[intro] hi everyone, welcome to the basic networktroubleshooting testing with iometer. we are going to be going through a few setups. first we are going to be setting up a perfmeterinstance to get a better look at what iometer is actually doing on the network interfaceitself and we are going to be setting up some basicnetwork workers in iometer, a dynamo work, and we are going to be setting an access specificationto deal with the networking component of it and then we are going to run the actual testand get our results. so the first thing we need to do, is we needto click "start", "run", type "mmc", we are
going to use mmc with a snap-in to add theactual perfmeter, because going to the actual performance monitor tool in the control panel for some reason windows 7 just doesn't letyou save it off, so we're going to use this technique which will allow you to save a shortcuticon for future reference purposes which i find extremely handy perfmeter is great for doing any kind of focusedtroubleshooting and getting some real-time dynamic feedback in the process we'll go to file here, let's see, add/removesnap-ins, look for performance, "performance monitor" "add", okay. i'm going to clean up some ofthis and make it look good. expand it out,
there's our performance monitor. now we can get rid of that window pane hereon the left, and get rid of the one on the right, shrink it down. alright! now let's clean up the look of this, justto make it easier to read. right-click on the background, go to properties, graph. we're going to add in the horizontal grid.for the vertical scale, we are going to change this to 200 because we're going to be workingwith a gigabyte connection so this is going to be in megabyte so the maximum megabytes you can get witha gig connection is approximately 124/125
theoretically, so 200 should cover our scale.our vertical scale anyway. okay. go to the appearance tab. i like toset my background to a nice and drab gray here. alright. now that's our basic look onperfmeter here. so let's go add some counter objects. clickon the "plus". it's going to populate here. let's go find our network object. there'sour network interface, expand that down. it auto selects them all, but you definitelydon't want them all. it's just information overflow, so let's go in here and pick bytes received, bytes sent, output queue length,that's a good one, and packets received/sec, and packets sent/sec. those are all good.
okay, so, if you have multiple nics, you'regoing to get two listed down here and you may not know which one it is. this isn't exactlyintuitive. so there's a little trick for figuring thisout. bring up a command prompt. select "start", "run", type "cmd". alright, now that we brought up the commandprompt, we're going to go ahead and type "ipconfig /all" that's going to give a listing of our interfaces.now you find your preferred interface, which should have the ip address of your systemthere and you will also notice name of that interface,and you can see there it's controller #2. so now that we have that piece of informationwe go back to perfmeter, select #2, click
add, and our counters are moved over for thatparticular interface. ah, back on the instances for the selectedobjects for our interfaces you could say "all interfaces", but it's going to double up onthe actual counters because we have two interfaces it's going to add twice what you selected up here, and again it'sjust a jumble of data so it kinda keeps it nice and clean if we just get specific. alright, hit okay. a lot of the time it pullsit in, this scale is not exactly accurate. so we're going to need to go in and adjustthe scale. to do that, right-click on the graph, againgo to properties.
for the data tab that is already selected,scroll over here to see what we are looking at here. okay, we're going to adjust all of these 5of these here. for the bytes received it's sent. we want to put that as 6 decimal places tothe left, so that is the fourth one down here. that will get us nice and accurate, in megabytes,reference in the hundreds. okay, doing the same thing for here, six tothe left, fourth down. ah let's see. i like to set my bytes sent, i like that to be red. that's my particular preference, and receivedgreen. then i'll visually clue in on that. okay, for the output length we're going todo two to the right.
well, i guess that would be one to the right.okay, packets received and sent; we are going to do one to the left. two to the left actually. alright, scale looksgood. we're going to hit okay there. alright, now that we got our perfmeter session allsetup, let's go ahead and save that off. so to do that we just go to file, save, andsave it to where ever you want. we'll just call it network. keep it simple. okay, perfmeter is good to go. now we aregoing to move on to setting up iometer. so first thing we are going to do; we are goingto startup iometer. and it should come in looking something likethis. it may have spawned a couple of worker
instances, if you have a couple of disks itmay have more or less than this. but we are not interested in testing the disk speed orperformance, so we're actually going to clean up thesedisk worker instances. we are actually going to be working with this network worker instancehere so let's detach these two disk workers so go to the detach icon: remove, remove.alright, now before we do anything else we have to create a manager on another host thatwe are going to be running the network test with. so access the remote host, or go to your remotehost. i have mine in remote session here. pull up either a power shell window, i'm usingpower shell here, or a dos window and you have to have iometer installed on this instanceof this host.
change to that iometer directory. okay, therewe go. and we are going to be running our "dynamo.exe". a dynamo process is a commandline version of iometer basically. it's a manager that runs in the background and itallows port communication for networking or disk process testing thatcreates a slave process and basically allows iometer application or gui to control it andpass and receive as it generates or receives data. so lets look at the help for that. so we'regoing to be using "iometer_computer_name" parameter and "manager_computer_name", sowe are going to be using the slash i switch [/i] and the slash m [/m] as in mike switch.
okay, so the iometer computer is where youare running your gui which is back on my other host, and the manager is this system. i justuse ip addresses to keep it simple. it may be a little easier for you that way. okay, so run dynamo slash i, okay that's theiometer computer where you have the iometer running, now for the manager, that's thishost, so we're pointing it back to itself. and go ahead and hit "enter". and you should see it startup. a series ofcommunications between the host where you have iometer and the host itself, and it'sdoing a quick check for physical drives and it found one on host-a which is the remotehost in this case.
