Problem:1 cable modem 150/20 mbps & 1 fiber line 5/5mbps how to distribute best

[questorfla]

Situation is a stated.  We have those two separate sources.  The fiber is where they threw us off.  New install, new building.  we are first fiber user. They put in a  24 port capable switch at their demarcation point,  We were the ONLY users.  Somebody forgot to cap the ports.
We got 5mbps for our phone system and .... well it looks like ALL the rest for our Fiber switch. This was a regular off-label 8 port unmanaged gigabit switch and all5 systems connected to it shows 100+ on multiple tests at multiple times of day.  needless to say, we had NO bandwidth issues

Until recently,  The problem was that I have always used OPEN DNS or GOOGLE DNS as they are always MANY times faster than COX's DNS (They are our provider)

A week ago we started getting all kinds of DNS issues.  somewhere back up the line cox DNS link had blown out a ram module and after than our dns ran to nothing. 

<< ADDED:  We have also been informed that NO MORE Freedom of DNS.  EVERY system MUST USE ONY COX DNS!! No OpenDNS  No Google even though they are MUCH faster!

No options!!.  If I try ANYTHING but the COX DNS it  I get windows error saying unable to contact dns server.  I was told it was a "security thing" but I have not heard from anywhere as to what this is or why

WHen they came in to check, they found the "error" and capped us back to 5gbps.  Unfortunately by them we ha offloaded a bunch of our files to cloud access.

20 workstations using cloud storage instead of a local server was convenient and worked fine at 120/120mbps+ but not so good at 5mbps.
The upgrade to fiver to PAY for 120 was MEGAbux,  But we could get 150/20 CABLE and did.

However:  the same feed into the same switch ended up at 5gbps on the systems after running through the same switches that had been running fine on the fiber.
Cable tech blames the switch and shows me where it drops off after the first Gigabit switch as though it was only a 100MBIT switch.  He says it is due to it being an "Unmanaged" switch.

Even though it worked fine on Fiber,  Wont work right on cable.
However.  If I run the cable into a gigabit Router with combined gigabit switch.  each output feed DOES get full 150.  If I add switches after router, it STAYS at 150.

The problem is, I have to provide this input to systems which normally pulled from the switch after the cable modem so I could have the static IP's for each system.

If I run them through router, I am confused about the ability to have separate IPS on each device,  These are small webservers and must have static IP plus have
required ports passed through to be a webserver.

I will try to draw up a ma and upload it but I am still a bit fuzzy on how I can get full bandwidth through a router but drop to 5mbps through a switch.

Any Info on this appreciated.  I need to know the best way to split up the output from a cable modem so hat all devices show close to max as possible.
What combination and what layout would do that.

Next part is  one of the static GOES to the office router which feeds maybe 30 workstations.

in my original setup, ALL showed almost full speed in every speed test .  Now all show 5mbps.

Is anyone out there well schooled enough in this to tell me if I am getting a load of bs?
And if not, how can I fix?

[40hz]

Only have time for a quick reply right now.

Regarding the DNS lock-in:

It's not that common - but it's also not unheard of for a Cable or Telco ISP to want you to use their DNS servers. If that's their new rule for which they won't grant any exceptions - and you don't want to switch ISPs - you're out of luck. If it becomes a real hassle you can always install a proxy server (like Squid) in your office to cache DNS queries and provide some speed up there.

FWIW, not letting you use your own choice of DNS servers in the name of their network's security is a bit of a stretch. I'm guessing it's more for your ISP to gather data on browsing patterns, block certain queries or sites, cap bandwidth to things like Netflix (and so much for "net neutrality" right?) or for something similar they've instituted that new rule. Your choice of DNS servers can't (well ok...shouldn't) really do anything to jeopardize their network's security. Even if you hit a DNS server that's been 'poisoned.' So nope...the official 'reason' they're giving you doesn't pass muster in my book.



-----------------------------------

Addendum: when you do show us your network map, also include the make/model of the routers and switches you're using? That would be very helpful to know.

