I am not that interested in wireguard, but its proponents always claim that it is so much faster than IPsec, so when they are right it should be useful.
IPsec on those devices will do like 20-30 Mbit/s and that is usually okay for the normal usages of a VPN.
(I am not talking about the usage of routing all your traffic from/to internet over some remote party’s network)
ok, another example
MT7621, Could you make untagged vlan on Switch chip ? NO
does OpenWRT could do this? YES
is new FW useful? NO
MikroTik internally uses VLAN to be able to configure each port completely independently. It looks like the switchchip-level VLAN is used for that.
Maybe in other software it is done differently, e.g. when they do not have that capability.
Or maybe there is some lack of support of a certain switch chip.
Does OpenWRT support hardware-accelerated routing on switch chips, like new RouterOS now does?
and, maybe the most brutal question
32/16 devices, or maybe 64/16 devices intended for home use, mostly equiped with 100mbps ports
what is the use of new FW ?
multi core BGP? Hhahahahaha
No, but new firmware could have support for newer WiFi features that the hardware of those devices could support
but the software until now didn’t. (because of the use of in-house written drivers instead of those provided by chip manufacturers and used by other companies)
It also could have improved IPv6 support which everyone can use, also in home use.
so, what i was wrote, maybe not so clear, that v6 need to stay as it is
with slight improvements to support old devices, few security patches per year, and this is it
old HW is NOT cappable for new FW.
This has not been shown. Apparently there is work being done on some issues with TILE and I do not like that releases for other architectures are held back because of that, and I think the hAP lite and mini have to be dropped (they do not work well with v6 either), but other than that things seem to go a lot better than at the named other companies.
I think we as LONG time MT customers should ask and suggest to put more time and resources to development of V7 for the ARM and CHR and release a weekly or so beta and then when V7 became stable or in the LTS then start adding more devices in if it possible.
And I’m sure changing for 16MB to even 1GB nand will cost MT less than $1, can we ask then to prevent using abd developing hardware with less than 128MB.
Ok
we are talking about 20/30 mbps as MAX on old devices
So, at this moment, i am “home” user, with FTTH drop
FTTH is not so SCI-FI today. Or internet speeds > 100mbps
These devices have 100mbps ports, capable of 20mbps VPN, and ?
It is normal to force kernel 5.x on 100mbps devices?
How could you expect that somebody with FTTH connection say: OK, i have a giga input, but my old router is 100mbps. Woow
And i could do 20-30mbps VPN. Woow x2
no, it is not enough. Even for internal network
So what is use of wasting time/resource for these old devices? From my point of view, security releases for v6 is enough
Modern kernel on old device for what? And other people who have modern device, suffering from MKT decision to postpone v7 because compatibilty issues…
ok, about OpenWRT
Yes, they are running DSA, HW offload, on most architecture with Kernel 5.x
I am using it every day
No, i don’t say it is perfect, but, it is community driven project, and it is WORKING!
MKT is paid project, and it is NOT working. Do you see difference?
and no, i am not arguing
only higlighting some facts and wrtiting my oppinion why MKT should drop support for 32MB ram and 16MB flash in v7
have a nice weekend
73 & DX from from ex HAM, op Peca, yu7ady/yz7rly
@npeca75: Amuse yourself and look up average internet speeds. I’m also lucky home user with 200Mbit FTTH, but most people have much less. So someone who is happy with 20Mbit VPN is pretty normal. If they now use TCP OpenVPN with v6, then new v7 with UDP OpenVPN is good example of useful improvement. And WG even better.
I won’t defend all MikroTik’s decisions. I think that some of them are really bad (16MB flash). But long term software support is good one, in comparison with others, where you get almost no updates. A something in between approach, where older devices would keep v6, but would still get new useful features, would probably be ok too. I mean for customers, but I’m not sure that it would help MikroTik that much. Backporting stuff needs work too. But once you get the new version going, you have just one to worry about.
In retrospect it seems unfortunate that they used so many different architectures. But it’s not necessarily what’s slowing them down now. Once you get the kernel going, lot of other stuff is user space and common for all architectures. Yes, they publicly admitted that they are waiting for TILE right now, but it’s a special case, they can hardly cut off architecture used by their high end devices, especially when they still sell them. It would be much easier for e.g. SMIPS or PPC. I’d actually agree with SMIPS, because they already removed some features, so even v6 is not complete RouterOS, and the hardware is clearly too limited to have any bright future. And nothing against PPC, but no recent models have it, so they could get away with it (but I’m not saying they should, unless it’s too big burden).
