Microsoft Does Not Want You To Use iPerf3 To Measure Network Performance on Windows 60
An anonymous reader shares a report: iPerf is a fairly popular cross-platform tool that is used by many to measure network performance and diagnose any potential issues in this area. The open-source utility is maintained by an organization called Energy Sciences Network (ESnet) and officially supports Linux, Unix, and Windows. However, Microsoft has now published a detailed blog post explaining why you should not use the latest version, iPerf3, on Windows installations.
Microsoft has highlighted three key reasons to discourage the use of iPerf3 on Windows. The first is that ESnet does not support this version on Windows, and recommends iPerf2 instead. On its website, ESnet has emphasized that CentOS 7 Linux, FreeBSD 11, and macOS 10.12 are the only supported platforms. Another very important reason not to use iPerf3 on Windows is that it does not make native OS calls. Instead, it leverages Cygwin as an emulation layer, which obviously comes with a performance penalty. This alone means that iPerf3 on Windows isn't really an ideal candidate for benchmarking your network. While Microsoft has praised the maintainers who are trying to get iPerf3 to run on Windows via emulation, another flaw with this approach is that some advanced networking options simply aren't available on Windows or may behave in unexpected ways.
Microsoft has highlighted three key reasons to discourage the use of iPerf3 on Windows. The first is that ESnet does not support this version on Windows, and recommends iPerf2 instead. On its website, ESnet has emphasized that CentOS 7 Linux, FreeBSD 11, and macOS 10.12 are the only supported platforms. Another very important reason not to use iPerf3 on Windows is that it does not make native OS calls. Instead, it leverages Cygwin as an emulation layer, which obviously comes with a performance penalty. This alone means that iPerf3 on Windows isn't really an ideal candidate for benchmarking your network. While Microsoft has praised the maintainers who are trying to get iPerf3 to run on Windows via emulation, another flaw with this approach is that some advanced networking options simply aren't available on Windows or may behave in unexpected ways.
microsoft can lick my balls (Score:1)
I'll use AI to measure network speed if I'm in the mood. And of course not "on windows", windows has been out of the window for 2 decades now.
Re: (Score:3)
For iPerf - the common use case is not to measure the network performance on the host machine itself, but to measure network performance on the network. Ie, you're a router tester and want to measure throughput. In those cases you want a host machine that is not going to inadvertently affect your tests because it's a slower machine or, in this case, if the porting to the host machine is inefficient.
Microsoft is not saying "don't test us because we'll end up looking bad". They're saying that a non-Windows
Re: (Score:2)
Re: (Score:2)
People still use Windows? (Score:1, Insightful)
It's 2024 - people still use Windows? Why?
Re:People still use Windows? (Score:5, Insightful)
Re: (Score:2)
Obviously not. But people are always slow to find out things and are always willing to believe empty promises.
Re:People still use Windows? (Score:5, Insightful)
It's 2024 - people still use Windows? Why?
How detached from the world do you need to be to ask that question? People use Windows because it works for them and they don't give a shit about culture wars. People use Windows because it comes provided for them without effort required to change something. People use Windows because their employers mandate it.
Asking "why" facetiously isn't funny, it just makes it seem like you live in a fantasy.
Re: (Score:3)
I use both Windows and Linux. Linux is not some wonderful OS where everything just works. It's got plenty of its own problems and annoyances.
Okay, it doesn't spy on you out of the box, but if you can install Linux you can use Rufus and a few simple tools to disable all that crap on Windows too. With WSL you can run a lot of Linux stuff too.
Pick you poison.
Re: (Score:3)
A lot of businesses still use Windows. So especially if your work/home separation isn't 100%, and a computer is a tool rather than something you like tinkering with or playing games on, you'll probably just leave Windows on your home machine too (because that's probably what it came with when you bought it).
Rather than asking why people still use Windows, you might ask instead why we're still waiting for the Year of the Linux Desktop, even though Microsoft's relationship with customers has steadily progres
Re: (Score:2)
One big reason for that is that the meaning "the year of the Linux desktop" has changed. Back when the expression was coined, it meant the year when Linux was good enough for the average computer user to use it as their OS without needing regular intervention from a tech support guru just to keep it working and up to date. From my POV, that came almost a quarter of a century ago, but by that time, it had mutated to mea
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:3)
Allow me to suggest a slightly different definition. The "Year of the Linux Desktop" isn't when it overtakes Windows, but when it starts eating a decent-sized share of Microsoft's lunch, or maybe overtakes MacOS to move into second place.
