Reply
Member
Posts: 296
Registered: ‎05-09-2013
Kudos: 173
Solutions: 27

Re: Requests/suggestions for new hardware models

[ Edited ]

ERL with one copper/sfp shared port ( i think 3 port is max for CN5020). No switches, no poe in/out. Price as close as possible to current ERL.

Edit: forgot to mention metallic case.

SuperUser
Posts: 3,892
Registered: ‎06-30-2010
Kudos: 1748
Solutions: 167
Contributions: 9

Re: Requests/suggestions for new hardware models

[ Edited ]

Instead of a ER-Pro8 and 2xTS-Pro at our main/edge location and large POP's, a ER-Pro8 with PoE 24/48V out would be nice:


UBNT-ancheng wrote:
  • Number of ports? 8-12 port
  • 100M, 1G, or 10G ports (how many each)? 1G
  • Copper (RJ45) or SFP/SFP+ (how many each)? 8-12 RJ45's + 2x SFP/SFP+
  • Form factor (e.g., rackmount or small like ER Lite)? Rackmount
  • Power supply (e.g., AC vs. DC, dual redundant, etc.)? Dual AC HotSwap
  • Is PoE input (use PoE to power the router) needed? No
  • Is PoE output needed? If so, what kinds (24V vs. 48V, passive vs. auto-sensing) and how many ports? Yes, 24/48V with auto-sensing on all ports.
  • "Real" ports or "emulated/switched" ports? For example, all three ports on ER Lite are "real" (1 Gbps each to CPU), but the last three ports on the ER PoE are "emulated/switched", i.e., they are connected to a switch which is then connected to the CPU with 1 Gbps. For a 12 Port, 2x5port Switches and 2 independent ports on RJ45's.
  • In addition to the network interfaces, any peripheral requirements, for example, serial console port (or not), external USB ports, minimum memory size, etc. etc. 1: A serial to USB internal, to have a USB console port. 2: 2xUSB port for UPS monitoring and such.
  • A deployment scenario (e.g., with diagram etc.) would be really helpful! At our core location where we have 1 ER-Pro8, 2x TS-Pro's, CCR1036 and fiber, a HA redundant system with above mention specs. would be very good.
  • Expected price? No clue. approx 500-1000$.
  • Temp. and Current sensors

For End-User install's:

Instead of a ER-Pro8 and 2xTS-Pro at our main/edge location, a ER-Pro8 with PoE 24/48V out would be nice.


UBNT-ancheng wrote:
  • Number of ports? 5 Port
  • 100M, 1G, or 10G ports (how many each)? 1G
  • Copper (RJ45) or SFP/SFP+ (how many each)? RJ45
  • Form factor (e.g., rackmount or small like ER Lite)? ERL/ERPoE size
  • Power supply (e.g., AC vs. DC, dual redundant, etc.)? DC or AC
  • Is PoE input (use PoE to power the router) needed? No
  • Is PoE output needed? If so, what kinds (24V vs. 48V, passive vs. auto-sensing) and how many ports? 1 Port passive 24V, for UBNT radio (WAN).
  • "Real" ports or "emulated/switched" ports? For example, all three ports on ER Lite are "real" (1 Gbps each to CPU), but the last three ports on the ER PoE are "emulated/switched", i.e., they are connected to a switch which is then connected to the CPU with 1 Gbps. 2 Ports + 1x 3port switch like ER-PoE
  • In addition to the network interfaces, any peripheral requirements, for example, serial console port (or not), external USB ports, minimum memory size, etc. etc. 1: A serial to USB internal, to have a USB console port. 2: 2xUSB for NAS and/or 3G redundant.
  • A deployment scenario (e.g., with diagram etc.) would be really helpful! End-User installation like the RB951G-2HnD with PoE out.
  • Expected price? I guess 100-200$.
  • Wifi integrated or option via modul (miniPCI).

 /Paetur

Member
Posts: 142
Registered: ‎09-19-2009
Kudos: 30
Solutions: 3

Re: Requests/suggestions for new hardware models

[ Edited ]

We have quite a lot of power outages here. Therefore our sites are all DC only, for longer battery runtime.


We have been having to open up the 8 port toughswitches and remove the power supply in order to run them directly off the DC and have something that does POE. We use this in combination with an ERL and vlans in order to do OSPF to neighboring towers.


What we need is a small form factor 8+ ROUTED gigabit PoE ports, DC powered, with the capability to move ports into a switch group for the access points, so that you don't have to have a seperate IP block for each AP. I would suggest 48v DC with the ability to select between 24v for airmax or 48v for airfiber.


