How TCP Works - MTU vs MSS

In this video we will dig into the difference between the network MTU and the TCP MSS. Both Ethernet and IP MTU's will be explained as well as how the MSS is set in each direction. Demonstrations in Wireshark
If you liked this video, I’d really appreciate you giving me a like and subscribing, it helps me a whole lot. Also don't be shy, chat it up in the comments!
== More Training from Chris ==
▶Getting Started with Wireshark - bit.ly/udemywireshark
▶Getting Started with Nmap - bit.ly/udemynmap
== Live Wireshark Training ==
▶TCP/IP Deep Dive Analysis with Wireshark - bit.ly/virtualwireshark
== Private Wireshark Training ==
Let's get in touch - packetpioneer.com/product/pri...

[@skakfinn]

Simple and to the point without any waste of time ...Perfect...

[@lakku121]

Very crisp and clear . Thankyou Chris!

[@nikulvirpariya]

Thank you for sharing. Fantastic and easiest way to understand. Thanks a lot Chris !! Keep Sharing.

[@kenz0bux]

the simplest explanation! thank you

[@samirshaikh52]

Wonderful explanation as usual. Thank you Chris

[@wiltonwell]

Wow Chris, thank you so much, this has helped me so much. Looking forward to learning more from you. Now getting on the TCP with you and David.

[@akashranjan1631]

Thanks Chris, your videos really helped me clear my interview for Akamai, really appreciate your contribution.

[@dogood2848]

Thanks for this clear and concise break down between TCP MSS and Network MTU

[@FantonMr]

Thank you for your explanation in an easy way.

[@ijajahmed4301]

Thank you Chris for this video. It really helps me to understand clearly about MTU and MSS.

[@ChrisGreer]

Glad it helped!

[@josephmbimbi]

great video, very clear and informative, thank you.

[@alialbushama539]

Chris, Thank you for all the educational videos that you post. I have learnt so much by watching them.

[@ChrisGreer]

Glad you like them!

[@sigmadews5403]

Great explanation! Thank you!

[@PolymathTheDiver]

Excellent video. Good job Chris.

[@tuliksky]

Thanks for your explanation!

[@kimmy0795]

Very clearly. Many thanks !!

[@prdpbhadoo]

Its very helpful, Thanks Greer

[@christoffer913]

Thank you, Chris!

[@gaulogao]

Thanks, Chris. Your videos help me a lot!!

[@tarekz6934]

Very clear explanation , thanks!

[@cjcjonesit]

I don't what it is about you, but you make them seem so easy. Thanks a million.

[@ChrisGreer]

Thanks for the comment Carlton - happy the videos are helping.

[@ashwinramachandran118]

Thank you . Great explanation

[@Zeus-fm8tv]

Some serious stuff is explained in the easiest manner!! Thank you.

[@ChrisGreer]

Glad it was helpful!

[@ciscology8367]

Amazing video Chris very informative! I never knew knew about MSS . Keep them coming please.

[@ChrisGreer]

Will do!

[@vv-gd1et]

Well explained. Thanks Guru🙏

[@upelister]

Another great video thank you.

[@williebrown4266]

Excellent presentation. Both the graphics and the presentation complement each other which makes it very understandable.

[@ChrisGreer]

Thanks for watching Willie!

[@ChrisGreer]

Thanks Willie!

[@pcpmtiservicos9855]

Thank you for this great content!

[@ChrisGreer]

Glad you enjoy it!

[@KD-vc5mf]

Thanks! I was confused that the same word "MTU" was used on different layers. Now I got it :)

[@fluffy1973]

DUDE, you need to be on Udemy, this is real quality work.

[@ChrisGreer]

fluffy1973 hey! I have a “Troubleshooting Slow Networks with Wireshark” course on PluralSight. Check it out! www.pluralsight.com/courses/troubleshooting-slow-networks-wireshark

[@rudrashahi8261]

nice explanation...!!!

[@rookiery868]

Great video ! Thanks!

[@FerdsTechChannel]

Good video about Ethernet MTU and L3 MTU and TCP MSS! Thanks!

