• Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 832
  • Last Modified:

Why do large files take longer to copy than small files?

We are looking to use Veeam to backup our current VMWare infrastructure and I was researching cheap NAS solutions.  Well, after 2 days I'm starting to realize just how slow these things are.  It seems to have much to do with copying LARGE files as opposed to small files.  I haven't been able to understand why.

Example - one vendor (won't name names) sells an 8TB solution with 4x2TB SATA drives running at 7200RPM.  I found out that when copying large files (50-500GB each), the throuput is a paltry 35-45Mbps which means a 2TB backup could take over 100 hours!

1) whats the technical reason why large files take so much longe rto copy than small files and
2) can anyone recommend a cost effecient network storage solution?  I have a 10 hour backup window.

1 Solution
the nas solution use a sync mode, that become more the copy so slow, if you want to increase the speed, try to recofigurate the nas server, in async mode, if you nas device, permits.
it's a common know problem, that present nas+vmware

some example here.

as a solution, try to reconfigure in async mode your nas server, or use another I/O method, directly attach eSATA, USB (maybe v3), iscsi, obviously if you can do it.

i hope this helps
Fred MarshallCommented:
The nugget is in the data rate.  The data rate *may* be different for large files like this:

Send a small file and there's little competition for bandwidth.
Send a large file and there's more comptetition for bandwidth - so the data rate drops.
With switches in the mix it's a bit harder to know if this is actually going to make much difference as collisions are avoided.

What is the NAS interace?  General ethernet, right?  So you can get best case 100Mbps = 10Mbps with overhead included or 1Gbps = 100Mbps with overhead included depending on whether you have GBit ethernet or 100Mbit ethernet end-to-end.  So, 35-45Mbps doesn't sound all that surprising for 100Mbps network.

Some hard drive test programs measure the speed as a function of file size - I don't recall which one.  So that might be a place to start.  Yet, the numbers may be expected as-is.
Look for solutions that offer both nfs and iSCSI.  This was, depending on how you provision you LUNs on your NAS, you can offer up both protocols if you want  to your ESX environment.  Otherwise, iSCSI will probably give you the best performace.
Could there also be some issues maybe with the network config?  
Have you done a packet capture on your host to see what MTU the server is using or if there are any errors with protocol negotiation?  
Also, possiblty try a higher MTU than the standard 1500 (9000 if your switch supports it).  That might require some reconfig on your ESX servers though so be sure to test on a dev box first.

You should be getting faster bandwidth with larger files than smaller.  What rates were you getting with small files?


Featured Post

Prep for the ITIL® Foundation Certification Exam

December’s Course of the Month is now available! Enroll to learn ITIL® Foundation best practices for delivering IT services effectively and efficiently.

Tackle projects and never again get stuck behind a technical roadblock.
Join Now