It would be nice to have the option of an SFP port in case we have a home at the base of the tower, and provide service on fiber to isolate it electrically from tower strikes. But not if it puts the price up anything significantly, we can just use a media converter in those few cases.


Previous post...

http://community.ubnt.com/t5/EdgeMAX/3-ports/m-p/328860

 

Member
Posts: 269
Registered: ‎03-19-2014
Kudos: 21
Solutions: 1

Re: Requests/suggestions for new hardware models

[ Edited ]

either 1x SFP+ or 1x 10GBase-T should be fine for (hypotetical) ERL5 as upstream port cause you always can use any kinf of media-converters for copper 10Gb port for fiber, unless you had to go for 40G or 100G, which is out question anyway, cause CPU limitations.

about power - would be nice to have full-range DC port in it, ie which can eat power from 12v-48v range, for example.

Established Member
Posts: 1,620
Registered: ‎02-24-2012
Kudos: 401
Solutions: 60

Re: Requests/suggestions for new hardware models

1 or 2 x SFP would be nice, or like ProCurve does, you get an RJ45 port and an SFP port below/next to it, and you can use either or, not both at the same time, which I assume reduces cost.

Auto-detecting PoE Ports - is that possible?

Earthing it pretty important - I have an EdgeRouter PoE which has the earthing lug on the chassis, but I don't recall if the power supply is two or three pin (with or without earth).

 

Veteran Member
Posts: 5,044
Registered: ‎03-12-2011
Kudos: 2495
Solutions: 120

Re: Requests/suggestions for new hardware models

After reading some of the suggestions here with some wanting a switch chip, some wanting all routed interfaces, etc - a thought I had is why not both?

All ports can be switched at layer2 line rates (presumably by a layer2 switch chip), but have a sufficiently fast interconnect back to the CPU so that it's not a bottleneck (ie, 10x gige ports, with 10gige back to the cpu). I'm not sure if there's any chips out there that can achieve this, but if there is it would increase the flexibility a lot.

Of course the other option is to do layer3 routing on a layer3-capable switch chip, but I imagine layer3-capable switch chips probably can't do fancy things like firewalling, nat, which would mean most of the featureset of EdgeOS is useless - but again, I don't know what chips are out there, so this may be a possiblity.

The other option of course is going with the current architecture of dedicated lines back to the cpu, but having a soc that can do line rate layer2 switching as well (heck, maybe the current ones are capable of this given some resources to implement it in the offload module?).

The requirement for layer2 line rate (which currently often means a ToughSwitch or similar is required) is that you usually have more than one PtMP AP at a site, which at least in my setup means a shared layer2 across them. Also for the management network although speed isn't really important for that. At the same time of course most sites have more than one PtP link too, which are (again, in my setup) on their own layer2 per link which makes having their own proper routed port useful.

Member
Posts: 269
Registered: ‎03-19-2014
Kudos: 21
Solutions: 1

Re: Requests/suggestions for new hardware models

[ Edited ]

no, switching mess - way to Hell.

either that was MikroTik nightmare or other companies, monsters.

you may had smart Switch OR you can have Router. last category("EdgeMax" products)currently was case and topic of discussion in terms of improvements, welcomed by consumers.

of course you may had different opinon about "optimal" design of gear, but lucky for you "sufficiently fast interconnect" - perfectly described Cavium chip itself aswell as offloading/accelerating feats, among other benefits/caps, so basically no need to "invent weel" - buy EdgeRouterPro today and tommorow - Ubiquiti send you teacup, "mFi" t-shirt and "AirMax" hat !! hurry ! limited offer, unlimited performance !!

p.s.

you're just snatched new "EdgeMax - Doing Fancy, cool thing ! faster than Ever !*barrel roll" advertisement, thats devoted both to EdgeMax hardware and EdgeOs, backing/driving it.

New Member
Posts: 6
Registered: ‎03-28-2014
Kudos: 1

Re: Requests/suggestions for new hardware models

We wouldnt mind seeing a version of the EdgeRouter Pro with 10G ports. Dual PSU's with C14 ports would also be nice to see for a router that is rack-mountable.

Member
Posts: 269
Registered: ‎03-19-2014
Kudos: 21
Solutions: 1

Re: Requests/suggestions for new hardware models

[ Edited ]

yeah, dual-power, 24-ports, all-10Gbase-T copper ports beast would be cool. sadly processors for such stuff - wasn't created, yet Man Sad

even mainstream intel celeron J1900 ? or G-series from AMD  and full-blown desktop nor server chips - can't. even their chipsets, itself - cannot back even pair of such links.

my original suggestion about five ports ER model with 10Gb uplink port and wire-speed routing on it or all 1Gb as ubq/simple device for small companies and home users without huge networks nor complex needs, but with fat traffic and supscriptions plan/uplink.