[@ChrisGreer]

Thanks for the comment!

[@mohamedhaneef6379]

Great video !

[@VyshfulThinking]

Amazing Chris

[@dimplejaral2239]

Thanks very much...

[@alexchamorro1884]

gracias por el video..!,saludos.

[@pawanshirke]

Hi Chris, Does the Ethernet MTU value be higher that 1500 if my Ethernet interface size if Tengig? Does the MTU size some how dependent on the actual interface size? In some routers we do define mtu for a an interface as 4484 or even higher value. This configuration of setting mtu size on an interface makes the topic MTU confusing to understand.

[@andygerard228]

Nice that was extremely helpfull.

[@ChrisGreer]

Glad it helped!

[@clarencealbay8327]

I always find your videos helpful. I am using NetScout solution. My comment is... You mention that the network device can interfere and change the MSS size.. If I want to know who modified the MSS size (if the network device or the server itself), then I should just compare the TTL and MTU of each capture of each network segment against the TCP dump. Because that can give me clue if the MSS changed against the TTL value Am I correct?

[@SaurabhKumar-lt9pk]

Very good explain 👌👌

[@SeroeKrevedko1]

Hi Chris! Appreciate your work! It would be great to add in video explanation about how work adjusts MSS in network devices. I can provide you with simple gns3 topology with c7200 routers and an example of how routers change the mss value

[@vasishtsubramanian4855]

www.cisco.com/c/en/us/td/docs/switches/lan/catalyst3650/software/release/16-9/configuration_guide/ip/b_169_ip_3650_cg/configuring_tcp_mss_adjustment.pdf

[@mrnetworker3544]

Thanks!

[@rodrigomunoz1556]

Hello Chris, thank you for this quality content, your videos help me in my daily work as a network engineer. I will really appreciate if you can give some advice about this issue. This is a TCP conversation, the client announces an MSS = 1452, server announces MSS = 1460, client uploads data to the server, this is a LAN environment, both devices are physically in the same Data Center in different VLANs. The problem: A lot of TCP errors: retransmissions (> 3%), dupACK, out of order packets, etc. What I see: fragmentation, If I perform a tcpdump directly on client OS, I see a lot of packets larger than MSS, ie 3000 or 4000. Why is client building that's long packets? could it be that TCP stack on the client side is not working as expected? There is other TCP conversations between the same hosts but in the opposite direction, in that case, all packets are equal or less than MSS, TCP errors are minimal (

[@dota2tournamentss]

I'm wondering if you solved it :) IMO application on the one or both side has incorrect TCP implementation and despite negotiating MSS they ignore it and they send packets much larger than expected.

[@F4t4lid4d3]

Wonderful explanation...I faced some issues with a pppoe link couple days ago. the root cause was the MSS configuration, after some adjustments the problem is solved.

[@F4t4lid4d3]

@Hussein Az Hi, you have to adjust the mss parameter on your device... In my case I'm using Mikrotik Routerboard. Google it for MSS and your device model...it helped for me.

[@tsushima33]

Thank you!! Great!

[@ChrisGreer]

You're welcome!

[@humayunakhtar2116]

Thanks chris

[@deepthireddy6354]

Hi Chris, Thanks for sharing the knowledge, I wanted to one thing. When I tried to change the MTU size to 9014bytes from 1500 bytes, am unable to see that change. I tried to change in configure option of particular ethernet interface. Please help me with this issue

[@saadqaddoori2633]

Hi, is it possible to disable MSS option in windows or linux?

[@abhijithks7419]

Great video Chris: talking about mtu, if I have an svi does the Mtu on my vlan interface have an effect ? Can Mtu on vlan interface and actual physical interface be different ?

[@ChrisGreer]

Yes, they can. On most switches you can set an interface-level MTU

[@yhytuncer]

Great video ! I wish you could add some throughput and bandwidth troubleshooting videos as well.