I agree with you 100% that from a practical standpoint, an average computer user could adapt to Linux right now, and it deserves a lot more love than it gets. I think the main driver at this point is the persistence of Windows in a lot of businesses, and the fact that it
Re:People still use Windows? (Score:4, Informative)
>"It's 2024 - people still use Windows? Why?"
1) Because it "comes with the computer"
2) Because people don't know any better
3) Because it is what many know
4) Because some businesses force it on everyone
5) Because, sometimes, it is needed for certain applications
That said, all my computers run only Linux, have for decades, and will likely continue to do so (plus hundreds of machines I am responsible for). And probably way more than half of MS-Windows machines out there could run Linux and do what the users need, just fine. The SNAFU's with MS-WIn 11 artificial hardware "requirements" will likely open more opportunity for Linux on millions of otherwise perfectly suitable machines that will, otherwise, end up in landfills.
Better questions is to ask are: Why do certain popular commercial software vendors not support Linux? Why do we allow a near-monopoly (MS) to essentially force manufacturers to include their OS on most, if not all, of their computers? How much more abuse will MS-Windows customers tolerate before jumping ship?
Re: (Score:2)
I think your reasons are right but your priorities are mixed up.
1) Because, sometimes, it is needed for certain applications
2) Because it is what many know
3) Because some businesses force it on everyone
4) Because it "comes with the computer"
5) Because people don't know any better (this isn't real)
#3 is because of 1 and 2
#4 is because of 1, 2 and 3
"People don't know better" is a honestly bit insulting. I would never presume as a ride-or-die Windows user to tell a desktop Linux user they should use Windows b
Re: (Score:2)
>"People don't know better" is a honestly bit insulting"
It wasn't meant to be insulting. But we (on Slashdot) are not the typical user, either. I deal with users of all types, all the time. And I can tell you that the majority don't know what an OS is, or that they even have a choice of OS on what they own.
>" If I am a customer for a $8K a year software license is a $100 Windows license really that big an ask for the customer?"
No, but it isn't just about price, it is about platform choice and every
Re: (Score:2)
Re: (Score:2)
Sure i didn't think you going for insulting but just say it out loud, it can read that way. Many home users may not be aware that today they can accomplish goals with Linux and that it's improved a lot. I am admittedly a little sensitive as the anti-MS around here can be browbeating at times.
What does come with the platform is the question. Is it security? In enterprise environments the IT team is in charge of security and most times they want AD which there really isn't a Linux equivalent to (they exist
Re: (Score:2)
Do you want to teach my 76 year mother how to use Linux and provide tech support? I certainly can't be arsed with that, meanwhile she uses Windows without any help.
Re: (Score:2)
Until my mom had a stroke at 93 she was running SCO Unix. I couldn't get her to change to something more modern because she was comfortable with what she had. Now that she's gone, I should clear out a lot of old hardware I kept around in case her DEC pc crashed.
Re: People still use Windows? (Score:1)
Re: (Score:1)
I use Windows, Ubuntu and MacOS. Other than a religious debate, each O/S works well in a desktop environment. There are quirks in each and I just ignore them instead of fighting. Overall, I rate them all very similarly. Privacy concerns in Windows seem to be somewhat overblown, considering the same users have no problems with platforms like Google or Meta.
Re: (Score:1)
Re: (Score:2)
Find me the Linux equivalent of SolidWorks.
Re: (Score:2)
Ask the employers, companies, etc.
Re: (Score:1)
A non-story story (Score:5, Insightful)
In a not-very-shocking turn of events, Microsoft seem to make several reasonable points about a piece of software in a measured way. There's almost nothing to see here, move along (after registering the mildest of surprise).
Re: (Score:2)
It probably would be more effective for them to apply patches to the OSS software in question to fix the problems rather than complaining.
Re: (Score:2)
they are not complaining. simply stating you probably shouldnt do this, for these reasons, but if you still really want to, go ahead.
ya know, as is standard for most things that are not supported.
Re: (Score:2)
Except it's not, read it in detail. The issue isn't just updating the software, it's writing it to work natively, AND getting users to use the correct version (they make a point of the fact that the most prominent hit when searching for the windows version gives you one that is massively out of date and also has bugs in it).
Use the right tool for the job. Not all software is the right tool on every OS.
Re: (Score:2)
WSL was discussed in the comments, which makes sense given the current solution was built on cygwin,
(blah blah blah that's not 'native' either)
If marketing were in cahoots with engineering, you'd see a follow up blog that WSL would be *faster* in benchmarks than native Linux through the use of hypervirtualization! :)
Re: (Score:2)
I used to work in MSFT marketing. We'd insist on something like that rather than something like this. Putting our own presales people on the OSS submissions wouldn't have been beyond us.