Member
Posts: 276
Registered: ‎11-16-2013
Kudos: 104
Solutions: 15

Re: Requests/suggestions for new hardware models


Basiley wrote:

yeah, dual-power, 24-ports, all-10Gbase-T copper ports beast would be cool. sadly processors for such stuff - wasn't created, yet Man Sad

even mainstream intel celeron J1900 ? or G-series from AMD  and full-blown desktop nor server chips - can't. even their chipsets, itself - cannot back even pair of such links.

my original suggestion about five ports ER model with 10Gb uplink port and wire-speed routing on it or all 1Gb as ubq/simple device for small companies and home users without huge networks nor complex needs, but with fat traffic and supscriptions plan/uplink.


Why not created? 

OCTEON III CN7XXX Multi-Core MIPS64 Processors

SuperUser
Posts: 21,755
Registered: ‎11-20-2011
Kudos: 7878
Solutions: 233

Re: Requests/suggestions for new hardware models


Basiley wrote:

yeah, dual-power, 24-ports, all-10Gbase-T copper ports beast would be cool. sadly processors for such stuff - wasn't created, yet Man Sad

even mainstream intel celeron J1900 ? or G-series from AMD  and full-blown desktop nor server chips - can't. even their chipsets, itself - cannot back even pair of such links.

my original suggestion about five ports ER model with 10Gb uplink port and wire-speed routing on it or all 1Gb as ubq/simple device for small companies and home users without huge networks nor complex needs, but with fat traffic and supscriptions plan/uplink.


x86 chips and custom built ASICs and FPGAs are a completely different beast.



isp builder | linux sorcerer | datacenter automation conjurer | blog: blog.engineered.online
link to our slack channel on the blog
Member
Posts: 269
Registered: ‎03-19-2014
Kudos: 21
Solutions: 1

Re: Requests/suggestions for new hardware models

[ Edited ]

multiply 24x copper ports to 10Gb port resource demands in each. thats would be NASTY/TERRIFIC load.

i don't know if top-dog Octeon III chips(say 8x or 16x core ones or with further widened/fat offloading, L2 and interconnect)can Really handle that, but seriously doubt about

their top-dog model - 48-core chip. present chips can drive about 1-2Gb/per core, so to drive 24x 10Gbports they need 100-200-core chip ? :/ which sounds no-practical.

for updated Octeon III they advertised up to 3Gbps/per core, but thats anyway, ruled-out anything bigger than 8x-12x -ports 10Gb product, right ?

but portions of description, really attract me, like "up to 120GHz of 64 bit processing" description =)

even pair of 8x core Xeon or Opteron chips  - probably wouldn't drive 24x 10GBASE-T ports, even moderately-saturated.

about about x86. Silvermont (for example J18xx, J1900 and J24xx models) intel chips and R-series AMD chips(for example 5350 Fs1b/AM1 CPU) - quite popular both among homemade twin-ethernet, backed by PCI/PCIe-based(with smart swiches)downling and mPCIe(usually for WiFi), routers and SDN/OpenFlow boxes, offered by some companies, nowadays, but MANY 10Gbase-T ports - imply WAY more serious processing power, necessairy/critical to properly process/route/control all that stuff.

basically its major 10Gb(either copper nor fiber)holding factor - lack of fast, cold and cheap chips to run products on.

there was plently of good Cat6a and Cat7, relatively easy to deploy(esp if you had STP experience nor IBM-related past)cables, good customers expectations, decent amount of traffics, ready to enjoy/flow all that, thru, just lack ... networking gear(in all portons of market - from SOHO to corporate). presented stuff funny overpriced, dangerously hot, sometimes and ridiculously outdated. ASIC usage in 10Gb switches slightly tone-down stuff, but WHERE is Routers ? and they remain really Hot. heck, even two first 10Gb interfaces/PHY generations, itself - really too warm.

p.s.

sorry for my probably naive calulations and probably slightly-irrelevant side-note at end.

bottom line: its would be groundbreaking, market-shaping product, if so, but probably would be bit more safer -start from something smaller, say 8x or 12x all-10Gb product, for example. also termal emission and power consumption of biggest OcteonIII chips - probably would concern designers, engineering end/complete products, based on it.

but they anyway - Way better than Tile chips, nor Freescale e5500-based chips. only probably OpenPower-derived Power8-based many-core chips/SoC can provide some Really challenging/competive numbers.