[@ChrisGreer]

Hey Leo! Be sure to check out the TCP congestion control video, I talk about those topics there - kzbin.info/www/bejne/gn_Ii42QlNSmoqs

[@anuraagtummanapally1395]

@Chris Greer what software do you use for presenting switches and nodes etc... can you share the info... at 0:27 a slide is there, I want to make similar slides, which software should I be using to get such figures... Thanks

[@ChrisGreer]

It's just Powerpoint with some Visio icons for the routers and switches. Nothing too crazy. Thanks!

[@tsrinivasnetwork6143]

Awesome video..

[@ChrisGreer]

Thanks!

[@VivekError]

Chris, would like to understand the difference between window size vs MSS.

[@adhak2011]

In 5:59, it should be pointed that MSS is not a negotiated value in essence between the 2 end points. However, the way Operating Systems implement it and deal with it might differ from the original description in the RFC and choose to work with the lowest MSS value advertised between the 2 sides.

[@ChrisGreer]

You are correct. I need to clip that part of the video - the MSS is an advertised value, not a negotiated one. Thanks for clarifying.

[@universalponcho]

Was told to watch this guys videos while reading my books for uni. Now I am watching this guy simply out of enjoyment lmao... 😄👍

[@ChrisGreer]

Thanks for the comment Jamey! Thanks for watching and thanks to whoever recommended the channel.

[@roswithadusa8673]

hello chris timeline 3:46 why in mtu are not the frame header (18bytes)included?

[@mustafaabdelfattah2493]

Thanks chris for the vidio , we hope the plalylist include new vidio for fragmentation process and its details, and what the offset number refers to ? , how pcap will be different at the client before fragmentation and at the server side after fragmentation happend in between , and is that fragmentation must lead to slowness ? How can i decide ?

[@ChrisGreer]

Hey Mustafa, great suggestion on a video. I'll take a look at it and see where I can work it in. It might be on a new series - How IP Works. :-)

[@akashvikraman7946]

Assume we have a host and 3 routers connection serially with different mtu. How does the tcp sender and receiver calculates mss value

[@rupees6]

Hi, you mentioned at around 6.05 that the sender and receiver should use the lowest value of MSS. But in RFC 879, it states that "The MSS can be used completely independently in each direction of data flow. The result may be quite different maximum sizes in the two directions".

[@ChrisGreer]

Hello Johnny - that is correct, the RFC does mention that it CAN be used independently. However, in most implementations, most TCP stacks will use the lower of the two values. For example, if a client sent a SYN with an MSS of 1300, it is rare to see the server send a 1460, or anythign higher than 1300. Usually it will send back a 1300 as well.

[@nayan271984]

Hey Chris, What is the exact difference between MSS and Window size ? I gone through the your videos and I found that in wireshark window size is 65535 bytes and MSS size is 1460. Now MSS value is local to interface and Window size is receiver buffer size. So if the Window size is 65535 bytes and mss size is 1460 then will client/server send data as per 65535/1460=44 segment of chunks ?

[@kanchadabest]

Great question, I want to know the answer of this question too ??

[@sujitprakashjha]

Hi Chris, First of all thank you for these TCP amazing explanation. I never knew this much TCP does and this series of video made me to understand it deeper. Could you please explain why MSS default value is 536 bytes?

[@ChrisGreer]

Hey Sujit! glad you enjoyed the video. The MSS default value is based on the IP minimum (maximum) packet size of 576 bytes. An IP endpoint should be able to accept packets of at least 576 bytes. They can be smaller than that, but the station should always be able to accept at least that as a maximum packet size. When you slice off the 20 byte IP header, then the 20 byte TCP header, you would get 536 as a payload, which would be encapsulated. Hope that helps!

[@exolicious1416]

I have a question: What i learned recently was, that the network layer (i am talking about the IP layer 3, which i am not sure if that is the correct english term), would not "look into" the encapsulated headers (layer 4 and 5). You however mentioned in your video that some network device along the way could have changed the MSS. I am a little confused and would really appreciate if you answered my question. Thanks in advance and thank you so much for the great videos on this topic. really appropriate speed and great visualisation!

[@exolicious1416]

Wow after scrolling through the comment section i just found the exact same question by someone else, whcih you already answered, disregard my comment.

