PsPing Power: The Ultimate Guide to Testing Bandwidth

foreign [Music] ER pool and this channel is dedicated to it students I.T professionals and anyone who enjoys learning technical subjects [Applause] thank you [Music] [Applause] we're on a series called PS tools part of this system internal Suite we are focused in on 12 tools called PS tools they are 12 utilities that are very powerful remote tools that can help the it pro do a lot of things in his daily tasks part four we're going to be focused on PSP looking at Network bandwidth testing all right so what is bandwidth well bandwidth is the data transfer capacity of a network in bits per second BPS U.S Federal Communication Commission defines broadband internet speeds as connection with a bandwidth of 25 megabits per second for download three megabits for upload let's hope that continues to rise notice the difference here in latency we looked at the time how long does data take to get from this point to this point bandwidth we're looking at the capacity how big is the hype how many bits per second can we move from this point to this point now bandwidth monitoring is very important in Enterprises there are consoles like this open source software third-party companies provide a lot of this live bandwidth monitoring of your Wan connections your VPN connections whatever you have that is incredible to monitor within your Enterprise you can get these kind of dashboard tools some of them are outrageously expensive and again some of them are open source it takes I don't know how long to get them set up but they're not easy to set up if they're free let's take a look at PS ping bandwidth one you have to have a PSP server process enabled on another device between the two areas that you want to measure bandwidth PS ping must use a port number for the server and you must use it in the client so that they can communicate via a designated port number Dash B will indicate to the client that you're going to do a bandwidth test ESPN Server doesn't care it will recognize what you're doing on the client you can use TCP or UDP UDP remember has no flow control you can really pump data through this you can receive from the PSP server rather than send now a lot of your parameters are the same with some of the other test modes let's take a look at some of the unique ones minus U will enable UDP bandwidth instead of TCP Dash H I had some problems with histogram where it said it did not have enough data it was hard to tell whether I had just instability just be aware that I have seen that there are also times that I was requested for a dash l in other words a requested size was required in the test you can use 150 bytes you can use 1500 bytes I'm going to show you we're going to actually play with some jumbo packet size and I'll show you all the details on that so something new for the bandwidth is the dash I this actually allows us to bump up the number of threads in the PS Hing process so that we can move more data more quickly the default is two times the number of CPUs up to a maximum of 16. now I actually did threads like 150 120 threads I don't know whether that gave me more data or not it says up to a maximum of 16. I ignored that and did 120 150 and I could actually see those threads so I'm not sure whether the maximum is 16 and ignores the rest of the threads that you create or not here's an example of my test setup and it's very important that you understand what is it you're testing for example this is my admin station Windows 10 hyper-v my parent partition is right here this is what I'm calling my admin station it's actually the parent partition on this physical box I have a virtual adapter then I go into a hyper-v switch and I go into actual physical Hardware this is a dual 10 gig physical Nick that I purchased on AliExpress it's extremely inexpensive did have a Microsoft driver which I was very happy the stability of all of this is always suspect when you're dealing with retail grade products and you're really pushing them to their maximum now if this was a server with server grade network card server grade drivers high quality switch you should get Rock Solid performance out of all of this no matter what but in my case I'm dealing with a lot of consumer grade subpar equipment because of cost restraints reliability is always a question my hazovo 10 gigabit 8 Port switch I purchased it's a sub 300 eight Port 10 gigabit switch it was so new it didn't have a manual I have yet to get a firmware update but it's 10 gig and I don't know how optimized it is it's basically default I modified some settings we'll take a look at it but whether it's fully optimized in a situation like this or not I have no idea at this point then I'm going to a physical almost a duplicate of this physical box again the same dual 10 gig hyper-v switch virtual adapter and then we're going to go into two different VMS that's our bandwidth path this is all software hyper-v switch is a virtual adapter is all software you really only have a few Hardware components the 10 gigabit dual Port the has a bow 10 gig switch and this 10 gig dual Port network card everything else is software and we're going to pump Allah data through these entities so we'll see some instability this is where my ESPN server was right here on this VM and I blew that up quite a lot it just became unstable if you're doing this type of testing with retail boxes it's probably best to reboot a lot give everything a chance to reinitialize and restart because you are maxing out everything as you start pumping data through now I do have another virtual machine and it's on a one gigabit network card over here same thing it's got a virtual machine hyper-v switch one gigabit network card connected to an HP one gigabit 12 Port switch I've got it connected to my 10 gig so we'll see this port over here on this 10 gig is going to be one gigabit to the HP switch to this one gigabit Nic and to the VM so we're going to from my admin station run a bandwidth test through this circuit right here so our second test will be from the admin station