0

We are using a CAD program on multiple computers in a network. All files and projects are stored on a network storage.

The users moan about long waiting times while opening from and saving to the network.

We (IT department) assume other issues than pure network latency. But to be sure we are thinking about a program to monitor and log opening and closing times of the CAD projects.

Maybe you guys have some ideas in mind how to achieve this goal?!

Our first idea was to log the start and end time of the network "stream". But how can we identify when the opening process of the file(s) start and end?

Some side notes:

  • a CAD project consists of a lot of files
  • all CAD files are stored on the same and separated network share
  • the data is shared from a Windows Server 2008 R2
  • all clients run on Windows 7 64Bit
  • two HP Procurve 4208vl switches are between client and server
  • the two HP switches are connected over 4gbit/s (4x fiber cable pairs)
  • about 6 clients uses the CAD program
  • about 50 other clients uses the switches and other network shares on other servers

Any help is well appreciated :-)

2
  • Run a packet capture program from one of the clients. Start a capture right before opening a project and stop the capture once the project is loaded. Then look at the capture.
    – joeqwerty
    Aug 28, 2015 at 14:45
  • Just for kicks, check your Procurve switches and see how evenly traffic is being distributed amongst the 4 fiber links (I'm always suspicious of lag groups/portchannels/trunking ports/whatever-you-want-to-call-them). In spite of my mistrust of them, you might consider a lag group between your servers and the switches. Aug 29, 2015 at 0:45

1 Answer 1

0

Thanks @joeqwerty for the tip with the packet capture program.

We now use https://www.wireshark.org/ and it works very well with the smb2 filter: https://wiki.wireshark.org/SMB2

We are using this filters:

  • do not show broadcast traffic: !(eth.addr == ff:ff:ff:ff:ff:ff)
  • only get traffic between client and server: ip.addr == _server-ip_ || ip.addr == _client-ip_
  • smb2 protocol flags: smb2.create_flags || smb2.create.rep_flags || smb2.close.flags
  • only show the CAD folders: smb2.filename contains "_foldername_"

so all together the filter is:

!(eth.addr == ff:ff:ff:ff:ff:ff) && (ip.addr == _server-ip_ || ip.addr == _client-ip_) && (smb2.create_flags || smb2.create.rep_flags || smb2.close.flags || smb2.filename contains "_foldername_")

with this filter we can now see the start and end request of all files in the CAD folder. with the associated timestamps we can calculate how long it takes to open the whole file with all sub-files.

not as automatic as we wanted, but it's a good approach IMHO. Maybe we have the time to automate the time measurement with a piece of java code and the jnetpcap library in a far away future...

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .