 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"We started with the humble goal of being the first to get a PC router to 10
[gigabytes per second], but we pushed it to 40," says Sue Moon, leader of
the lab in which the research was conducted. Her students Sangjin Han and
Keon Jang developed software called PacketShader that made this possible.
PacketShader uses a computer's graphics processing unit (GPU) to help
process packets of data sent across a network.
http://www.technologyreview.com/communications/26096/
Altho I'm not real sure how they mean "based entirely on software" if it's
using a GPU. :-)
--
Darren New, San Diego CA, USA (PST)
Quoth the raven:
Need S'Mores!
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
It is an interesting way to use commodity hardware, instead of custom
stuff. Though, the article does make it hard to take them seriously when
they use lines like "Modern routers are rarely dumb switches anymore."
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
packetshader... sounds very photorealistic... :)
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Sabrina Kilian a écrit :
> It is an interesting way to use commodity hardware, instead of custom
> stuff. Though, the article does make it hard to take them seriously when
> they use lines like "Modern routers are rarely dumb switches anymore."
Most comemrcial routers HAVE been switches for the last 20 years. The
first packet to come through would be routed by the cpu, but the result
of that routing table look up would be sent to an interface cache and
all the packets for the same destination after that would be switched
directly by the interface hardware, leaving the CPU alone.
I think that what they meant by this is that the routers nowadays have
to do policy routing and load-balancing, that is inspect up to the
application layer before deciding where to send the packets, rather than
just based on the destination address.
--
/*Francois Labreque*/#local a=x+y;#local b=x+a;#local c=a+b;#macro P(F//
/* flabreque */L)polygon{5,F,F+z,L+z,L,F pigment{rgb 9}}#end union
/* @ */{P(0,a)P(a,b)P(b,c)P(2*a,2*b)P(2*b,b+c)P(b+c,<2,3>)
/* gmail.com */}camera{orthographic location<6,1.25,-6>look_at a }
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Francois Labreque wrote:
> Sabrina Kilian a écrit :
>> It is an interesting way to use commodity hardware, instead of custom
>> stuff. Though, the article does make it hard to take them seriously when
>> they use lines like "Modern routers are rarely dumb switches anymore."
>
> Most comemrcial routers HAVE been switches for the last 20 years. The
> first packet to come through would be routed by the cpu, but the result
> of that routing table look up would be sent to an interface cache and
> all the packets for the same destination after that would be switched
> directly by the interface hardware, leaving the CPU alone.
Switches have moved up to handling some of the layer 3 protocols and
doing some of the stuff typically thought of as router work. Managing
VLANs and the like, and switching stuff between them. If the router
never had to handle BGP, and the like, then a smart switch could replace
it. Depending on the switch, it might even work where all you need is
QoS and some switching between local disparate IP subnet and VLAN.
However, a smart switch won't replace a router where one is really
needed; handling all the annoying route finding when you have a more
complex topology outside the area of your network. At least, I haven't
seen one that manages BGP. Wiki tells me they may exist.
Still, all of this is smart switches moving up to fill the need of
simpler routers. Any router, barring marketing names, has never just
switched packets on layer 2 like a dumb switch will.
I feel like maybe I am over explaining the joke of "Layer 3 devices are
rarely layer 2 devices anymore."
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Sabrina Kilian wrote:
> I feel like maybe I am over explaining the joke of "Layer 3 devices are
> rarely layer 2 devices anymore."
I remember when we moved to the new building at work. I was just going
to pay maybe £300 for some new network switches. But no, the Director of
IT insisted that we had to buy Cisco switches, which are £1,500 each.
As soon as we started setting them up, I quickly realised that they
aren't switches at all. They're routers. You can filter traffic, route
between different network numbers and God only knows what else. (Not, of
course, that *I* have access to do any of this. I'm only the person in
charge of the network, right?)