[@ChrisGreer]

@@exolicious1416 Not a problem at all! Thanks for the question. I totally understand - I didn't know that L3 routers could adjust a layer 4 MSS value either, until it was something that I ran into on a client network. It is a good question!

[@ChrisGreer]

@@exolicious1416 Not a problem at all! Thanks for the question. I totally understand - I didn't know that L3 routers could adjust a layer 4 MSS value either, until it was something that I ran into on a client network. It is a good question!

[@ranjanadissanayaka5390]

hey, I got a question. The syn,ack from the server to the host -- is that the MSS negotiated between the two devices ? or is that the MSS on server's side (MSS of server) ??. Please answer.

[@ChrisGreer]

MSS is advertised, not negotiated. Both sides can have different values while respecting the one set by the other side.

[@gaborm4767]

​@@ChrisGreer Why my computer send maximum only 1400 bytes TCP payload if the other peer sent MSS value 1412?

[@neelabhsrivastava70]

Hi Chris. Can u explain the difference between Window Size Value and and MSS. Not able to find any tutorial for the same.

[@ChrisGreer]

Hey - sure thing - here is a video on the RX Window. kzbin.info/www/bejne/h6HOo5JnZ7eEbc0 That should take care of ya.

[@shkelqimemini1425]

If we have ipv6 and we know ipv6 have 40byte header and my question is Does the mss decrease even more ?

[@anonMarks]

Good

[@mitpatel4268]

Hi Chris, So does this imply that MSS must match on both ends? And, the negotiation "should" be such that they should settle up to the lowest MSS value between the two?

[@ChrisGreer]

Thanks for the question Mit - No, the MSS does not need to match. It is an independent value advertised by each end of the connection and is not negotiated. It is simply an advertisement of the largest segment size that a host can receive. So in other words - "Don't send me anything larger than this.... 1400" Or whatever value is used.

[@mitpatel4268]

@@ChrisGreer Great, thanks Chris. And, both MSS and Window Size would be determined by the local device itself and will be advertised to the opposite end, depending on internal buffer capacity, network conditions, etc?

[@dhruvneo]

Hi, Assuming only one interface and a P2P connection b/w two devices. If MTU size is configured to 9000, then changing the MSS to more than 1460 would make sense ??

[@ChrisGreer]

Hello Dhruv, That is a great question. In short - yes, if your MSS is 1460 on each side, then TCP will only take advantage of that segment size. However, in some cases the application in use can be jumbo-aware and adjust the MSS on its own. It really just depends on the stack/app in use.

[@soniaketh5784]

nice

[@techevangelist8373]

Great content. I have this doubt - In your other video named "how tcp works - the handshake" at 6:59, you said MSS value is something getting negotiated between sender and receiver.. and both the parties fall back to the lowest one..but here in some answers to questions in the comment session, you said MSS is not negotiated and independent of sender and receiver..are these contradicting statements?

[@ChrisGreer]

Hey - thanks for the comment. MSS is not negotiated. If I said that or implied it then I was in error. I may have meant that some stacks will respect the lower of the two values. But the value itself is not negotiated. If it is not advertised in the handshake at all, then the opposite side will generally assume 536 as a default value. Thanks for the question!

[@techevangelist8373]

@@ChrisGreer thanks for your reply

[@user-jd3jg1gx6b]

It is possible view MTU value on Wireshark?

[@StefanDekanski]

The most confusing part about TCP and MSS is about Options in the IP layer and TCP. From my understanding, you only 'count' fixed header size of both TCP and ID (20 each) and then trim off any data from the TCP segment, by the byte amount of Options of IP and TCP (that are used). Right? For example, if I have MSS of 1460 and IP without options and have one TCP option (4 bytes, or smaller but with added padding, for 32bit rule) then I can't really send 1460 but 1456 of actual data?

[@ChrisGreer]

Yes that is correct - If the TCP header is carrying options for SACK or timestamps, or other options, this takes space away from the encompassed data. So in these cases it is not uncommon to see payloads that are less than 1460 bytes. Remember the MSS is just an absolute maximum, given no options in the tcp header. However if it is smaller than that it is acceptable.