[Stoic Joker]

However:  the same feed into the same switch ended up at 5gbps on the systems after running through the same switches that had been running fine on the fiber.
Cable tech blames the switch and shows me where it drops off after the first Gigabit switch as though it was only a 100MBIT switch.  He says it is due to it being an "Unmanaged" switch.

Thanks for that, I like to start my morning with a good laugh. A switch being managed or unmanaged has absolutely nothing to do with it's backbone speed. Managed switches just offer one the ability to segment traffic to mitigate congestion from spreading so a few bad apples can't take down the entire broadcast zone.

Chances are that the switch is getting feed a network frame type (packet size if you will) that it can't handle causing it to badly fragment the traffic flowing across it.

Even though it worked fine on Fiber,  Wont work right on cable.
However.  If I run the cable into a gigabit Router with combined gigabit switch.  each output feed DOES get full 150.

As above theory suggested, this says jumbo packet size definition mismatch to me ... That sound about right to you too 40hz?


On a side note, if you have two ISP's and a DNS preference (I'm an OpenDNS fan too), bring both providers together into a dual WAN router and then configure the router to route all the DNS requests out through the one that isn't being an ass.

[Shades]

Malfunctioning port on the switch could be a problem. If you were using the WAN port of the switch, it could be capped, so try another port. Try a different cable (following a different route) might help as well. Without knowing how your LAN looks, that would be things to look at.  A year or so ago I could lay my hands on a very affordable Zyxel 24-port switch that included basic managing options. Works really nice.

As suggested in another post, get a good PC with old single-core specs and install Untangle on it. You have fiber and cable coming in, so get 3x 1GBit NIC's into that PC. 2 for for incoming signals and one for your LAN. Feed the LAN output cable into a normal port of your switch and use the switch to connect all workstations.

Configure DHCP on the Untangle box so every workstation will receive their static IP address from the Untangle DHCP server, no matter where they connect their PC. It allows you a lot more control over which workstations/devices can be connected, how much bandwidth can be used by them, which provider has preference, traffic management etc. This solves more than one problem you have. A system like this is also easier to maintain and keep redundant parts for in case of an emergency.

[Stoic Joker]

...WAN port on a switch?

[40hz]

As above theory suggested, this says jumbo packet size definition mismatch to me ... That sound about right to you too 40hz?

Yup! That was my first thought. Which is why I wanted to see the equipment list in order to get an idea if that might be the case, and more importantly, if it was fixable without needing other hardware.

But Shades raised another interesting point. If the ISP supplied router is capping them on the WAN side that's another possibility. Kinda like business grade DSL service around where I am. Most circuits can provide the full bore. But the ISP sets the rate in their locked-down router so you only get what you're paying for. Which is why upgrading to the next level "asymetric" usually requires nothing more than a phone call and maybe a credit check. One hour later and presto you now have a faster pipeline courtesy of the tribbles over at central office who went in, changed a setting, and bounced your router.

The giveaway would be whether or not the client has admin access to the router. If they don't - and I'm guessing they don't - there's a real chance The ISP is capping them. And the fact it got "fixed" so easily tells me it was a local router setting that was misconfigured. (Some providers set you up initially at the next highest rate to give you a taste of what the deluxe service level feels like. It usually reverts to what you've contracted for after a week or two if you don't take advantage of the "special offer" upgrade. Maybe, in this case, they forgot to drop it back down?

So anyway yeah...frame mismatch or capping in effect. Those would be my two best guesses.

Regarding Untangle NG:

I think Shades made an excellent recommendation. I like Untangle too. I have a few small SOHO clients plus my two sister's houses (one with a home-based business) running off Untangle. And it's a dream to set-up and administer. What's really nice is you can see into everything it's doing - so if you ever need to troubleshoot, everything from your WAN port inward is visible and configurable. And it supports web caching and bandwidth control - so that should help with web access. Plus there's built-in WAN balancing and failover - so if they do end up keeping two ISPs, those features can provide some interesting possibilities.

[40hz]

...WAN port on a switch?

