Butch, You completely missed my point, ot the background to the thread.. Of course "you" can build a tunnel of just about any MTU size on "your" network. The issue at hand is what max MTU size OTHER upstream ISPs allow on their network.
Scott was talkng about doing a tunnel accross an end user Comcast circuit. With the open Internet, the two end point end users don't really have control of what ISPs in between gets traversed from End Point A to B. The ISPs in the middle could chance at any time. This is not a new problem.... For example.... A number of years back Universities built a "private" experimental transport network to support high MTU above 9600, so that their GB and 10GB networks could pass full capacity. As you know, max transfer rate is directly proportional to latency times packet size. Most common ISPs only passed 1500MTU, therefore the Universities had to make their own net. This has been a challenge for years for even passing VLAN tags or MPLS data, where layer2 fiber carriers would only pass a 1512 packet. When you are the end user, the answer is to shrink your MTU, so after the tunnel overhead it fits into the ISP's max 1512 MTU. But when one is an tranport ISP that transports many customer's data, it is not appropriate for the ISP to shrink his MTU below 1500, as all the other end users would not know that the MTU was shrunk, and would not have their routers set to a smaller MTU to fit. Sure you can allow fragmentation, and TCP will automatically split the packets to fit, but it has been common ISP management practice to disallow fragmentation for various reasons that I don't want to get into in this thread. And yes, there is MTU autolearning, but again, not supported by everyone or all protocols. So sure, the ISP can make a tunnel setting a lower MTU, so after tunel overhead, it will fit in the uipstream's 1512 MTU. But then full size packets (because packets comming from end user customers will be 1512 size) inside the tunnel will get fragmented to fit into the tunnel. For long haul backhauls, there can be side effects of just simply allowing fragmentation on the routers without any further consideration. Again, we have a good solution for this... It is called CIPE. Its a tunneling protocol that splits the packets appropriately for optimal efficiency. I understand how CIPE works because it is what we use. I can't say I understand the methods that Mikrotik may use. So, what I asked is how Mikrotik can deal with that problem, because Mikrotik does not support CIPE. Tom DeReggi RapidDSL & Wireless, Inc IntAirNet- Fixed Wireless Broadband ----- Original Message ----- From: "Butch Evans" <but...@butchevans.com> To: "WISPA General List" <wireless@wispa.org> Sent: Friday, April 24, 2009 11:03 PM Subject: Re: [WISPA] Interesting BGP Redundancy Opton for FREE > On Fri, 2009-04-24 at 19:37 -0400, Tom DeReggi wrote: >> Over a Layer2 PTP its usually not an issue, but it is over a standard >> transit connection. >> (customer and Internet needs to see 1500 bytes, but an ISP's tunnel >> causes >> packet size to exceed 1500 MTU. > > I have built tunnels that carry 12000 byte packets. Not sure where this > idea comes from. They can be built that will carry as much as 65k > bytes. > >> We use Cipe tunnels to solve that. To split the full size packets before >> it >> enters the tunnel, so tunnel stays at 1500MTU or less, required by the >> transit provider.. >> >> How do you do it with Mikrotik ? > > Of the tunnels I've done with MT, you just use PPtP and set the MRRU > (just like your tunnels). I've done this with standard Linux, too. It > is actually quite an elegant solution. > > -- > ******************************************************************** > * Butch Evans * Professional Network Consultation* > * http://www.butchevans.com/ * Network Engineering * > * http://www.wispa.org/ * WISPA Board Member * > * http://blog.butchevans.com/ * Wired or Wireless Networks * > ******************************************************************** > > > > > -------------------------------------------------------------------------------- > WISPA Wants You! Join today! > http://signup.wispa.org/ > -------------------------------------------------------------------------------- > > WISPA Wireless List: wireless@wispa.org > > Subscribe/Unsubscribe: > http://lists.wispa.org/mailman/listinfo/wireless > > Archives: http://lists.wispa.org/pipermail/wireless/ -------------------------------------------------------------------------------- WISPA Wants You! Join today! http://signup.wispa.org/ -------------------------------------------------------------------------------- WISPA Wireless List: wireless@wispa.org Subscribe/Unsubscribe: http://lists.wispa.org/mailman/listinfo/wireless Archives: http://lists.wispa.org/pipermail/wireless/