--
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Sabrina Kilian a écrit :
> Francois Labreque wrote:
>> Sabrina Kilian a écrit :
>>> It is an interesting way to use commodity hardware, instead of custom
>>> stuff. Though, the article does make it hard to take them seriously when
>>> they use lines like "Modern routers are rarely dumb switches anymore."
>> Most comemrcial routers HAVE been switches for the last 20 years. The
>> first packet to come through would be routed by the cpu, but the result
>> of that routing table look up would be sent to an interface cache and
>> all the packets for the same destination after that would be switched
>> directly by the interface hardware, leaving the CPU alone.
>
> Switches have moved up to handling some of the layer 3 protocols and
> doing some of the stuff typically thought of as router work. Managing
> VLANs and the like, and switching stuff between them. If the router
> never had to handle BGP, and the like, then a smart switch could replace
> it. Depending on the switch, it might even work where all you need is
> QoS and some switching between local disparate IP subnet and VLAN.
> However, a smart switch won't replace a router where one is really
> needed; handling all the annoying route finding when you have a more
> complex topology outside the area of your network. At least, I haven't
> seen one that manages BGP. Wiki tells me they may exist.
>
Even if it does appear that the author of the article's experience with
routers is limited to the Linksys or Netgear home stuff, I'm not talking
about "layer-3" switches. I'm talking about a router that does
switching in hardware based on mac-addresses.
> Still, all of this is smart switches moving up to fill the need of
> simpler routers. Any router, barring marketing names, has never just
> switched packets on layer 2 like a dumb switch will.
>
The Cisco 7000 or 7500 series have been doing just that, and the newer
GSR and CRS models do it as well. As I said earlier, the first packet
to come by will be CPU-routed, but all the following ones* for that same
destination will be handled in hardware based on mac-address look-ups.
This is how a Cisco 7000 was able to handle 16M pps with only a 33Mhz
CPU, 20 years ago.
--
/*Francois Labreque*/#local a=x+y;#local b=x+a;#local c=a+b;#macro P(F//
/* flabreque */L)polygon{5,F,F+z,L+z,L,F pigment{rgb 9}}#end union
/* @ */{P(0,a)P(a,b)P(b,c)P(2*a,2*b)P(2*b,b+c)P(b+c,<2,3>)
/* gmail.com */}camera{orthographic location<6,1.25,-6>look_at a }
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Orchid XP v8 a écrit :
> Sabrina Kilian wrote:
>
>> I feel like maybe I am over explaining the joke of "Layer 3 devices are
>> rarely layer 2 devices anymore."
>
> I remember when we moved to the new building at work. I was just going
> to pay maybe £300 for some new network switches. But no, the Director of
> IT insisted that we had to buy Cisco switches, which are £1,500 each.
>
As I mentioned to you at the time, the IT Director might have had very
good reasons for going with Cisco gear. (One of which may or may not be
that they took him on a good golf outing!)
What kind of troubleshooitng facilities exist for your £300 switches?
What's their MTBF? Can they survive a broadcast storm? Can they
PREVENT a broadcast storm?
> As soon as we started setting them up, I quickly realised that they
> aren't switches at all. They're routers. You can filter traffic, route
> between different network numbers and God only knows what else.
See? That's probably because the director of IT wanted to use some of
these features.
> (Not, of
> course, that *I* have access to do any of this. I'm only the person in
> charge of the network, right?)
>
Most companies will not grant more than read-only access to their onsite
tech, if any access at all. As you mention, these switches can do much
more than just local switching between ports. Untrained people have
been known to mess with stuff they didn't fully understand and cause
network-wide outages. (Trained people too!)
I'm not suggesting you're a cowboy who does any of this, but if you're
interested in helping out your network folks (who are in the US, IIRC)
Cisco offers courses. I'd suggest bringing that up with your manager
and explain that this way, you'd be able to provide faster help when a
problem arises and the boys from the Mothership are still asleep.