through the virtual Nic hyper-v switch 10 gig network card and gig switch another 10 gig Nic hyper-v switch hyper-v adapter into another virtual machine we'll look at the bandwidth through this and we'll look at the bandwidth through this we're going to start our test to this virtual machine we should see a 10 gigabit connection here and then a one gigabit connection to this virtual machine and we'll look at bandwidth here's my virtual machine and it's set up as a PS Ping Server it's using Port 5000 I'm back to my admin Center psping minus B so we're going to force a bandwidth test minus U we're going to use UDP minus I is how many threads we want to activate to create this traffic we're going to do 16. minus L we're going to fully take advantage of jumbo packets minus and we're going to do a thousand seconds of packet pings because we're dealing with a 10 gig switch and it doesn't take very long for that to happen we're going to go to 1.120 using Port 5000 so let's go ahead and do it it begins the sending of data and we can go look at our switch and we can see over here we're at 99 percent on that one gigabit Port so we are saturating that one gigabit port on that switch down here on the 10 gig notice we're getting only about 34 percent remember this is what we have the 10 gig is 34 percent going this way we've got 99 percent of this traffic right here just saturated so going this way is just totally saturated this one is fine it's got plenty of reserved but this is what we're testing right through here we're at six percent and this will take some time so I did a thousand seconds just so that it would stay active and we could look at what was going on I'm also going to look at my process on my admin station I'm going to go to process Explorer and if you notice as we come down to PSP I'm going to come up to columns and add a column come up to process performance and add we're going to look at threads and I'm going to grab this column and move it over and we can see right now we have 18 threads running on this psping process on my administrative workstation because we asked Dash I we set a maximum of 16. obviously two threads are being used by just the activity of the process but we've told it a maximum of 16 right now it's got 18. and that is what's pumping out this data it's interesting to know that our virtual size is relatively small we've only got about two Megs in virtual memory we're only taking about eight Megs in physical memory and that is found under working set and private bytes I'm going to stop this test because this would go on for a long time I'm going to go ahead and Ctrl C and let me check on my workstation my server and it said existing connection was forcibly closed by the remote host so it saw that control C it's back up again let me go back to my admin station and let's just do a reasonable test let's just do a thousand packets and I'm going to see if we can capture some histogram data and just leave it at the default which is 20 buckets go to end and notice it's done let's take a look the UDP Center bandwidth statistics that's how many packets were sent the 9014 jumbo packets this was the amount of data the bandwidth was 385 megabit I'm sorry megabytes be be really careful with that megabytes 385 megabytes 385 megabytes received we see nothing here packet rate was zero percent said not enough data for histogram because I have 10 gigabit Nix and a 10 gigabit switch I went ahead and set up jumbo packets we're going to walk through how to configure all of your components for Jumbo packets so everything has to be set for Jumbo packets the switch the Knicks the hyper-v switch and the virtual adapter let's take a look at the process that we take so that everything is set for Jumbo packets and then we'll show you how to test to make sure everything's right we'll start with our hasavo 10 gigabit switch if you come down I'm looking at the web console for the switch and if you come down to Port here is where we have the jumbo frame option I've enabled it and put in the 9014 value and then applied then of course you have to save your running configuration to your Startup configuration so make sure sure you do that on your switch now let's go to my admin station we're going to go to my physical Nick and set that for Jumbo then the hyper-v switch and then if we had a virtual machine which right now the parent partition actually uses the hyper-v switch so if you look here at my diagram for my admin station over here is my network connections from control panel we're going to start with a physical switch and that is one of these two ethernet converge network adapter x540-t2 now I know it's this one right here I'm going to open up the properties of that network card come up here to configure Nick go to the advanced Tab and under jumbo packet we're going to have the 9014 say okay now we've configured the switch the physical Nick now we need to deal with the hyper-v switch which is this one right here so this represents the hyper-v switch so again right Mouse click properties come up here to my virtual adapter under advanced and jumbo packet and 1914 bytes so my jumbo packet is now configured for basically these two because this is a parent partition these two act as one for the parent partition so for my second physical box notice again looking at control panel because it just is a better way of viewing your network cards here I see my physical I've got an Ethernet converge network adapter x540-t2 this is the only one that's enabled this is the one that I'm using so I'm going to right Mouse click again same thing configure it go to Advanced and make sure that jumbo packets are enabled come down here to my switch again configure and then make sure jumbo packets are set now this is this machine what we're looking at right now is the parent partition let's go into the virtual machine and set up the virtual adapter so on this machine this virtual adapter is separate I have to actually go into the VM and configure this virtual adapter so let's go into it this is actually the virtual machine that