SuperUser
Posts: 21,755
Registered: ‎11-20-2011
Kudos: 7878
Solutions: 233

Re: Requests/suggestions for new hardware models

[ Edited ]

*sigh*

http://netgear.com/landing/10gigabit.aspx#tab-10gbefiberswitchoptions

http://datainterfaces.com/ecs5510-24s.aspx

http://h17007.www1.hp.com/us/en/networking/products/switches/HP_5820_Switch_Series/index.aspx#.U0Xp0...

These are the low end. They you get up to the large chassis-based switches from Brocade, Juniper, etc. Point is, there are TONS of large port count (192 ports and above) 10Gb/40Gb/100Gb Fiber switches out there.

They're using ASICs, not x86-type chipsets, or SoC designs (for the most part).



isp builder | linux sorcerer | datacenter automation conjurer | blog: blog.engineered.online
link to our slack channel on the blog
Member
Posts: 269
Registered: ‎03-19-2014
Kudos: 21
Solutions: 1

Re: Requests/suggestions for new hardware models

[ Edited ]

oh no !! no ASIC/Switches+SoC overclomplicated/incoherent mess Man Sad

they can work, for sure, but at what price ?(aside pricing, itself, btw) Man Sad

i strongly beleive (for number of reasons) thru experience in CPU-driven routers design not in "partially managed" multi-chip monstrocities.

and for end-users market such things are out of question not only cause pricing.

10Gb switches was made some time ago(and thats commented above a bit), but where is 10Gb routers ?

and switches itself - predictibly costly(cross-out about 4/5 potential 10Gb adoption/penetration), "thanks" to ASIC in design, partially and still hot.

if i was forced to multi-chip mess choice - i will pick multi-CPU x86 SDN boxes instead. most of them slower in L2 10Gb drive, but complete L7 Routers with fully-tranparent and manageable neflow(and most SDN brands finish their OpenFlow transition within this year).

but my original  message, from which this started - actually answered to "why not both?" question about practical benefits from attempts to combine "wire-speed L2 chips" with router-specific CPU/SoC, which is popular among some vendors stuff. i don't think it provide benefits of both elements/worlds/approaches, more than it pack drawbacks/flaws of them, which is in my opinion, im frequently see in gear(made such way)im had to manage both at work and in home(nor friends/society setup).

anyway, regardless my possible childish and pessimilstic suggestions and calculations - someone suddnly created SOHO-targeted small/mid-port capacity, affordable, reliable and easy to manage products with 10Gb ports - it would be BIG THING. both for consumers, vendor and market development.

Veteran Member
Posts: 5,044
Registered: ‎03-12-2011
Kudos: 2495
Solutions: 120

Re: Requests/suggestions for new hardware models

Something something layer2 != layer3?

Member
Posts: 269
Registered: ‎03-19-2014
Kudos: 21
Solutions: 1

Re: Requests/suggestions for new hardware models

[ Edited ]

i see you point, despite my attempts to critize it(from experience far from flawless with such stuff/design).

so "switched at layer2 line rates (presumably by a layer2 switch chip), but have a sufficiently fast interconnect back to the CPU so that it's not a bottleneck" sounds like based on common sense, but they(switch and CPU) more frequently interfere with each other, than help. CPU lack control in result and switch chip frequently interfered by CPU to unleash potential. thats writen, ironically thru device, based on AR9344 SoC which include, aside WiFi, switch chip to save on dedicated PHY's during production and resulted performance nor consistency of netflow - far from perfect :=/

others may correct me, but rarely such things as "l3 switch chips" used, its more frequently "L2 wirespeed" switch with fast, but naive/simple controller/SoC, atop it, that do mangement above L2, frequently within same package/die.

thats would be interesting otherwise, but definitely too expensive kind of ASIC, in my, strictly personal opnion.

Cavium creators kept in mind that, likely, judging from what they packed in.

switches and routers remain separate beasts in result. until switches extinct, in which im firmly beleive.

p.s.

i wish Ubnt folks not eaten all their pop-corn, reading us Man Happy

and nobody hurt nor offended during viewing that. atleast i can say such about my part.

New Member
Posts: 2
Registered: ‎01-26-2014
Kudos: 2

Re: Requests/suggestions for new hardware models

I would really like to have a version of the ERLite-3 router with one of the ethernet ports replaced by a VDSL/ADSL interface.

While ethernet works for pretty much everything else these days, including VDSL/ADSL services using PPPoE bridged through a modem, there are other DSL framing formats in fairly common use which either can't be bridged or are difficult to bridge to ethernet modem-style (the products which try to handle these that way generally don't work well).  These include PPPoA, which a few US DSL providers use but is much more popular in, e.g. the UK, and RFC 1483 LLC encapsulation, which is often what you get in the US if you order your DSL service with static IP addresses.  Here's an article which lists who uses these encapsulations in the US:

    http://www.hiroinc.com/article.php?article_id=34

I've had to find routers for RFC 1483 on a number of DSL circuits ordered from AT&T, and have had to find something to terminate PPPoA in London.  There aren't a lot of good choices for these (to say the least).

