affect performance. However, if it's saved periodically, it does reduce performance and locks the disk for writing when saving is taking place. If you use these options, I recommend FAT32 rather NTFS as the file system.
  • The mount as removable option slightly alters the Windows caching strategy, but I doubt it affects performance noticeably.

Andrew

Reply to this topic

Sometimes you can find a solution faster if you try the forum search, have a look at the knowledge base, or check the software user manual to see if your question has already been answered.

Our forum rules are simple:

  • Be polite.
  • Do not spam.
  • Write in English. If possible, check your spelling and grammar.

Author:

Email:

Subject

A brief and informative title for your message, approximately 4–8 words:

     

Spam prevention: please enter the following code in the input field below.

 **     **  **     **   ******   **     **  ******** 
 **     **  **     **  **    **  ***   ***  **       
 **     **  **     **  **        **** ****  **       
 **     **  **     **  **        ** *** **  ******   
  **   **    **   **   **        **     **  **       
   ** **      ** **    **    **  **     **  **       
    ***        ***      ******   **     **  **       

Message: