Post by thread:[EE] Managed switch versus unmanaged switch

On Nov 19, 2007, at 7:17 PM, Xiaofan Chen wrote: > http://en.wikipedia.org/wiki/Network_switch > > I am a bit confused with the term Managed switch and I am not so > sure if the above Wikipedia entry is correct or not. > > The thing is that I tend to beleive that the cheap wireless routers > (normally including 4-port switch) are not managed. However they > do offer a web interface to do some basic configuration. I think I agree with the Wikipedia definition. Whether the switch that is part of your wireless router is "managed" or not depends on whether the router CPU offers visibility into statistics and configuration of the switching chip itself; it COULD be a router with an integrated unmanaged switch, or it could be somewhat better; either is useful in the consumer setting. The dumbest of "managed switches" are certainly little more than a switch chip with a microcontroller added to give you that config and statistics interface to the chip. > What I heard from a seminar is that N-Tron managed switch > will offer features such as IGMP Snooping and they are more > expensive than unmanaged switch. "managed" is far from a complete description; there's a wide range of features you might find on a more complex managed switch. A top-of-the-line ci$co switch probably has enough features to be incomprehensible, and we work all the time to add more, so we can justify the high price. BillW

On 11/20/07, William Chops Westfield <spam_OUTwestfwTakeThisOuTmac.com> wrote: {Quote hidden}> > On Nov 19, 2007, at 7:17 PM, Xiaofan Chen wrote: > > > http://en.wikipedia.org/wiki/Network_switch > > > > I am a bit confused with the term Managed switch and I am not so > > sure if the above Wikipedia entry is correct or not. > > > > The thing is that I tend to beleive that the cheap wireless routers > > (normally including 4-port switch) are not managed. However they > > do offer a web interface to do some basic configuration. > > I think I agree with the Wikipedia definition. Whether the switch > that is part of your wireless router is "managed" or not depends on > whether the router CPU offers visibility into statistics and > configuration of the switching chip itself; it COULD be a router > with an integrated unmanaged switch, or it could be somewhat better; > either is useful in the consumer setting. The dumbest of "managed > switches" are certainly little more than a switch chip with a > microcontroller added to give you that config and statistics > interface to the chip. I see. Any example of a dumbest managed switches? In the Wikipedia definition, a router is a Layer 3 switch. So A Linksys WRT54G will be a router integrated with 4 switches. Is the integrated switch managed or unmanaged? I think the web interface only manages the router part and not the 4-port switch part. How do I tell? Then I read this and it is talking about the difference of a Layer 3 switch and a router is the implementation: MPU versus ASIC. Are these just marketing terms? http://en.wikipedia.org/wiki/Multilayer_switch {Quote hidden}> > What I heard from a seminar is that N-Tron managed switch > > will offer features such as IGMP Snooping and they are more > > expensive than unmanaged switch. > > "managed" is far from a complete description; there's a wide range > of features you might find on a more complex managed switch. A > top-of-the-line ci$co switch probably has enough features to be > incomprehensible, and we work all the time to add more, so we > can justify the high price. > I see. I've read more and more Cisco switches in the industrial automation related news. Example: http://phx.corporate-ir.net/phoenix.zhtml?c=196186&p=irol-newsArticle&ID=1078217&highlight= Xiaofan

In my limited experience, a managed switch is one where you can not only monitor it on a very fine grained level, but you can enable/disable ports, split it into sections (ie, four ports form one network, the rest another completely seperate network), copy and re-route traffic (ie, can listen in on 'switched' packets as if they were on a hub) as well as SNMP. They make it easy to diagnose problems and remotely fix them. An unmanaged switch may have some ability to monitor statistics, but rarely can be sectioned (this is the right term, but I can't recall what it is right now), re-routed, and generally don't have full SNMP interface and control. -Adam On Nov 19, 2007 10:17 PM, Xiaofan Chen <.....xiaofancKILLspam@spam@gmail.com> wrote: {Quote hidden}> http://en.wikipedia.org/wiki/Network_switch > > I am a bit confused with the term Managed switch and I am not so > sure if the above Wikipedia entry is correct or not. > > The thing is that I tend to beleive that the cheap wireless routers > (normally including 4-port switch) are not managed. However they > do offer a web interface to do some basic configuration. > > In the lab, we use either off-the-shelf commercial grade switches > or industrial grade N-Tron 405TX (all called unmanaged). > http://www.n-tron.com/html/400.html > > What I heard from a seminar is that N-Tron managed switch > will offer features such as IGMP Snooping and they are more > expensive than unmanaged switch. > http://www.n-tron.com/html/500.html > ethernet.industrial-networking.com/articles/articledisplay.asp?id=936 > http://www.cisco.com/warp/public/473/22.html > > > Regards, > Xiaofan > -

