Firmware Road Map

[mike99]

Oups, deleted, a little confuse when I answered LOL.

[tma]

This is confusing. To recap: Nasked said that port isolation is missing, and Chris said it is not - because for each port there is the ISO option. Nasked then pointed to a manual from TPLINK that explains that for each port, one would specify a list of ports that the given port may communicate with. This is obviously more than one could do with a single on/off option. So Nasked concluded that, in the Netonix implementation, port isolation could (at best) split the ports into two groups - the ports that are in the ISO league and the ones that are not.

It seems they were talking about a different kind (or level) of "port isolation". On a DLINK for instance, port isolation has the same semantics as on the TPLINK: it is a configurable many-to-many relationship between ports that probably acts on a very low level, i.e. even lower than the kind of isolation you get from VLANs. On the Netonix, since it's a single property of each port, what does it mean if a port is isolated? Is that like disabling the port except that the link state is still enabled but the port doesn't transport any packets? Or will isolated ports communicate (only) between them, as Nasked suspected?

[sirhc]

There is this new invention called a "search box"

Read these posts found with a search for "Port Isolation":
viewtopic.php?f=6&t=1126&p=8391&hilit=port+isolation#p8391
viewtopic.php?f=6&t=426&p=5096&hilit=port+isolation#p5096

Port Isolation (but discussed a couple times on forum)
Any port with Port Isolation Enabled will not talk to any other port with Port Isolation also Enabled. Pretty much the same as client Isolation on an AP.

I do not mind answering questions and I am always happy to help people out but "please" look and see if the answer is already here by at least attempting a couple searches.

Pretty much everything these switches do are discussed on here. People say we need a manual yet I can not even get most people to read the single sheet of paper I include on top of every switch we sale. But I agree a manual would be nice but here is a true story.

I used to be a programmer and one of the last programs I wrote in another life time was a civil engineering software package and at first I assumed the people using the software are all engineers and surely if the UI is simple no manual would be needed. Some engineering companies still use the software which you can actually download here http://www.hydrosoft.net/swm

Anyway everyone complained there needed to be a manual and these engineers would constantly call me and ask me the theory or equation used or whatever which could all be found in any civil engineering text book. SO anyway I eventually hired a top notch graphics person and a few co-worker engineers and we basically wrote not only a manual but explained all the theories behind the equations such as what a Froude Number is or where Manning's equation came from. This manual was not only a manual but basically a text book that could teach the average Joe the principles behind everything we did in the program. Next we paid a profession print shop to print up hundreds of copies of this manual and shipped them all out FREE of charge. NO ONE READ IT AND THEY STILL CALLED AND BUGGED ME DAILY.

[sirhc]

Since I mentioned hydrology it reminds me of grounding system and we all know I am pushing WISPs to use better grounding systems so stepping up on the soap box.

It is a very similar principle how a proper grounding system functions compared to storm water control systems.

The principles of engineering rather hydro dynamics or electrical overlap and intermix so much at the fundamental level.

In hydro dynamics we looked at capacity and resistance and how it affects the flow of fluids and when designing a grounding system it is largely the same principles.

Capacity and resistance determine the flow path.

Yes there are some very distinct differences but the principles are almost the same. Yes electrical current will flow uphill much easier then fluid current but in grounding designs you never want a ground cable to ever loop or bend up and then down. You should pretend your cable is a pipe and the electrons behave like fluid and they should all gravity flow down as efficiently as possible as bends and connections increase resistance for electrons just like it does for fluids but just like when designing storm drainage system you also have budgetary constraints so over design is kept to a responsible level with a reasonable margin of safety.

Electricity and fluid will both take the path of least resistance.

I always teach my kids that knowledge or experience is a lesson in how to think not what to think and the principle of something you learn today in one field can most often be applied tomorrow to another field.

Knowledge is power but knowing how to think and the ability to apply fundamental knowledge from one application/field to another is a super power for solving problems.

The only stupid questions in life are the ones you can answer yourself with the facts you already know.

[tma]

I was only pointing out that you, Chris, and Nasked were talking about different concepts of port isolation. To me it appears that one of your customers was making a feature request in a thread called "firmware road map". You told him that this feature exists, but he referred to the TPLINK manual which explained (better than he was able to) what port isolation concept he was asking for - which is not available on the Netonix switches.