I'm using to test so now I'm in the virtual adapter again go here jumbo packets so it's quite complexed in getting all this correct so I just finished configuring this adapter the reason why I didn't have to do it over here is this is the parent partition this is a virtual machine so that the difference is that when I configure the hyper-v switch it can automatically configures the parent partition when I configure the hyper-v switch if it's a secondary virtual machine not the parent I have to go into that virtual adapter and configure that one for Jumbo packets now remember jumbo packets are technically jumbo frames because this is a ethernet frame you can see in these dialog boxes they're calling them packets which is kind of confusing because they are ethernet frames so be aware that the word frame and packet is a little bit nebulous when we get into using jumbo just to remind you jumbo frames and notice in Wikipedia they use the word frame typically ethernet is 1500 bytes jumbo frames if you have all the equipment that supports jumbo frames you can use larger than 1500 they're saying that it's commonly used 9000 bytes but almost every setting that I've found was 9014. just remind you when you're dealing with a parent partition the hyper-v switch and the virtual adapter of your parent R1 when you're dealing with a virtual machine not the parent partition you'll have your hyper-v switch to configure as well as your virtual adapter a little bit confusing there between a parent partition and a virtual machine inside now to test my configured circuits for Jumbo packet availability what we're going to do is we're going to go to the admin station run a very large packet size that way exceeds the 1500 byte that standard ethernet is designed for we're going percent of 8 000 byte packet and we're going to tell Ping command don't fragment don't break it in other words normally if I send an 8 000 byte packet through this virtual machine through the hyper-v switch the physical Nick it's going to break it up it's going to say whoa 8 000 bytes let me break it up and put it into 1500 byte frames and everybody's happy takes longer to get there but it will eventually get there what we're going to say is don't fragment it if it runs into anything that's 1500 byte limit it's going to Simply drop it so let's go ahead and test it ping minus F minus L we're going to require an 8 000 byte packet and we're going to send it to 19268.0.123 go ahead and go to the end here and notice it's going through just to show you that if I misconfigure anything in this circuit incorrectly it's going to drop this right now it went beautiful so everything from my parent party Mission through the switch over to the other physical box through the hyper-v into the VM perfect let's misconfigure one thing and see what happens so I'm going to come back to my virtual adapter here on my parent partition and my hyper-v switch which is this here so I'm going to go ahead here go into properties go into the networks section here Network properties go to Advanced and we'll it's set for 9014 and we'll disable it so I misconfigured one element of this path let's go back and check I'm going to up arrow and go ahead and so you can see it's saying whoa whoa what do you mean I can't send this 8 000 byte packet through this component without being fragmented so it's telling me right away it can't do it so this shows you with this setup and everything pre-configured when it works and if you don't have anything anything in that path that is misconfigured for Jumbo packets it's going to stop it's either going to drop it or give you some information like this saying I just can't do this now in order to demonstrate all this because it's a 10 gigabit Network that we're going to be playing with and it's jumbo packets everything happens very quickly so what I'm going to have to do is make sure that each test runs a long time because if I do 10 000 packets it's done before I can move my dialog box so that you could see something completed so things happen very fast on 10 gigs so I'm going to run the very long test which makes things unstable because I'm pumping so much data so bear with me our first test we're going to do PSP minus B which is going to turn on bandwidth minus U is the UDP minus I we're going to do 16 threads in our client side minus L is going to be a packet size of 9000 bytes minus n we're going to do for a thousand seconds and we're going to send it to 192.168.0.123 colon 5000. It's Beginning the test let's go take a look at our so here's our switch and we're running at about 34 percent on one port 33 on another port not bad I'm really pumping data notice these are both 10 gig and limitations start coming in because of virtualizations you've got a virtualized switch you've got a virtualized adapter so a whole lot of limitations start kicking in not because of the physical connections but because of virtualization you do lose some due to virtualization I'm running at five percent this is a long test and we're just doing it so that we can see what's going on otherwise it gets it's finished before I can even move from one dialog box to the other this is the virtual machine that I'm sending it to this has got the psping server you can see it's connected it recognizes we're doing a UDP bandwidth test and we're at 10 percent so I stopped the first test take a look what I did on the second test PS ping minus B minus U minus I I decided to try 160 threads minus L left the bytes in the packet the same and everything else but I upped the amount of threads let's go take a look at the switch we're now pumping 47 and 47. so adding more threads gave us more data through the pipe so here you can see I had it 160 threads I stopped that one and tried 260 threads so down here psping minus B minus U minus I I went to 260 and when I went and looked at the switch it really was no change so adding more threads at this point did not seem to make any difference we were pretty much at Max here we understand that our quote band capacity is larger according to this this