Yeah. That's why I wanted the equipment list. There may be some nomenclature issues (i.e. switch/router/combo) we'll need to get straightened out too.  

BTW - I had a chuckle and an eye-roll (Oh puh-leez!!!) over that managed vs unmanaged switch "explanation" too. I guess that's another one of those "little known facts" you hear about.

[Shades]

There are several switches here (1x Netgear, 5x 3com, 1x Zyxel and 3x no-names) and all, except one have a WAN port in them. None of them mention that they are a router/switch combo. But I have never tried them either, so I couldn't tell. They came with a pile of other stuff, bought at the US embassy auction years ago. IMSM that stuff (2 complete servers, 2 incomplete IBM PC's, modems, cables, NICs, software, UPSes etc.) didn't cost much more than a 1.000.000Gs (around 150 USD in those days) and one of those servers was a PII Proliant.

Anyway, most of the switches are around 15 years old by now, 10/100MBit and come with a coax connector as well. Come to think of it, were those (incorrectly labeled) WAN ports not used to string multiple switches together back in those old days? With cross-cables? Sorry, too lazy to look it up myself.

Ah well, it's easy to make a mess of a network and keeping it as simple as possible will save your sanity and makes it much more manageable and has usually a longer up-time as well. Inheriting a network from some companies could be considered damn near capital punishment. 

[Stoic Joker]

Come to think of it, were those (incorrectly labeled) WAN ports not used to string multiple switches together back in those old days? With cross-cables?

That's really quite bizarre ... AFAIK those have always been called uplink ports. Translation issue? Admin being funny? *Shrug* If you can post a model number I'll look it up to see what it are.

[40hz]

Anyway, most of the switches are around 15 years old by now, 10/100MBit and come with a coax connector as well

Some old 3COMs used to have special matrix-type ports and cables to link the backbones of multiple hubs and switches together. That gave them better throughput between the switches than running a standard crossover cable because the network ports were mostly just 10Mbs on the boxes that had that feature. We're talking late 80s/ early 90s vintage on those. I don't recall seeing any with coax connectors though.

[attachthumb=#]     [attachthumb=#]

I'm guessing that was probably for twinax cables - which are shielded twisted pair although they look very similar to coax. IBM was big on that type of wiring for their ARCNET network topology. Their 52xx series terminals (usually connected to a System36 or System38 minicomputer) all used twinax cables. Early Ethernet did too IIRC.

Here's a '36' - sometimes called "the world's littlest dinosaur." It was about the size of a two drawer filing cabinet. It's preferred language was RPG-II - although you could also license COBOL, FORTRAN and BASIC for it. Love that 8" floppy drive!

[attachthumb=#]   [attachthumb=#]

Ah...those were the days! (NOT!!!)

[questorfla]

WOW!  Looks like I opened a can of worms.  Here is as much detail as I have this time of night because right in the MIDDLE of all this... God Help us...
they decided to jump on that Godaddy /Microsoft combo hosted 365 exchange  !!:(
I can say this .  IF???>>> they ever get ther bugs worked out,  This is one amazing commo package.  I am not a "frills tech".  I only care a bout reliability and speed.

I can give you some Honest triple checked times that border on the impossible.

You can open Your email app on computer.  Send a TEXT as a email to a cellphone in another state.  Have the other party one the phone when you do this.
If possible, have the volume set to high so both partied can hear the "ding" on each others phones.

As soon as you hit send.  and I mean as SOON AS, like maybe 2 seconds.  the party receiving the email in their cell phone as TEXT even (ie: sent to their phone number.txt.att.net (if using at&t) or similar for others.  Total time from you hit send to they get test.  3 seconds!  Maybe 5 if a bad time of day.  we have never seen more then 5 seconds.

Same test email to email.  Same results.  Same test phone to phone, same results.  This is the FASTEST linking I have ever seen and if someone out there knows how they do it I sure would like to find out.  In theory... sure,  Speed of light and all that,  But in FACT there is no straight path and the routing and overhead make this totally impossible.

It would make a great magic trick in Vegas!