I agree with Sob, my internet is 100 Mbps down 30 Mbps up so for a VPN my speed is not going to be above 30 Mbps. And that is even quite high, on other connections it may be 5 or 7 Mbps in this country.
Also, I prefer RouterOS over OpenWRT any time. But I think those that don’t are welcome to use OpenWRT (even on MikroTik hardware).
Custom developed (“home made”) devices drives and thread/concurrency problems like kernel-mode deadlocks are example of quite common issues and are particular error prone in case of low level development. The final 5% of the bug management often consume more than 90% of the total development time. That’s why it’s important to streamline the development process with ready made drivers and and skip as much as possible of the “not invented here” syndrome that might linger from the earlier days of RND.
People think that way but this is not only a +1$ for memory chip… and the same for switch chip, many times this is CPU internal feature who have Ethernet on-board, means inside CPU.
How many people say… MikroTik, why LTE devices use 100Mbps, this is only +1$ to price. This is so common.
Example of 1$ upgrade for ETHERNET:
179$ LHG LTE6 kit
199$ LHGG LTE6 kit
give you +20$ and the PCB must be reorganise for new CPU… arch was change… but the same LTE module, the same antenna gain.. only CPU must be changed to reach 1GbE
20$ is for many people big difference. My 90% will be still normal LHGR without lte module + m.2 adapter + Quectel EM12-G
My summary is differ.
People not see a real bottleneck and can not understand why 100Mbps is sometimes a good option, crazy ?
marketing bottleneck: 100Mbps, but people with SXTR/LHGR/LDFR buy this hw for build-in “external antenna” in one box for start reaching more then 10/2Mbps.
Many person who write to me connect from far away where other ISP/WISP not exist . This is my scenario. They always say… this 100Mbps, if not this my LTE will be faster ?
for me bottleneck a) - LTE modules base on regions SoC, not Global. Chateau use Quectel EG12-EA, why not global ? Because MikroTik belive lower price and differ devices for differ regions will keep the lower price.
for me bottleneck b) - SLOT mPCIe… probably with ROS7 stable MikroTik introduce 5G modules base at M.2 SLOT, I hope but this open path to insert many other non-mikrotik devices in future. Maybe they stay at mPCIe and this will for us start of era close unit for only MikroTik lte modules because m.2 adapter cannot solve heat problem…
for me bottleneck c) - R11e-LTE6 Cat.6 without 2CA for B1+3/7, I hope this will be solved by feature firmware update and new lte modules can do CA for UE B1/3/7/8/20 any combination.
for me bottleneck d) - They are others, as always
In my opinion, the bottleneck a/b/c should be solved before or with 1GbE Ethernet feature. Solving 1GbE first by +20$ not change situation for me a lot.
Basically I agree regarding the GBE interface but it feels like a slightly twisted comparison with the other $20 since the price with your configuration using a “Quectel EM12-G” adds another $330.
Another major functional bottleneck is the antennas in LHG which are marketed for use in extreme rural conditions. Considering these conditions when only the lower LTE bands are available in practice, the LHG antenna has effetely no signal gain at all which is quite absurd and directly contradictory considering the marketed use case.
I believe it’s on the edge of false marketing using a statement like this:
“The LHG XXX is a device for very remote locations that are within cellular network coverage. Mount it outdoors, on a pole, mast or any tall structure, and connect even where cell phones fail. Due to its large sized high gain antenna, LHG LTE6 kit can connect to cell towers in extreme rural locations, giving you the ability to provide last mile internet access where nothing else is available.”
Regarding LTE-modules I would love to see a simple DSL (domain-specific language) with hooks (for ie “enable/disable/connect” etc) that can be used as a replacement or add a new drivers to control MBIM/AT for whatever reason. In the bottom line, it’s just text to be acted on. That would make it extremely fast en easy to add new LTE-modules that make them functionally equivalent to for example R11e-LTE6.
EDIT: Changed DLS to the correct spelling “DSL” (Domain-specific language)
LHGR (without lte module) 102$ + m.2 adapter 5$ + 2x pigtails (u.FL F <5cm,10cm> MHF4 F) 2.5$ + EM12-G 90$ = ~ 200$ for finish unit with 100MbEthernet limitation.
Contact me PM or Telegram or in other thread about details bcs it’s to big OT.