[@hrishekessaraghueeti1258]

Can we change the MSS size manually for an on going TCP connection?

[@ChrisGreer]

Hello - you can change the MSS at the stack level. Here is an example of how to do it in Windows 10. At the bottom of the article it shows how to manually adjust the MTU. This would work theoretically for any new connections. You can also adjust the MSS at the routing level as shown in the video. support.microsoft.com/en-us/help/900926/recommended-tcp-ip-settings-for-wan-links-with-a-mtu-size-of-less-than

[@avishayhakim]

Hi Chris, nice videos on Wireshark and TCP do you have any video that explains how to calculate the throughput I can get between 2 end-points? thanks :)

[@clarencealbay8327]

My take on that is to get a sample capture of any protocol across those two endpoints. Then sum up all the frame lengths across a minute (60secs) duration. Then whatever value of bytes or bits (8) you get, divide that value by 60. With that you can get the bit per sec or bit rate. There you go. (sum of all frame lengths across 1 minute duration) divided by 60 = bit rate or throughput. If you want payload or volume per minute, then no need to divide by 60.

[@mrspacer1]

is MSS value the same for receiving and transmitting?

[@ChrisGreer]

The advertised MSS is a statement of what the sender can receive in terms of payload bytes. Each side can have a different MSS. If the other station can receive a higher MSS, the opposite station should be able to transmit at that higher MSS, even if it is advertising that it can only receive a lower MSS. In practice, TCP stacks may vary on this. I have seen a stack that would only transmit the same size as it's own TCP MSS receive size. This is not according to the standard though.

[@mitpatel4268]

Hi Chris, I am getting into this and it is quite a lot of details if I try to put all things together. Nevertheless, can you please address one question? Suppose interface MTU (L2) is set to 9000. Now, if MSS is having a value of say 1460, does it make any sense/advantage of having the interface MTU very large (i.e. 9000)? Segments are naturally going to be smaller because of limited MSS value right? I am now puzzled on how the concepts fit in. Does the 9000 L2 MTU have any significance or advantage in this case? Maybe yes - because, so many TCP segments would fit in to a single L3 and L2 PDU? I hope my question would make sense to you.

[@ChrisGreer]

Great question as well Mit! Remember that other protocols will use a connection as well - such has UDP, or other IP traffic. Traffic is not always TCP, so a larger MTU at layer 2 could be utilized in theory by non-TCP traffic that can support a higher payload. But TCP will always go by the MSS.

[@mitpatel4268]

@@ChrisGreer Thanks Chris!!! So what you mean to say is - for any non-TCP traffic, values of 9000 as the L2 MTU can possibly be advantageous, but won't be the case for TCP traffic, given that we have a TCP MSS of 1400? Bottom line is - a frame can be only as big as permitted by the TCP MSS ultimately. Is that correct?

[@deepthireddy6354]

@Mit Patel Can we set MTU value to 9000, I have tried using some commands but when I send the data and checked in wire shark I able to see still the size is 1500 which is default value. Can you please give me the steps to change the MTU size in windows

[@amisriva1]

My understanding is that the MSS value is 'announced' and not 'negotiated' during the TCP handshake. I have seen material over the internet that will tell you that the 'lower' of the two (client and server) MSS values will be used. I don't believe its correct. Chris, would you clarify? BTW, your videos are just amazing

[@ChrisGreer]

Hello Amit, you are correct. The MSS is not negotiated, it is an advertisement. I have seen stacks respect the lower value before, however the value itself is not negotiated between endpoints. Thanks for watching Amit!

[@techevangelist8373]

Is there a concept like L2 MTU and L3 MTU? What I heard is L2 MTU is 1500Bytes + Ethernet header where as L3 MTU is 1500Bytes(Payload+TCP H+IP H). I always thought ethernet MTU is how much of data that can go inside an ethernet header(excluding the ethernet header). Can you correct me?

[@ChrisGreer]

Hello! Short answer - yes. There is an L2 MTU and L3 MTU. Both are basically the same number though, since L2 doesn't include the ethernet bits. For the most part i have seen the two terms used interchangeably, but they are different.