WE tested against normal simple IM phone to phone and this beat it by a fair amount.

I still have not figured out the Go Daddy+ Microsoft Partnership it was the most out of the blue link I could imagine.

Anyway.  The Router layout was explained to me like this:

The COX modem supplies a full duplex signal which cannot be handled by "ANY" modern unmanaged switch.  He said the switch would only allow the single sided connection which would drop the speed to 100Mbit.  Those were HIS terms exactly.
He suggested i try a "better switch/  Fortunately, I already had.  Not just one but several.
ASUS, DLINK, CISCO, NETGEAR, and a few White label that are ALL marked as gigabit "self-managed" leaving any one port to be input with the rest as out put.

Not only did NOT of them work, they ALL has the exact same symptom.  The GREEN h-speed connect on the COX modem would ALWAYS be orange reflecting (to him at least0 that it was out "bad switches"
I have used ALL of these on mhy Mediacom Modem and never had one what forced the modem into a low speed state.
He the replaced whatever we had with a Motorola Surfboard and it was no different.

Right now, i order to give the office the speed it needs, I have the min office router running direct off the modem
I can then daisy-chain ALL the switches if i want to.  I tried out of modem to #1,oput of #1 to #2, then to #3, then to #4, THGEN finally to a system.,  Speed test:  139down+20 up.  Even with 15 other system plugged into various ports.

These are all 16 port switches.  the one at the END was feeding the one i ran the test on.

So.. i am baffled.  I have had a similar setup running for years on bonded T1's but I admit we were always getting the BENEFIT from wide open Intranet,  this moving of our largest database to the cloud is what is doing all this.  When I hosted the database in out office, NO problems (of course) since all equipment rated GB speeds.

We wont be getting GB on any kind of cloud connect but...
The REASON for the CLOUD is another post I had a while back about port 445.  This is blocked by the ISP's for outbound from their user on HOME internet service,  NOT blocked on Business service,  The reason id listed as being done to protect the uneducated masses from various "worms".  Apparently they don't care if businesses are protracted or not and besides it is OUTBOUND they block anyway,  Sounds backward to me.

But:  no port 445 both ways, no drive mapping,  SAMBA requires it.  Server 2008 (both R1 and R2) as well as any newer Ines all need that port to be open to map.  If I could map the drives direct for users outside the building,  we never had a problem. in all the years we ran it.  IF the users PAY for Business they are unblocked and they can map from home just as before.

I have researched everywhere I know to find an alternative but.. No luck  VPN's wont help because the blocking is done in the users home cable modem.  They can connect perfect with the VPN and even run the SQL database program (which is on a higher port).  If I could somehow squeeze in access to allow a mapped drive on another port, we would be fine.

Only alternative is CLOUD, cloud requires ALL use it (office and home) or else i have to keep literally hundreds of files per hour in syncing two way in a 50,000 file folder,  THAT is why we need all that bandwidth.

I am wide open to any suggestions and have tried some suggested by others but nothing holds a candle to direct mapped drives on a vpn.  WEBDAV wont do it, the ZCLOUD services are one more complexity and one more app to fail.

I will keep my eye on this thread and hope someone has found a loophole to get past this port block.

[Stoic Joker]

The COX modem supplies a full duplex signal which cannot be handled by "ANY" modern unmanaged switch.  He said the switch would only allow the single sided connection which would drop the speed to 100Mbit.  Those were HIS terms exactly.

I can't even make words to respond to this with.

Right now, i order to give the office the speed it needs, I have the min office router running direct off the modem
I can then daisy-chain ALL the switches if i want to.  I tried out of modem to #1,oput of #1 to #2, then to #3, then to #4, THGEN finally to a system.,  Speed test:  139down+20 up.  Even with 15 other system plugged into various ports.

hub and spoke would probably be safer performance wise than a linear daisy chain to avoid cascading the traffic through a series of choke points across the switches, because the backbone speed of a switch will be faster than going port to port to port..

But - I should mention we also have Cox here (Not. My. Idea...) - my preferred configuration especially with cable companies is to bridge their device to an internal main router that can actually be controlled properly. So if that part of the configuration is working for you I'd stick with it (just make sure their box is bridged).