TCP and UDP OpenVPN are both (almost) useless in MKT
There is millions,billions,zillions OpenVPN servers with lzo or lz4-v2
And without password auth, only Cert.
So everyone who want to use MKT to reach these servers, need to call the IT dep and say: I am using MKT, and i have only partly implemented OVPN, please, could you setup your linuxes according to my MKT ?
and no, comparing WG and OVPN is not correct. WG could not replace OVPN as L2 tunnel (ethernet mode)
You forget that there still is a use-case for the classical meaning of a VPN, a virtual private network.
It is e.g. a network between a company and its branches or employees working from home.
Both the server and the clients are under a single management which can decide what to buy to make them compatible.
The transfer rate is usually limited to the uplink speed of each connection, and it does not need to max out the downlink speed, it only need to match the requirements.
Today, many think that VPN always refers to the services that are intended to dodge your identity from law enforcement, media rights enforcement, etc.
But that is not the case and lots of users never use that form of VPN. I think those “newfangles” uses of VPN should use a different name, maybe EPN (enhanced privacy network) or so.
Of course in those cases you take the services from another company who decides on the parameters and when to change them.
Not a safe bet to buy equipment from another source which may not follow those changes.
This is clearly off-topic gone wild, but let me add my 2¢ anyways.
That 16MB flash thing is not only economical, but also technical. If you take a close look on the different RotuerBOARDs you’ll notice that all those 16MB flash devices use SPI Flash chips, whereas devices with a larger amount of flash usually use NAND flash chips. And that’s significant. For IPQ4018 SPI flash is the only available option, whereas IPQ4019 also supports SD/eMMC and ONFI interfaces. But as with a lot of other microcontrollers, some pins are usually shared between different interface, so you have to chose what to use. Have a look, for instance, at hAP ac3 and hAP ac3 LTE6 kit- the former has no mPCIe slots and 128MB NAND flash, whereas the latter has mPCIe slot (with LTE modem preinstalled) but only 16MB SPI flash. Are those related? I don’t know for sure, but I guess they likely are.
For SPI flash, 16MB (128Mbit) is the maximum you can access using a “traditional” 3-byte addressing. I know (I googled) there are IPQ4018 devices on the market with 32MB SPI flash chips. What stops Mikrotik from doing the same? I don’t know. There may be some hardware compromises to consider, or it may require architectural changes, or… I am just inclined to believe that they have a good reasons for that.
no ! @pe1chl
YOU forget that world is not made by MKT
There is plenty other network equipment from other vendors.
So, you think, when company buy few MKT device, it should change all other equipment to MKT to maintain compatibility ?
I am using on my work lots of different network device. And strugling with MKT denial to implement things as it should be
Sorry, but whenever somebody say something against MKT, you shown on the forum, and start advocating, defending MKT
i don’t care what is your personal reason for this behaviour, but it IS a forum, where user says oppinion about MKT gear,
eventualy, try to force MKT to implement stuff to full functionality
and if MKT implement only half of OVPN, only half of IPV6, only half of …
Should i paid HALF of price?
Ok, let it be, to stop this OFFtopic
Tomorrow i am goin in shop, and i will paid only the half. When seller refuse to give me MKT device on half of price, i hope you will be there with same entusiasm defending me against seller, speaking that HALF IS ENOUGH !!!
@npeca75: I think you misunderstood some parts. If for example MikroTik’s OpenVPN implementation is lacking, it’s not good, and I don’t think you’ll find anyone saying otherwise. But it still can be good enough for someone. If it’s simple VPN used by employees to access office from elsewhere and RouterOS is server (I have this in several places), you have all interoperability you need, because every client will be able to connect to it. If it’s the other way around and you need to connect RouterOS client to more capable server, it’s problem and it’s bad. But in first case it’s almost fine. It will be even better with UDP OpenVPN or WG (which in this case is good replacement, because there’s not need for L2). And it’s good that updates bring these new features. Who knows, maybe one day we’ll have full OpenVPN, and everyone will get it, because MikroTik releases updates even for old devices.
Damn straight says the CCR1009 owner!
Of course all the work should be done on TILE first and then ported to all other architectures.
Bets, … which comes first, v7 stable/long term release for my CCR, or a shot of covid vaccine ???
I’m also lucky home user with 200Mbit FTTH, but most people have much less. So someone who is happy with 20Mbit VPN is pretty normal. If they now use TCP OpenVPN with v6, then new v7 with UDP OpenVPN is good example of useful improvement. And WG even better.