I would pay $100 over the price of an ERL for one of these.  Even for PPPoE having the modem built into a router you like sometimes saves the cost of an external modem, and it makes (remote) management of the service much easier if you can see line stats from the router directly.  For non-PPPoE encapsulations having the modem built into the router becomes essential.  In fact I'd pay more than a $100 uplift if the VDSL2 chipset were compatible with Lucent Stinger concentrators since these are still used a lot in older VDSL deployments, but implement a slightly non-standard VDSL for which there are very few compatible products (all of which are awful).

Dennis Ferguson

Regular Member
Posts: 493
Registered: ‎02-02-2008
Kudos: 83
Solutions: 5

Re: Requests/suggestions for new hardware models

Dual AC version of the ER PRO.

Dual DC version of the ER PRO. (ISOLATED +48VDC)

Prefer both with hot swap power supplies.  Really you only need one version with two different power supplies you can slide in but I really prefer dual power supplies on anything "core". 

Speaking of which, has anyone hacked an ER PRO to DC yet?  

New Member
Posts: 35
Registered: ‎05-05-2011
Kudos: 20

Re: Requests/suggestions for new hardware models

[ Edited ]

I have been following the edgemax product line for a little bit now and am anticipating the carrier version of the ER. I will explain our current setup and talk about my criteria for a drop-in replacement for what we have currently.

Current Environment:

  1. Full Cabinet colocated in Dallas
  2. Redundant Power
  3. Redundant 1GIG BGP Uplinks - Will be upgrading to 10G in the next 6 months.
  4. Two poweredge servers running Vyatta. (We have been runnning vyatta for nearly 6 years now IIRC)
  5. Two 48 port Cisco 2960 switches
  6. iBGP between the two vyatta servers

 

What I am looking for:

  1. 8 Ports are fine
  2. We would have to have a minimum of 4 ports @ 10GB. (Uplink, switch1, switch2, iBGP).
  3. Copper.
  4. 1U Rackmount
  5. Dual AC (one for each PDU) - Someone mentioned hotswappable.. That would be a nice feature.
  6. No
  7. No
  8. Real ports. No switches.. UBNT needs to offer a "carrier" grade solution and switches and emulated ports are no way to do that.
  9. Serial console and USB
  10. High-Availability Network
  11. No clue on price since I dont know the cost to produce the above.. Given UBNT history it will be shocking I am sure!
  12. More CPU and Memory
  13. Config Sync with exclusions (interfaces, certain firewall groups, etc)
  14. Wishful thinking maybe but a REST interface to update the config would be nice. Our log collection server could push ip addresses to a firewall group when certain patterns are matched. Our auto-provisioning service could push new IP addresses into accepted firewall groups. The above would be nice over a REST interface..

As far as the UI goes for this deployment scenario it isnt important we could care less. I think a good approach for UBNT is to build out the features, let those that dont require a UI adobt them, battle test them and let the UI come behind once the functionality has been ironed out. With that said we recommend a LOT of ERL to our customers across the country and to them the UI is important for the SMB application.

There are a lot of naysayers out there that cant believe we would even consider something outside the norm or that we have been running Vyatta. I was speaking to someone the other day that manages data center build outs and was telling him about the ER. His first response was "I sure hope you arent running a $350 device at the edge of your network!"  I explained to him that we ripped out all of our Cisco equipment to put the vyatta in years ago. We push a ton of small packets and havent had an issue yet. I believe that if the above can be produced and the system proven to be reliable it could be a game changer in the industry.

 

Veteran Member
Posts: 5,044
Registered: ‎03-12-2011
Kudos: 2495
Solutions: 120

Re: Requests/suggestions for new hardware models

Possibly more in the "ToughSwitch" category than EdgeMAX, although I don't see much happening in that camp at the moment...

But, pretty much this http://routerboard.com/CRS226-24G-2SplusIN with PoE would be amazing. Combine with an EdgeRouter Carrier that has a couple of SFP+ ports on it as well and you've got yourself a pretty kickass combination. (NB: Layer2 is probably sufficient, especially if paired with a new EdgeMAX Carrier`with SFP+'s`- I'm not sure if that product is actually a layer3 switch, or if it's just a layer2 switch with a small router bolted on and probably maxes out at a couple of hundred mbit - aka useless...)

Reply