--
/*Francois Labreque*/#local a=x+y;#local b=x+a;#local c=a+b;#macro P(F//
/* flabreque */L)polygon{5,F,F+z,L+z,L,F pigment{rgb 9}}#end union
/* @ */{P(0,a)P(a,b)P(b,c)P(2*a,2*b)P(2*b,b+c)P(b+c,<2,3>)
/* gmail.com */}camera{orthographic location<6,1.25,-6>look_at a }
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
>> I remember when we moved to the new building at work. I was just going
>> to pay maybe £300 for some new network switches. But no, the Director
>> of IT insisted that we had to buy Cisco switches, which are £1,500 each.
>>
>
> As I mentioned to you at the time, the IT Director might have had very
> good reasons for going with Cisco gear. (One of which may or may not be
> that they took him on a good golf outing!)
>
> What kind of troubleshooitng facilities exist for your £300 switches?
> What's their MTBF? Can they survive a broadcast storm? Can they
> PREVENT a broadcast storm?
It's a switch? It connects two computers together. What could *possibly*
go wrong? In about 5 years, the existing switches we had worked
flawlessly. The only reason I was going to buy new switches was so that
they would be gigabit. Ironically, we bought Cisco switches that provide
PoE (which none of our equipment uses), and apparently you can have PoE
*or* gigabit, so they bought PoE.
>> As soon as we started setting them up, I quickly realised that they
>> aren't switches at all. They're routers. You can filter traffic, route
>> between different network numbers and God only knows what else.
>
> See? That's probably because the director of IT wanted to use some of
> these features.
Yeah. They actually split our network into several logically seperate
networks. I have absolutely no idea why. Other than needlessly
increasing the complexity of the network, I cannot imagine what useful
purpose this serves. It's not like there's any special routing or
anything; everything still routes to everything. It's just that now I've
got 5 network numbers to worry about instead of 1. And I have to
remember which ports are on which subnet.
(Oh, and did I mention? All the personell who had the access passwords
and the training to know what to do with them have now left the company,
so the switch configuration can never be altered again.)
>> (Not, of course, that *I* have access to do any of this. I'm only the
>> person in charge of the network, right?)
>
> Most companies will not grant more than read-only access to their onsite
> tech, if any access at all.
The guys at HQ seem to have the general opinion that anything more
complex than switching on a light switch is far beyond my capabilities.
(Presumably because for everyone *else* in IT, this is actually the
case. :-P ) When we had new servers come in, they had to send a guy from
America over here to set it up for me. Because, you know, I'm completely
incapable of installing Windows by myself, right? I mean, it's not like
I've already set up 3 different servers all by myself to far. Oh, wait...
> I'm not suggesting you're a cowboy who does any of this, but if you're
> interested in helping out your network folks (who are in the US, IIRC)
> Cisco offers courses. I'd suggest bringing that up with your manager
> and explain that this way, you'd be able to provide faster help when a
> problem arises and the boys from the Mothership are still asleep.
Hmm. Getting them to pay for some Cisco training might not be a bad idea
actually. When I do finally manage to leave, that may come in handy...
--
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Orchid XP v8 a écrit :
> (Oh, and did I mention? All the personell who had the access passwords
> and the training to know what to do with them have now left the company,
> so the switch configuration can never be altered again.)
>
Go to www.cisco.com
type "password recovery XXXX" in the search box (where XXXX is the model
number, e.g. 3560)
It usually involves holding a button pressed while you boot the switch
and then typing low-level command to load the IOS without the config and
requires a serial cable to connect to the console port (which means you
probably need a USB-to-DB9 adapter too)
I did not tell you this. In fact, I was never here.
--
/*Francois Labreque*/#local a=x+y;#local b=x+a;#local c=a+b;#macro P(F//
/* flabreque */L)polygon{5,F,F+z,L+z,L,F pigment{rgb 9}}#end union
/* @ */{P(0,a)P(a,b)P(b,c)P(2*a,2*b)P(2*b,b+c)P(b+c,<2,3>)
/* gmail.com */}camera{orthographic location<6,1.25,-6>look_at a }
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
|
|