On Nov 19, 2007, at 10:25 PM, Xiaofan Chen wrote: > In the Wikipedia definition, a router is a Layer 3 switch. So A > Linksys > WRT54G will be a router integrated with 4 switches. Is the integrated > switch managed or unmanaged? I think the web interface only > manages the router part and not the 4-port switch part. How do I tell? The WRT54G series make that question interesting. First off, there's something like seven sub-models of the WRT54G and WRT54GS, so every one has slightly different hardware. Then you mix in that the ones that have the nice Broadcom smart switch hardware in them, typically don't have any access to the switch from Linksys' default firmware... but... Firmware like DD-WRT and OpenWRT *can* and *does* have access to some of the advanced features of the switch chipset, through folks hacking it in. I still wouldn't put it even close to on-par with a multi-thousand- dollar Cisco switch running a later version of one of the feature- filled versions of IOS, though. > Then I read this and it is talking about the difference of a Layer > 3 switch and a router is the implementation: MPU versus ASIC. > Are these just marketing terms? > http://en.wikipedia.org/wiki/Multilayer_switch Yes. And you're not even guaranteed that anyone does anything in an ASIC in a switch advertised as a "layer 3" switch. It's all just hype. Even within the BIG vendor's model lines, a switch with ASIC switching might have backplane speed limitations or other problems throughout the model line. (The early Cisco 7206VXR's with the original SUP1 cards were pig-dogs and couldn't do what Cisco's marketing department said they could -- handle gigabit upstream links -- for example.) With there not any serious engineering test labs for such equipment who can afford to do non-partisan third-party tests of the high-end gear, the only real information network that counts is staying in touch with others in the industry, especially if you're handling router/switching purchases say for a large carrier, at gigabit or faster speeds. It's a small tight-knit group who all help each other evaluate equipment and they mercilessly beat up vendors who lie to them, in public. Usually on various mailing lists devoted to carrier network managers and gear-specific lists. {Quote hidden}>>> What I heard from a seminar is that N-Tron managed switch >>> will offer features such as IGMP Snooping and they are more >>> expensive than unmanaged switch. >> >> "managed" is far from a complete description; there's a wide range >> of features you might find on a more complex managed switch. A >> top-of-the-line ci$co switch probably has enough features to be >> incomprehensible, and we work all the time to add more, so we >> can justify the high price. Some of the Cisco gear is overkill, and all of it is expensive... but... often when your network gets to a certain size and you need Cisco gear for edge-connectivity, you end up with it everywhere else due to familiarity and brand "loyalty", I guess. And I must admit, when you first look at a Cisco 2950 type switch for what looks to be a "simple" deployment and then you show up and the customer has some WILD network configuration that requires you jump through some strange spanning-tree, or routing protocol hoops... that 2950 starts to look like a great deal. There are others that do a good job, of course... Foundry Networks are well-known for good gear, too. > I see. I've read more and more Cisco switches in the industrial > automation related news. > > Example: > http://phx.corporate-ir.net/phoenix.zhtml?c=196186&p=irol-newsArticle&ID=1078217&highlight= Cisco is like Microsoft... the old joke, "No one ever got fired for buying Windows." Or the older version, "No one ever got fired for buying IBM." Same thing with switches/routers... "No one ever got fired for buying Cisco." Brand loyalty is strong. Frankly, Nortel (after going through some years of really bad/annoying switching products) seems to have their act together again, and are making some very nice (much cheaper than Cisco with comparable or even better features -- Nortel has both a command line and a MENU structure, that blows Cisco out of the water, if you're not an IOS expert) equipment. The problem is... Cisco's market domination and one of the only truly successful training/certification programs in "Internet" technology space -- being a CCIE actually means you know something, unlike just about every other IT certification out there -- means that most people you'd likely WANT to employ to take care of your network, are Cisco- trained and know IOS. Kinda the same thing as saying the people you want handling your core production systems that make money and have almost no downtime ever, know Unix. Probably the commercial "flavors" like Solaris, HP-UX, AIX, and such -- not just Linux they learned in their spare time. (GRIN! Hahaha... just kidding, maybe! And nowadays that Linux comment may or may not be true, depending on distro and professionalism of the system admin deploying it! It can save a lot of money when done right, but lots of bad sysadmins out there slap Linux on a box without any thought to anything about the system's performance, recoverability from failures, etc... and then reap the rewards of using a $500 pizza-box for a critical system.) -- Nate Duehr nateKILLspamnatetech.com

On Nov 19, 2007, at 9:25 PM, Xiaofan Chen wrote: > I see. Any example of a dumbest managed switches? I dunno. Some of the early switch chips were only manageable via serial port. I think nowadays "managed" usually means "manageable via the network itself." > > In the Wikipedia definition, a router is a Layer 3 switch. Um. Sort of. > So A Linksys WRT54G will be a router integrated with 4 switches. Most of the cheap routers consist of a routing "engine" running linux with a 5-port (or whatever) switch chip on one of the ports. One router and one switch, I would say. (perhaps the switch chip is now built in to some of the SoC chips...) > Is the integrated switch managed or unmanaged? I think the web > interface only > manages the router part and not the 4-port switch part. How do I tell? I don't understand why the label matters. Do you have some particular management task you want to achieve? If so, see if it's supported by reading the manual. If not, why do you think you care about the differences between managed and unmanaged? > Then I read this and it is talking about the difference of a Layer > 3 switch and a router is the implementation: MPU versus ASIC. > Are these just marketing terms? Probably just marketing terms. I would say that a layer three switch is capable of switching packets around based on data in the layer three headers of the packets, but a router has to be smart enough to speak at least one high-level routing protocol for distributing info about which packets to ship where. Any high-end "router" box is going to have ASICs doing some or all of the actual packet switching; otherwise your pps is going to be limited to only about 3Mpps per CPU (one bi- directional Gigabit ethernet interface...) There may be uses for "layer 3 switches" that are particularly dumb, but I'm not sure what it would be... BillW

On Nov 19, 2007, at 11:27 PM, M. Adam Davis wrote: > In my limited experience, a managed switch is one where you can not > only monitor it on a very fine grained level, but you can > enable/disable ports, split it into sections (ie, four ports form one > network, the rest another completely seperate network), copy and > re-route traffic (ie, can listen in on 'switched' packets as if they > were on a hub) as well as SNMP. Right, but nowadays you have to be VERY specific about those features, because even the WRT54GS with the Broadcom chipset in it can do most of that... with the non-supported non-Linksys firmware in it. The chipset has VLAN tagging, QoS feature on the chipset itself, etc etc etc. But that little device won't keep up with certain speeds at all, and isn't exactly "industrially" built to any kind of hardware standard other than a nice board in a plastic box... > They make it easy to diagnose problems and remotely fix them. That's debatable. They make it easy for folks with proper training and understanding of networks to diagnose problems. If the person is a clueless noob, the managed switch is an insurmountable problem -- they can't even figure out how to log into it, let alone fix the network! (But they were probably going to have to escalate to someone with a clue, anyway.) The real difficulty I see in new engineers/techs right now is the lack of ever having studied the OSI model, or having seen any of the trials and tribulations of each layer as it was built up to the giant scale it's at today, when you talk about carrier-class networks. They go to "Cisco Boot Camp" and learn Cisco's version of the Layered model (it more accurately defines their marketing terms than matches the real original OSI model above layer 3) and they can't speak fluently about how all these interconnected layers interoperate anymore -- they're in over their heads on their first day, and it takes a year or more for them to really "get it"... usually longer. (I remember when devices were all "dumb" and I could get away with teaching an abbreviated OSI model to technicians... because the only layers they could effect changes to in a real network back then were 1, 3 and 7. The "stuff" inbetween either couldn't be changed because it was embedded by the manufacturer in the chipsets in the routers and switches, or it had to be escalated back to an engineer for a code change and a recompile. I remember when I thought it was "cool" that my 90's Linux box could change its MAC address! Now MAC cloning is a feature in every SOHO "router" in virtually ever household that has broadband Internet access... and is even less understood than ever.) > An unmanaged switch may have some ability to monitor statistics, but > rarely can be sectioned (this is the right term, but I can't recall > what it is right now), re-routed, and generally don't have full SNMP > interface and control. Yeah, you're looking for VLAN switching for the term... and generally you're right... but to really spec out a piece of network gear these days you MUST know a lot more than the general definition of "managed switch" -- especially after manufacturers moved what was traditionally the "router" into the switch hardware for Layer 3 and up, and put it in an ASIC so it'd go ultra-fast. We have customers that complain that in any failure scenario imaginable, they want less than 5 second reconvergence times on networks that have upwards of 12 devices. Only the Cisco's and other big guns of the networking world can even do it, and it always requires a deep knowledge of every protocol involved in the failover monitoring/switching and proper design of the network to accomplish that. The price tag jumps and the negotiations start... "could you handle a 20 second failover in the worst-case scenario, and we'll use some cheaper gear and non-proprietary routing protocols for half the price?" -- Nate Duehr .....nateKILLspam.....natetech.com

On Tue, 2007-11-20 at 13:25 +0800, Xiaofan Chen wrote: > I see. Any example of a dumbest managed switches? > > In the Wikipedia definition, a router is a Layer 3 switch. So A Linksys > WRT54G will be a router integrated with 4 switches. Is the integrated > switch managed or unmanaged? I think the web interface only > manages the router part and not the 4-port switch part. How do I tell? Managed switch is very much up to the manufacturer to define. That said, a "managed" switch to me, at the very least, will allow me to configure each port of the switch (i.e. set it to 10Mbps only, or turn it off), and see the status of each port (i.e. a 100Mbps is connected to port 2). There are MANY more features available on managed switches, but I would classify this as a bare minimum. > Then I read this and it is talking about the difference of a Layer > 3 switch and a router is the implementation: MPU versus ASIC. > Are these just marketing terms? > http://en.wikipedia.org/wiki/Multilayer_switch Frankly I believe the distinction of whether a piece of hardware does it's work mostly in software or mostly in hardware is pointless. Yes, generally an ASIC will be more reliable then software running on an MPU, but that isn't necessarily so, and there are certainly benefits of doing more stuff in software (i.e. in the field fixes or upgrades). TTYL

On Nov 21, 2007, at 2:14 PM, Herbert Graf wrote: {Quote hidden}> Managed switch is very much up to the manufacturer to define. That > said, > a "managed" switch to me, at the very least, will allow me to > configure > each port of the switch (i.e. set it to 10Mbps only, or turn it off), > and see the status of each port (i.e. a 100Mbps is connected to port > 2). > There are MANY more features available on managed switches, but I > would > classify this as a bare minimum. Setting full or half duplex is also a typical feature. Comes in handy on some gear that won't auto-negotiate properly. > Frankly I believe the distinction of whether a piece of hardware does > it's work mostly in software or mostly in hardware is pointless. Yes, > generally an ASIC will be more reliable then software running on an > MPU, > but that isn't necessarily so, and there are certainly benefits of > doing > more stuff in software (i.e. in the field fixes or upgrades). Agreed. Metrics like how many 100 Mb/s (or faster - GigE) connections the device can handle at the same time (backplane speed), etc... are far more important in server room work. For home use, probably not so important. Other things to look for in the datacenter include things like dual power supplies, etc. VLAN partitioning also typically splits up the backplane speed between the VLAN's so there's some metrics there that should be paid attention to, for real professional networks. -- Nate Duehr EraseMEnatespam_OUTTakeThisOuTnatetech.com

Xiaofan Chen wrote: > On 11/22/07, William Chops Westfield <KILLspamwestfwKILLspammac.com> wrote: >> On Nov 21, 2007, at 5:43 PM, Xiaofan Chen wrote: >> >>> A hardware serial modem is way better [than winmodems] >> A "hardware" serial modem does most of its work in >> software. But having a separate CPU (or several) in the >> modem itself is a fine example of when it makes sense >> to use multi-processing :-) (which would you rather >> use - 20% of your $2000 desktop system, or all of a >> $50 peripheral? The fundamental flaw is that the CPU >> cycles in the modem were never really expensive enough >> to economically replace using the desktop CPU.) >> > > I see. I have not used a modem (ok the Dell notebooks > at work and at home still has one) for a while (three years+) > due to the availability of Broadband. I opened the old modem > a while ago and it has an Intel 80186 (no more from Intel) > inside. I think a few of the current modem chips are based on the 6502 family of chips. I remember my 300 baud modem used an 8031. -- Linux Home Automation Neil Cherry RemoveMEncherryTakeThisOuTlinuxha.com http://www.linuxha.com/ Main site http://linuxha.blogspot.com/ My HA Blog Author of: Linux Smart Homes For Dummies

On Nov 21, 2007, at 9:30 PM, Neil Cherry wrote: >> I see. I have not used a modem (ok the Dell notebooks >> at work and at home still has one) for a while (three years+) >> due to the availability of Broadband. I opened the old modem >> a while ago and it has an Intel 80186 (no more from Intel) >> inside. > > I think a few of the current modem chips are based on the 6502 > family of chips. I remember my 300 baud modem used an 8031. I think most modems these days have two processors. One tends to be a low-performance general purpose micro like an 8031; it does the user interface (such as it is), manipulation of relays for taking the phone off hook, and so on. The second processor is a "data pump"; usually an application specific DSP of some kind that does the actual modulations and demodulation. Data-side compression and error correction can happen in either one, and that's where you see some of the performance differences between a "good" modem and a "cheap" modem. Sometimes they're both on one chip. I have in front of me a digital modem dimm from one of the older high-end cisco Access servers (AS5800, AS5300); it's got two intel 80960 cpu, six TI DL17337 (DSP things, I think) , 12 64k*16 RAM BGA chips, and some other stuff. IIRC, this implements 12 "universal ports" - the DSPs can be loaded with different code to run either digital modems or Voice CoDecs depending on what's on the other end of the line. BillW

On Thu, 2007-11-22 at 09:43 +0800, Xiaofan Chen wrote: > On 11/22/07, Herbert Graf <spamBeGonemailinglist3spamBeGonefarcite.net> wrote: > > > Then I read this and it is talking about the difference of a Layer > > > 3 switch and a router is the implementation: MPU versus ASIC. > > > Are these just marketing terms? > > > http://en.wikipedia.org/wiki/Multilayer_switch > > > > Frankly I believe the distinction of whether a piece of hardware does > > it's work mostly in software or mostly in hardware is pointless. > > This depends on the product. Remember the famous WinModem? Obviously, but that is a particularly extreme example. > A hardware serial modem is way better. Mostly true, but not universally so, I encountered some pretty bad bugs in a USRobotics 33.6kbps hardware modem, had to exchange it 4 times before I got a version where the bug was fixed. Had it been a software modem theoretically it would have been alot less bothersome for me to get a properly working modem. At first yes, winmodems were complete and utter crap. Much of the blame can be thrown at the driver makers, but some falls to windows itself. That said, pretty much all internal modems these days are winmodems, and they work perfectly well (at least the ones I've used in the past 4 years). Frankly, the winmodem example is apples vs. oranges. Remember, you are comparing a technology that was a LONG time in the making (hardware modems have existed for a LONG time) vs. winmodems which were brand new both in concept and design. Given the same amount of development time I don't see why a hardware modem is "better" these days. It's like comparing a DVD player today vs. the first BluRay player. The BluRay player has WAY more bugs then the DVD player. Why? DVD players have been revised and fixed for over a decade, BluRay is brand new. TTYL