alright, so this looks good. so it's waitingfor action from the iometer controlling client, so we'll minimize this. and as you can see,host-a was brought in, so let's expand that out. and it picked up the worker disk as itdid when we brought in our initial host. we're going to clean this up, because we'renot doing any disk testing. detach that thread. okay, good. alright, now what we'll do, andit doesn't matter what host you start from for the purpose of this tutorial we'll goahead and start from our iometer controlling host which is tank. we'll go over here to disk targets. noticethat it's grayed out since we don't have any disk target workers going on. we'll move onto the network targets, that's also grayed
out because we don't have any created. so on your tank host go ahead and select thathost or host where you are running iometer and click on this icon here "start new networkworker" and that's going to create a network worker right here. so we'll click on that, "worker 1" on tank,and notice under targets, it give us an option to select where we are going to be communicatingto or from. now we can essentially loop back to ourselves by selecting tank, but we don'treally do that since we want to do a network bandwidth test, we want to communicate it with the remote host on the other end,which in this case is host-a, so our target
is from tank to host-a. so we're going toselect that, and notice when i do that it creates a ghost worker on host-a in the topologyscreen window. and if you happen to have multiple networkinterface cards on host-a, or actually i should say on tank you can actually select what ipyou want it to come from in this drop down. but as we only have one, it just defaultsto that so it's not an issue. something to be aware of. alright, so we have our networktargets all configured so we are going to move on to our access specifications. what we want to do here is, what i normallydo is i create a series of send and receive specifications so you can independently testyour send and receive. we're going to do a
receive in this situation, so just for brevity, we are going to click the "new" button here.call it "net receive". let's start with 3k, change this to 3k. your sequential and randomaren't going to have an effect here. percent access specification access spec isn'treally going to effect it. what is important here is the write and read. write is "send",read is "receive" network-wise, so we're doing a receive test we're going to move that toread 100%. alright. and everything else should be good. we'regoing to click "okay". we're going to create another copy of this. let's do one at 16k.alright, good. we created a copy, so everything else shouldbe the same. and we're going to create one
more copy at 24k. and we'll save that off.okay good, so we have three new access specs created here, so let's add them to the assignedaccess specifications side. add 16, add, and 24, add. good. okay we wantto go to the results display and set that up. i'm going to set my update frequency to2 here. um these are a few important things here youwant to take a look at. total ios, since we are only doing a receive and send you wantto change that from total to write, actually we're doing read. that way it's not going to double up on it,so it will appear like you are getting twice the bandwidth. you don't want to do that,so adjust that accordingly.
actually lets change this top one here to"network packets per second", now it makes more sense. and for the megabytes per second,we are going to do "read megabytes per second". that will be much more accurate. alright,and here we'll do "packet errors", and here we'll do "tcp segments retransmits" just incase we do have some faulty media, we can see that. that will be very apparent, sincethat is a test that will be a good indicator for what's going on so ideally you want both those to read zero,then you're not getting any errors on your media. if you are, well, problem solved iguess. just replace the media. alright, and you have your latencies of courseso you can see your average and your maximum
response time for network latencies, so thatshouldn't be particularly high at all, otherwise you could have some other equipment issues or potentially firewall issues if you're testingthrough your firewall. okay, so this looks good. so now we are going to go to the testsetup. this is more for reporting purposes and runtime. we are just going to run our test for 15 secondshere. ramp up time, you really don't need that with networking, but i'll through a fewseconds in there. it's more of a disk setting. alright, recordresults, we don't want to record any results to a file, so i'm just going to say "none"here.
and everything else should be okay. normalis good for this scenario. alright, let's go ahead and run our test.uh, before we do that i'm going to pull up the perfmeter instance we created. okay, nowwe are going to go ahead and start this test. you can see the results from a previous testdone earlier. ah, go ahead and click the green flag to begintest, and it does the two second countdown ramp up, and we're off and running. now you can see we are on the first iterationof three iterations, we are on the 3k, you can see it doing about 20. perfmeter seemsto be a little more accurate than iometer, but he numbers should be fairly close.
i'm getting about 16 there, and we're rampingup for the second test run two, and we're getting about 24 meg now. now these numbersare subdued because i'm running this video screen casting software, so it's subduing my numbers quite a bit, otherwisei'd peak out on a third range but i won't. so we're into test three now, we are rampingup, and we are hitting about 44, 50 roughly and you can kind of see the step process thereand what is going on and another little handy feature is to clickthe highlighter there to get a highlight at the received which is what we are lookingat, and it helps you visually clue in on things when you got a lot of stuff going on yourperfmeter.
and the test is complete. and that's essentiallyhow it works. it helps you kind of gauge your bandwidth, do some testing, check for errorsusing iometer. get a good general sense of what's going on. it's essentially a load generator for yournetwork. you can do it with sending, you can do it with receiving. just make sure to adjustyour iometer metrics whether you're reading or writing. setup your access specification, and makesure it's appropriate in your results display. it's a very flexible tool you can do a lotof things with. you can some really neat advanced things or you can
pull in multiple hosts using the managersfrom the dynamo process and you can configure multiple workers to stream across there independentlinks into a single pipeline potentially and you can test the bandwidth of that pipeline. say they each have 1 gig each on the clientsand a 10 gig pipeline, and you want to know you are getting that 10 gig and then you canconfigure your workers to perform such a test. either monitor with iometer or tool you prefer. that's essentially how it works and it's agreat tool. i use it for a lot of stuff. it works great for network testing. i enjoy iteven more for testing disks and drives and i invite you to check out the article on itechstorm.com.
please feel free to leave comment, subscribe,and click the links and i'll look forward to producing the next video for you guys.take care.
0 comments:
Post a Comment