So the question should be: Is this kind of port isolation available on TPLINK (and DLINK and probably others) worth considering for the road map? The answer could be that the silicon used doesn't support port isolation present on TPLINK (and DLINK and others). If it does, though, we should discuss whether there's a practical need for having it.

At that point, I'd say that I haven't used either form of port isolation so far. However, if at all, I'd see a need for the TPLINK kind of port isolation. Why? In your video, you were suggesting to split uplink ports (with AF gear) using VLANs in order to minimize flow control packets from 100 meg ports with half duplex gear having a negative impact on GbE uplink port(s). I found that to be a really good suggestion and so I went back and had look at my installations. And yes it seems this is something I should implement. However, I won't be able to use VLANs always because in the same VLAN there's a mix of 1 gig full duplex and 100 meg half duplex gear.

You may call that bad design, and it probably is, but at present I don't have a quick way to fix it. Port isolation - the way it is done on the TPLINK - could be helpful here because it adds another way of splitting the switch into port groups. I imagine that I could have the uplink port separated from other ports with the same VLAN ID. The kind of port isolation that is currently available would not work in this scenario.

Although my scenario seems to be a valid use case for the TPLINK style port isolation, I would not ask for it to be implemented - because, as said, it would only help to maintain a bad design. Others, Nasked for instance, may come up with more valid uses cases ...

[sirhc]

I was not referring to you Thomas or Mike or even just Nasked asking how Port Isolation worked, simply trying to get people in general to use the search tool more. This is why people who do not know me very well often think I am a jerk.....well, my wife and kids think I am a jerk all the time but that's a whole other story.

[sirhc]

Even Dave gets me wrong so it must be me and how I come across and not the rest of the world. Guess I better learn a little more finesse or I will find my self banned from yet another forum.

please give nicer answers to below..this guy is taking time to offer good comments
your a jerk chris

viewtopic.php?f=17&t=295&start=60

Sorry guys, it is just in my nature to be a blunt jerk I suppose even when I do not mean to be, I blame it on watching too much Discovery Channel.

[tma]

I've got something that may turn out to get away with a simple answer like "yes, you can do it with the present firmware", or it may translate into a feature request:

As pictured in this post, we build all major distribution sites with 4 switches. One reason for doing so is that we need more than 24 ports (the 2 routers eat 12 ports to start with) and the second reason is that on each switch we use only the upper or lower port row - so in case of failure we would simply reconnect devices to the other row on the remaining switch. These 4 switches are connected in an RSTP ring and each part of the ring is made from two cables being in a LAG. With firmware 1.3.9, this works beautifully.

However, as your video predicted, I found that the S3 and S4 switches - where we connect all PoE powered devices and, among them, 100 meg gear - have large number of FC pause frames sent towards the S1 and S2 switches. These FC pause frames will likely throttle traffic towards GbE devices also connected on the S3/S4 switches, like AF5 or AF5X. Your suggestion was to give these GbE devices their own uplink connection(s) using a different VLAN ID.

In order to marry our concept with yours, I would sacrifice the LAGs and split them into two separate rings - so one RSTP ring would be for VLANs 10, 20, 30 feeding the GbE devices, and a second "superimposed(?)" ring would be in parallel but reserved for the 100 meg devices in VLANs 40, 50, 60. From a cabling point of view, there's no change at all. But the spanning tree topology changes: we'd have two independent rings over all 4 switches and use VLAN IDs to keep the rings separated.

Question: Is that possible with current firmware or would we need to have something like MSTP implemented for it to work?

[drbaker]

Same here, we replaced a few mikrotik routerboards with only switching functions by Netonix and are really very happy with the resuts in terms of throughput, POE power, etc. But we loose MSTP (having a few independent vlan rings before that we can't reproduce now). Is there an estimated date for a firmware with MSTP enabled? Thanks!

[sirhc]

Same here, we replaced a few mikrotik routerboards with only switching functions by Netonix and are really very happy with the resuts in terms of throughput, POE power, etc. But we loose MSTP (having a few independent vlan rings before that we can't reproduce now). Is there an estimated date for a firmware with MSTP enabled? Thanks!

I will talk to Eric next meeting