[@BrunoVernay]

At about 1' it is only true for IPv4. When using IPv6, there can be no fragmentation along the way. The MTU must be set no just with the other interface locally, but along the whole path. (For the 0.01% that already use IPv6 out there ...). Excellent videos by the way, I am watching the whole series backwards !

[@ChrisGreer]

Thanks for the comment Bruno! That is true - I should shoot another vid with an IPv6 slant since that one was v4 focused. Glad to hear you enjoy the content.

[@rendroyce]

Hi, I'm reporting a problem I've had in the past. An application is running at 1400 MTU. There was a problem with application access, I pulled the interface to 1400 MTU on the firewall. App access has been improved. Why did we have access problems when the application was 1400 and interface 1500? Will it not cause different access problems when you reduce the interface to 1400?

[@ChrisGreer]

Hello, I would need a pcap to be sure about what happened.

[@DikiciBurak]

Could you share the trace file with us Chris ?

[@rajivgoud1000]

I have a doubt Chris. As per your explanation, MSS is set during the 3 way handshake. So the data being segmented would not cross the MSS value. When does the IP fragmentation come into play then?

[@ChrisGreer]

That is a great question Rajiv. IP fragmentation would come into play if there is a lower MTU along the path than the packet size, including the TCP and IP headers. So you could have an MSS value of 1400 (which, if the TCP and IP header values are both 20 bytes would give you an IP packet length of 1440) but an MTU somewhere of 1000, which would then require fragmentation. Hope that helps.

[@rajivgoud1000]

@@ChrisGreer Thanks Chris. IP Packets can take any path and encounter a lower MTU device. Cleared. :)

[@Udaridamarakula1234]

you have course in Udemy ?

[@ChrisGreer]

Hey, I don’t. Just Pluralsight.

[@subhamthemusicalguy8851]

You are a gem

[@aboutturn]

Question-Routers work on layer 3 of OSI model..they deal with IP network layer Where as tcp mss is layer 4. How can router along the path change tcp layer 4 information(mss). I am really very confused.Appreciate if anyone can answer my query. Thank you.

[@ChrisGreer]

Hello Himanshu - Great question. You are right - routers do deal with layer 3, network layer stuff. However, they also pay attention to layer 4. Think about access lists at the port level - those are all layer 4 things. So many of them have the ability to look beyond the layer 3 info in the packet. Many enterprise level routers (and even branch routers) have the ability to make adjustments to values in the TCP SYN. On Cisco routers, you can use the command 'ip tcp adjust-mss' to change the mss of the SYNs. The reason to do that is if there is an MTU on the next connection or later on in the path that will require a smaller MSS in order to avoid splitting up the packet, especially for packets with the DF bit set. Thanks for the comment!

[@ankanshrivastava7517]

@@ChrisGreer Hello Chris - Thanks for all the details , i have a followup query that even if router has decreased the MSS of the source TCP packet , will it inform about it to source ? Because destination will get this updated detail and will send the reduced MSS size segment , however source will expect initial published MSS ? please correct me if am wrong in interpretation . how this will be resolved ? have you posted any video that can answer all these queries from scratch ? Thankyou !

[@ChrisGreer]

@@ankanshrivastava7517 Great questions! The simple answer is no - the router will not report back to the sender that it changed the MSS. But if there is an MTU problem down the road and ICMP is enabled, the original sender may see some ICMP fragmentation needed messages. But this will happen at the network layer, not the transport layer.

[@saifu62]

So tcp window and mss is same!?

[@ChrisGreer]

Hello Saif - no they are quite different. The Maximum Segment Size is an advertisement of the maximum encompassed data that a station can receive in one segment. The TCP receive window is the amount of data that the receiver has space for in its receive buffer. So they are quite different. Check out my video on the TCP Receive window to clear up that topic. Thanks for the comment.

[@howardcai1851]

router can change mss？😂

[@ChrisGreer]

Yup.

[@facebookinger]

Thank you for explanation. :)