So our config (with Gb link all the way through) is Web-->Bridged Cox Box-->WatchGuard XTM26-->DMZ with public access servers in a CIDR /16-->WatchGuard XTM26-->LAN with users and stuff


Now if you Cox box isn't bridged...that could easily be half the problem.


Did I mention bridging?

[40hz]

The COX modem supplies a full duplex signal which cannot be handled by "ANY" modern unmanaged switch.  He said the switch would only allow the single sided connection which would drop the speed to 100Mbit.  Those were HIS terms exactly.

I can't even make words to respond to this with.

Man! When it comes to that cable 'tech', the problem's not how much he doesn't know. The problem's how much of what he DOES 'know' is completely wrong.  

[40hz]

they decided to jump on that Godaddy /Microsoft combo hosted 365 exchange  !!:(
I can say this .  IF???>>> they ever get ther bugs worked out,  This is one amazing commo package.  I am not a "frills tech".  I only care a bout reliability and speed.

Can't speak for GoDaddy's offering, but I have clients that subscribed directly to the full MS Office365 package which they're accessing through a "business grade" AT&T DSL connection, and the results are exceptionally impressive. At least to my eyes.

The critical factor is having a fast and reliable web connection - and a properly set up LAN. Since it's cloud, success is all in the plumbing. So it behoves you to pay extra for a higher bandwidth data plan and get some decent quality (i.e. non-consumer grade) networking hardware in place. You definitely want decent routers and switches if you're committing to a cloud solution for your business.

That said, one of my clients is using the usual Linksys cruft you can pick up at any Staples or OfficeMax - and it works just fine for their three person office. YMMV.

About the only hassle I ran into getting them set up was handling the cut-over for e-mail and DNS. That can get a little tricky if you don't understand what goes down when you do that. However, even if you don't, you can still manage to transition them very nicely provided you read the caveats and carefully follow the recommendations and instructions Microsoft provides.

In some respects. it's potentially more confusing for people who have been handling this stuff for years. Because the usual tools used to do it don't apply to Office365. You have learn (and use) the O365 admin interface. Microsoft has made it as simple as possible for a non-IT person to use. So much so that we "pros" sometimes get a little tripped up and confused because...well...it really can't be that easy, right?

Well guess what? It is.

[40hz]

But - I should mention we also have Cox here (Not. My. Idea...) - my preferred configuration especially with cable companies is to bridge their device to an internal main router that can actually be controlled properly.

Now that suggestion is worth it's weight in gold. Even if you have to argue with your supplier to let you do it.

I almost got into a fistfight with a cable service to get them to turn off the friggin' DHCP and NAT on their router,  supplied to my client who was paying for a /29 fixed IP service. They had wanted me to use "sticky" IP assignments for our fixed addresses because it somehow magically made the cable company's "ability to remotely test and maintain the client's router easier." (Which really doesn't make any sense at all. This hokey techno-BSing must be a cable thing!)

It took about twenty minutes and conversations with three separate idiots before I finally got a grownup on the line who reluctantly put the router in bridge mode and left the rest of the driving to us.

Sometimes you just have to stand your ground to get the job done right.

[Stoic Joker]

they decided to jump on that Godaddy /Microsoft combo hosted 365 exchange  !!:(
I can say this .  IF???>>> they ever get ther bugs worked out,  This is one amazing commo package.  I am not a "frills tech".  I only care a bout reliability and speed.

Can't speak for GoDaddy's offering, but I have clients that subscribed directly to the full MS Office365 package which they're accessing through a "business grade" AT&T DSL connection, and the results are exceptionally impressive. At least to my eyes.

Being that we're an MS Cloud services partner, I had to hang back and let you take point on this one. But yes, even with my unbridled hatred for all things ~Cloud~ I really like the Office365 hosted exchange ...(which is why we're an MS Cloud Partner)... And have a few quite happy clients on it as well.

