To troubleshoot a nasty network issue I am looking for a Windows tool that can generate file reads/writes on a SMB share.
I checked benchmark tools like fio and diskspd but they all prepare a fixed file and then read/write into that. This doesn't seem to trigger the issue, so I was hoping to find something more life-like. Any ideas?
#Searching SMB Share Load Generator
void ferry
burnt ledge
I usually use Total Commander (or Explorer) and just copy big files from C: to the share (or vice versa). That obviously doesn't help if you're looking for particular load patterns (IO heavy etc) but at least it produces some "load"
weak socket
you can use diskspd, you just need to have it run multiple tests at the same time in smaller batches and mixed settings.
couple of references for SMB/NFS performance testing
https://unhandled.wordpress.com/2016/07/20/madness-testing-smb-direct-network-throughput-with-diskspd/
https://enterprise-support.nvidia.com/s/article/windows-smb-performance-testing-references
https://www.starwindsoftware.com/blog/hyper-v/whos-got-bigger-balls-testing-nfs-vs-iscsi-performance-part-3-test-results/
From MS
https://learn.microsoft.com/en-us/archive/blogs/josebda/using-file-copy-to-measure-storage-performance-why-its-not-a-good-idea-and-what-you-should-do-instead
https://learn.microsoft.com/en-us/archive/blogs/josebda/diskspd-powershell-and-storage-performance-measuring-iops-throughput-and-latency-for-both-local-disks-and-smb-file-shares
You can also use Lanspeed Lite or Intel I/O Meter
fyi, I have these notes handy because I just recently did this exact same thing for our Azure systems comparing NetApp CVO to other options
orchid lily
There's even IOmeter 2006 version. 🙂
What issue are you trying to solve? It's been my experience that benchmarks aren't reliable performance troubleshooting steps to take.
void ferry
It is not about performance troubleshooting. Our Windows VMs produce a lot of "netif.tcp.conn.bad.checksum" events on the NetApps. I've been in contact with VMware and Microsoft but they blame each other as which side corrupts the packages. I have been tracing with a productive Win 2016 environment so far, Microsoft now asks for tests with Win 2022. I am only the storage admin so I search for any possibility to emulate the production workload and hopefully trigger the issue.
clear hornet
rdfc.exe is one of my favorites
sleek bloom
why not just sniff the traffic to see if the checksums are actually ok or not...
void ferry
We did that but still are unable to identify the root cause.
sleek bloom
if you're actually getting frames with checksum errors then the remote port is generating them
get the fiber ports and ends "washed" and if that does't fix it, change the sfp's, probably on the transmitting side first
burnt ledge
if you're actually getting frames with checksum errors then the remote port is g...
it depends on the switch type though. cut-through vs. store-and-forward