Fiji/ImageJ on a Windows server

[Gelman, Laurent]

Hi everybody,

We recently purchased a server for image processing: 32 Cores (Intel Xeon E5-4627 V2 @ 3.3 GHz, 4 CPUs, no hyperthreading), 518 GB RAM, windows 2012 R2.
We then compared the speed of this server to that of a desktop machine: iMac, 8 Cores (Intel i7-4771 CPUs @ 3.5 GHz, multithreaded), 32GB RAM, Windows8 (I can explain why we run Windows8 on a iMac if some people are shocked, let's call this computer a "WinMac" :)).

These are results of the speed tests:
Stitching (Stitching plugin from Fiji): My "WinMac" is twice faster than the server.
Denoising (with the PureDenoise plugin): My "WinMac" is 3.6 times faster than the server.

The server has 32 cores, my WinMac 8 cores. I would not expect the server to be 4 times faster (CPUs are different, some writing to disk might be involved, etc.), but I would not expect it to be more than twice slower as well.

IT tells me, maybe ImageJ or Fiji or Java is not optimized to run on servers.

Is this the case? Is there a problem of configuration of Fiji or Java or Windows on the server? (I checked already "memory and threads" parameters).

Thanks for your help,

Best regards,

Laurent.

NB: When running Matlab, the server outperforms our desktops.

___________________________
Laurent Gelman, PhD
Friedrich Miescher Institut
Head, Facility for Advanced Imaging and Microscopy
Light microscopy
WRO 1066.2.16
Maulbeerstrasse 66
CH-4058 Basel
[phone]
[phone]
www.fmi.ch<http://www.fmi.ch/>
www.microscopynetwork.unibas.ch/<http://www.microscopynetwork.unibas.ch/>



--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Ben Micklem]

Hi Laurent,

I’m far from an expert on Java, and in fact have never run ImageJ on a
server, but the following thread hinted at optimisations that might be
needed:
http://imagej.1557.x6.nabble.com/A-note-on-JVM-performance-td3692131.html


Are you running the Server Java Virtual Machine?

Just done some more Googling, and it looks like the 64-bit version only
runs in the server JVM, so I guess you probably are using that.

This looks like a good place to start for tuning the server:

http://fiji.sc/Java_Options


Ben




On 26/08/2014 11:28, "Gelman, Laurent" <[hidden email]> wrote:

>Hi everybody,
>
>We recently purchased a server for image processing: 32 Cores (Intel Xeon
>E5-4627 V2 @ 3.3 GHz, 4 CPUs, no hyperthreading), 518 GB RAM, windows
>2012 R2.
>We then compared the speed of this server to that of a desktop machine:
>iMac, 8 Cores (Intel i7-4771 CPUs @ 3.5 GHz, multithreaded), 32GB RAM,
>Windows8 (I can explain why we run Windows8 on a iMac if some people are
>shocked, let's call this computer a "WinMac" :)).
>
>These are results of the speed tests:
>Stitching (Stitching plugin from Fiji): My "WinMac" is twice faster than
>the server.
>Denoising (with the PureDenoise plugin): My "WinMac" is 3.6 times faster
>than the server.
>
>The server has 32 cores, my WinMac 8 cores. I would not expect the server
>to be 4 times faster (CPUs are different, some writing to disk might be
>involved, etc.), but I would not expect it to be more than twice slower
>as well.
>
>IT tells me, maybe ImageJ or Fiji or Java is not optimized to run on
>servers.
>
>Is this the case? Is there a problem of configuration of Fiji or Java or
>Windows on the server? (I checked already "memory and threads"
>parameters).
>
>Thanks for your help,
>
>Best regards,
>
>Laurent.
>
>NB: When running Matlab, the server outperforms our desktops.
>
>___________________________
>Laurent Gelman, PhD
>Friedrich Miescher Institut
>Head, Facility for Advanced Imaging and Microscopy
>Light microscopy
>WRO 1066.2.16
>Maulbeerstrasse 66
>CH-4058 Basel
>[phone]
>[phone]
>www.fmi.ch<http://www.fmi.ch/>
>www.microscopynetwork.unibas.ch/<http://www.microscopynetwork.unibas.ch/>
>
>
>
>--
>ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Gelman, Laurent]

Hi Wayne,

I just tested Gaussian Blur. My computer is still twice faster.
As far as I can see on the task manager, those plugins I tested use multi-core processing.

Best regards,

Laurent.

-----Original Message-----
From: Rasband, Wayne (NIH/NIMH) [E] [mailto:[hidden email]]
Sent: mercredi 27 août 2014 02:38
To: Gelman, Laurent
Subject: Re: Fiji/ImageJ on a Windows server


On Aug 26, 2014, at 6:28 AM, Gelman, Laurent wrote:

> Hi everybody,
>
> We recently purchased a server for image processing: 32 Cores (Intel Xeon E5-4627 V2 @ 3.3 GHz, 4 CPUs, no hyperthreading), 518 GB RAM, windows 2012 R2.
> We then compared the speed of this server to that of a desktop machine: iMac, 8 Cores (Intel i7-4771 CPUs @ 3.5 GHz, multithreaded), 32GB RAM, Windows8 (I can explain why we run Windows8 on a iMac if some people are shocked, let's call this computer a "WinMac" :)).
>
> These are results of the speed tests:
> Stitching (Stitching plugin from Fiji): My "WinMac" is twice faster than the server.
> Denoising (with the PureDenoise plugin): My "WinMac" is 3.6 times faster than the server.

Hi Laurent,

Not all commands and plugins are multi-threaded. You should try a command, such as Process>Filters>Gaussian Blur that is known to be multi-threaded.

Best regards,

-wayne

> The server has 32 cores, my WinMac 8 cores. I would not expect the server to be 4 times faster (CPUs are different, some writing to disk might be involved, etc.), but I would not expect it to be more than twice slower as well.
>
> IT tells me, maybe ImageJ or Fiji or Java is not optimized to run on servers.
>
> Is this the case? Is there a problem of configuration of Fiji or Java or Windows on the server? (I checked already "memory and threads" parameters).
>
> Thanks for your help,
>
> Best regards,
>
> Laurent.
>
> NB: When running Matlab, the server outperforms our desktops.
>
> ___________________________
> Laurent Gelman, PhD
> Friedrich Miescher Institut
> Head, Facility for Advanced Imaging and Microscopy Light microscopy
> WRO 1066.2.16 Maulbeerstrasse 66
> CH-4058 Basel
> [phone]
> [phone]
> www.fmi.ch<http://www.fmi.ch/>
> www.microscopynetwork.unibas.ch/<http://www.microscopynetwork.unibas.c
> h/>
>
>
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html