In some respects. it's potentially more confusing for people who have been handling this stuff for years. Because the usual tools used to do it don't apply to Office365. You have learn (and use) the O365 admin interface. Microsoft has made it as simple as possible for a non-IT person to use. So much so that we "pros" sometimes get a little tripped up and confused because...well...it really can't be that easy, right?

Well guess what? It is.

Yepper, it is that. My first trip out was a total baptism by fire because I had to pull 60 calendars from Google, 20 mail boxes from Yahoo, and 1,000+ contacts from all over hell and gone...and get it all stuffed into the remotely hosted Exchanger server straight and under/accessible to the right users. Only took a day or so to get sorted out and it's been smooth sailing from then on.

[questorfla]

As usual. after a whole week of asking every board I could fine (most of whom wont even bother to reply to an IM )
Almost everything I need is right here at "home"  DC.!

OK. I can't fight COX.  Like fighting a pig.  You cant win and just get dirty.  The fact I am old enough to be his father and been doing this since before he could walk makes NO difference.
I agree.  IDIOT.
BUT
unfortunately he speaks the idiocy of our ISP.  What was really FUNNY  was seeing the look n the techs face the first time he connected his test set and ran a speed test on the fiber.    PEGGED the meter 5 times.   

Started blathering something about some "speed boost" technology that "sometimes' gives false readings!
LMAO. 

He turns away fast after the 5th run and calls office.  45 minutes later we are smoothing along at 5gbps.  No more 150 fiber for us.  Glad I made plenty of screen shots to prove it was there to start with. 
But in all honesty. the company was NOT paying for it so... we should not have ever had it anyway. Sure was nice

But their BS about the managed and unmanaged switches was just too much to handle and he refused to back down.
Said the fact I could plug my GB router in to their Modem and the "idiot light" stayed green proved it.
Plug in a switch.  ANY switch. any port. did not matter.  Dropped off to 5gbps.

Plug same switches into router and even cascading them did not slow down a whit.
So.
What to do?  If I buy my OWN CABLE MODEM.  (Which we can do) ..  Any recommendations of which one easiest to bridge?  The officer manager has decided I "don't know a thing about networking" and he is falling down to please the COX tech (because the COX rep is his golfing buddy).  So I am voited an >>>Idiot!<<<

I REALLY need that speed split up before the router.  The Web Servers are Web-Service Servers and need the 150 if we can get it.
The office router also Needs to 150 to provide enough BW for the users to live by the "Cloud" .  They need CLOUD so they can access from anywhere.  Until they blocked port 445 I used to map them back to the office server with ZERO problems for years now.
Some (those not on COX like ME    still do get full speed mapping.  But COX has this area pretty sewed up.
IF anyone (Stoic and 40hz you guys seem to have a similar setup
But mine is done for cheapskates who wont put up the bux for the good stuff.. It STILL SHOULD get better than 5mbps though)  AND  it always DID until they went to 150.  at 50 we got 5.  At 100 we got 5  At 150 we still got 5.

At least they are consistent.  BUT.  Slap in a router.  ANY router it seems and off she goes. 

I'm sort of OK with that IF... there was a way to pass through the static IP's to the end-points.  Each of those little web-service boxes has its own small router (just for use as a simple firewall) .
They are not the problem either. 
IF>> I go to the COX modem> Office router  >  Other router on DHCP (or static in the 192.168.xxx.xxx band) with the final server on a single static ip at 10.0.50.100 with port 80 and a few others forwarded to it.

I think that looks right. Anyway. I can get the full 150 (or close).
but it is worthless as a webserver.  They gave me 6 static IPS to use with the Cable Service.
Then started blathering about me not having a managed switch so it would not work.
I can't argue that if I use a single GB router. it DOES pass through.  Nor that no matter how many switches I run it thrugh after that router I still get 150.

But I don't see a way to get 5 website servers on the 0other side of it  unless  split off before the first router.
This is the way I have done it for years and never had a single problem till now.

I know this is all Cheap consumer grade stuff but a $200+ GB switch should be able to handle the input fro a single cable modem shouldn't it?