switches graphic monitor but our actual thorough put capability is relatively limited it could be because of any one of these components or a combination of these components virtualization virtual adapters virtual switches we don't know what is really producing the actual thorough put this circuit but we know we're not anywhere close to exceeding the bandwidth of that switch let's do some more testing PSP minus B minus U minus I I'm going to go back down to 160. it didn't seem to do any good going above that minus L we're going to do the 9000 byte packet minus n here I'm actually going to just do a thousand packets and we'll see the end results here here we see our maximum bandwidth was about 521 Mega bytes so you could multiply that by eight and you would have how many bits per second which is almost a gig and 521 521 and then the receive site is slightly different or about half of the Cent a udb packet rate is 67.939 and again in the documentation there is nothing explaining this UDP packet rate nothing you're just having to well I think it's because there's nothing in the documentation not even in the book so it's very frustrating when something that is so much effort has been put in to develop and you don't have the necessary information to interpret the results next I did the same exact test this time I added the dash H so we'll collect some histogram data and here it says not enough data for histogram which again the error message there's nothing in the documentation that says oh you have to do this it says increase the request size well it's already 9 000 bytes after much playing and trying and tweaking I finally found a combination in which I got some histogram information one I had removed the UDP completely I had to up the amount of packets sent so I went 10 000 packets my packet size was 9014 histogram just 10 buckets or 10 bins and I finally got histogram data from my output what was my takeaway having spent days and days with psping especially with this bandwidth section very frustrating the parameters I think I understand I've played with enough parameter options to get a feel of the parameters the results were always frustrating in that the results would vary in their output there was no documentation explaining what was in front of me you're at the mercy of just trying to well I think that means something which is very frustrating you can get an idea of bandwidth latency all of these things it's useful it's powerful I wish more documentation would have been included I don't believe for one minute that Mark racinovich who wrote this tool didn't understand everything that's happening just a little bit more documentation would have made this tool a lot more useful and less frustrating Mr Vanderpool why are you wanting people to become a member of your channel don't you realize they're going to have to pay 299 every month out of their income yep I do and actually YouTube takes 30 percent of that so we don't even get 299 when you become a member of the tech savvy production Channel I started teaching technology almost 32 years ago and I taught primarily electronics at first Electronics technology and then I switched to Information Technology about 25 years ago I had a little gentleman from Bolivia he was just incredibly energetic student and he was teaching me how to edit ini files in Windows 3.1 and I just was hooked on Information Technology never looked back started doing Cisco started doing all kinds of courses so that I could move my program that I was teaching at the time from Electronics to information technology and I'm still in contact with yavier from Bolivia my first technical course in Information Technology was Windows for work group 3.11 some of you are just drawing a blank you should it's a long time ago troublesh around January 3rd 2010 I began to take lectures that I taught in class every day and we would just put a camera and point and shoot and it was raw I mean you there was no editing there was just a little bit of editing on the front end of the video we edited a little bit on the back end and we just stuck it up on YouTube so after I retired over the last three years I went back to YouTube can I get an audience of serious technical Learners who would generate enough ad Revenue that could pay the basic cost that required to make those videos if you're thinking Mr Vanderpool you have a nice home can't you just afford to do this for free well that's not actually my home this is so this is my real office and over there is my dachshund under the blanket and my chihuahua on the floor and my HP server behind me is sitting up against the wall I got a Linux box back here I got a whole lot of technology in front of me that's really my home so we're not trying to generate a whole lot of Revenue just to pay the cost of this kind of training but here's the bottom line the type of material that we're creating on this channel is for serious technical Learners we don't get a lot of views in other words if I did a cooking channel and I was teaching you how to cook barbecue I could probably get a hundred thousand views and we wouldn't be having this conversation but because I'm creating very very technical content and I try to make it very accurate and engaging the audience shrinks down people that really like this kind of content that want it engaging want it accurate and want to learn are the people who watch these videos and that's the group that pays for the cost so if you become a member that 299 subtract what YouTube takes is what allows us to move forward you don't have to become a member for six years just become a member for five months I would like to continue creating content that is technical engaging and it's after that 10 out there in YouTube world who really want to learn so why do I need you a member I need you to partner with me to make this channel successful help us cover our costs that's all we're asking and become a member [Music]

Channel: TechsavvyProductions

Views: 1,823

Keywords:

Id: fpbQ1J_Cfbs

Length: 31min 36sec (1896 seconds)

Published: Sun Sep 24 2023