That's where I got to quickly too... (Score:2)
Seems they'd be the ones that could measure how badly Cygwin is breaking things, and fix it if they wanted to.
Meaning are there issues it doesn't correctly diagnose? Or as they claimed... call types that have undue performance burdens? Then work on them.
Guessing Microsoft could be relatively good at changing their software in bad ways, and then make sure iPerf3 notices them as expected. Even contributing a test harness or fake driver to improve the QA process.
Or they could just ignore the issues as they
Re: (Score:3)
>"Microsoft seem to make several reasonable points about a piece of software in a measured way."
That was my take as well. A reasonable and valid argument as why to not use it. So used to FUD from them, that it is easy to jump to a wrong conclusion....
What a clickbait title (Score:5, Informative)
Re: (Score:2)
Shockingly, M$ isn't wrong here (Score:3)
I have a bunch of servers at work on 10gb and 25gb connections, some Windows based and some Debian based. Windows based Iperf3 (even the newer one on the budman.pw site) can often give results that don't match up vs what the machines can transfer via SMB or HTTPS or NFS.
It's opensource (Score:2)
Is Microsoft able to spare the resources to fix the code or advise the opensource developers?
Or are the hooks just not available without paid libraries?
Re: (Score:2)
Cygwin is the reason (Score:2)
The performance hit from Cygwin is the primary reason it's not good on Windows. It's like running on the beach with Cygwin being the sand.
Its like anything (Score:2)
Anytime you are doing "science" you need to know what you are measuring.
Cygwin inst emulation its a compatibility library. I highly doubt its use impacts network performance at all for certain parts of the scale.
CPUs are fast network cards mostly are not. You 14th i5 is going to outrun that 2.5GbE adapter cygwin or not. So if what you are bench-marking is the peer, say some router or IoT thing etc; I don't see the issue.
On the other hand if you are bench-marking the host with PCI-E 10GbE card or something;
Still better than running a test in a browser (Score:2)
And isn't Cygwin only a problem if the test ends up being CPU-limited? I've run iPerf3 on Windows on fairly old computers and they still saturate Gigabit-Ethernet, even on long distance links with non-local latency. What's the issue?
Cygwin isn't 'emulation'. (Score:2)
Cygwin is a way to turn *nix code into functional Windows results at compile time. Meaning yes, there is 1 more layer between any iPerf3 code and the native Windows calls, but it's not like there is a huge amount of work being done between the expected *nix calls and whatever Windows does.
Or there shouldn't be... I'm not a network diagnostic programmer. Nor have I written much at that level to compare both API's for differences.
Funny enough, Windows used to have a subsystem to allow *nix program API call
Another nail in Windows (Score:1)
PS And the sheer stupidity Windows API: for example you use mostly the same
They fucked up again? (Score:3)
And are now trying to lie about it? Seriously, who do they think is left that would believe them?
Re: (Score:2)
Who are you talking about? Microsoft, or the creators of Iperf3 who themselves don't recommend it for Windows for the same reason?
gweihir your ignorant anti-MS posts are getting tired. Stick to topics where you can at least get other people on your side, rather than topics where original creators of the software agree with MS's position on it. And be a bit more aware of your topic, literally half the posts here right now are talking about exactly the performance issues from TFS.
Problems with iPerf3 on Windows (Score:2)
I have used both versions of the utility along with the very old first edition for about 25 years and I did notice that they were very inconsistent results with the iper 3 utility on Windows, especially when using faster than gigabit connections such as 2.5 GB or 10 GB NICs.
They were also incompatibility issues trying to get version 2 to work with version 3 since they wouldn't talk to each other. So I ended up getting everything working with version 2 so I could cross-test speed between Unix and Linux opera
WSL? (Score:2)
I use Cygwin under my Windows notebook from work and my workstation at home. It works pretty well.
I tried WSL when it first came out but it is a per-user install so I don't use it. I wonder if running iPerf3 under WSL is better than using a version compiled with Cygwin libraries.
Re: (Score:2)
Not especially. iPerf3 is very performant and sensitive, which makes it a great tool for measuring network performance, but that also makes it sensitive to even small perturbations.
WSL2 is a virtual machine, so all network traffic goes through the virtual NAT system. WSL1 is technically a compatibility layer, but its recreating/emulating of Linux I/O typically comes at a pretty hefty performance hit.
Makes sense (Score:1)