Time limits do make sense, but this destroys the high frequency trading market. HFT makes money for the market and provide liquidity. It is valid that they have speed limits, the market just needs to be prepared for them to disappear overnight.
Cancelling before execution is a bad design? Maybe a rate-limiter can exist to detect too many cancelled orders at size, but I cancel orders all the time because I see a better trade or realize it won't fill any time soon. I should have to keep the orders up all day, according to thos rule you propose?
Yea, not sure how they got their numbers. I can say that Huawei is rightly not in the list, they just steal R&D, must be easier. I guess it doesn't matter in China anyway.
I would argue that computer proliferation to the masses means that your average computer user knows LESS "these days" than the average computer user a few years ago. Laptops outsell desktops in the consumer market and the few make it easier for the many. The more people on the network, the easier it gets to find help and guides to do just about anything. More work is done for you and you don't have to be on your own. Think about maybe your grandmother using the emailz on the interweb, it's possible now but probably not likely 10 years ago.
The iPod argument is also part of my point - my little sister can use an iPod because it's designed to be easy and intuitive. These days just means more people are using your application or system. I still don't think that means the systems running your application or service have changed that much in terms of what is supposed to be monitored, sure there may be more parts. Is it up? - is still paramount, but obviously I think it's a good idea to get more info. I've used nagios and it's just not for me, but that's cool.
I believe these days is just overused.
You are mostly right and partly wrong. Store and forward vs. cut through isn't really an issue anymore as it is very rare to see store and forward (unless it's a broadcast frame). These are buffering issues, where CEF is a propriatary forwarding technology whereby the entire routing table (FIB) is put into memory in a quickly accessable hash. Fast switching and CEF are both routing enhancements.
You sound as though you are explaining the original multi-layer switching model: netflow switching. The first packet of a stream was inspected at process level and then a netflow entry was built. Subsequent packets in that stream hit the netflow shortcut and required no process lookup. This is specific to multi-layer switches.
I would say the majority of latency on the megaweb is not from switches - it's probably an mostly distance and then some is router latency.
Dude, they surely tried this. Don't assume you fixed the problem when in fact you weren't there! Depending on the platform and functions applied to this particular device, it could have been much more complicated. Usually they are running traffic tunnelled through the FWSM module and it forgets to take into account the.1q tag or they are using an encryption module which had calculated on pre encryption sizes.
agreed, the number of modules is astounding, seriously. The time it would take to find that kind of market to make all those little modules for a computer would be asinine.
Besides, the future isn't in 7200 series routers, it's CRS, GSR, 6500(7600), etc. You would also be hard pressed to do better than cisco overnight. cisco has an army of programmers and a solid foundation around which endless features are added.
I like the idea though!
Solid comment... a little proofreading never hurt eh?
Another note - Americans, as in North Americans? I think that's pretty ignorant to assume the U.S. comprises all Americans.
Dude, because someone sells a product, that does not mean they are the ones using and operating it. For instance, take the same company: cisco makes networking products, but they are not an ISP - interesting concept eh?
What happened in china? Oh yea, the free market of the world allows a company sell products to another country. Cool. Does your company use netscreen? That's just as bad, but it's not illegal. What China decides to do with it is not up to cisco. There needs to be an uprising of the people, new legislature, etc -- hardware is not going to stop china. They make everything anyway! If they really wanted it, they would just skim it from the factory.
Re:I don't care what you say, surveillance works.
on
Cisco Aquires SyPixx
·
· Score: 1
"If you arent a criminal and arent commiting crimes what do you have to worry about?" - spoken like a true neo-conservative.
You totally missed the point man.
I understand the connectionless small packet idea... but I was referring to a reference Ogun made:
"UDP is done in full software mode, you can't use CEF etc on UDP." I have never heard of this, just wondering what he meant (I work in TAC;))
Welcome to the real world, where supply chain management is not perfect. That is why it is so hard to produce hardware and not just anyone can do it. The good news here is that you can keep track of these issues. When you talk to your sales person come contract renwal time, I am sure they will compensate you for this and other issues like this.
UDP? I think you mean IPX maybe? CEF applies to IP routing and UDP is IP.
You also forgot to mention the GSR and CRS.
The 6500 may not be what these guys are competing against though, I see them trying to compete with the 3600's and ISR's at this point. Either way cisco spends a great deal of time optimizing software algorithims since it is a core component of networking. Some cases hardware helps, but there are quite a few memory models throughout the different lines, and to say most is the same hardware is just not true. AIM encryption module, FWSM, 6k, 4k, 3550, 3750/3560, VPNSM, etc are all examples of hardware accelleration. Heck, even the 2950 does QoS in hardware.
That's my school! By the way if spending money on an education seems like a bad investment, maybe school isn't your thing. I made tuition back not even a year after graduation:)
Slashdot Mirror
Time limits do make sense, but this destroys the high frequency trading market. HFT makes money for the market and provide liquidity. It is valid that they have speed limits, the market just needs to be prepared for them to disappear overnight.
Cancelling before execution is a bad design? Maybe a rate-limiter can exist to detect too many cancelled orders at size, but I cancel orders all the time because I see a better trade or realize it won't fill any time soon. I should have to keep the orders up all day, according to thos rule you propose?
This article is competent. If you would like to share your story with the class, we can tell you how great your writing is.
Yea, not sure how they got their numbers. I can say that Huawei is rightly not in the list, they just steal R&D, must be easier. I guess it doesn't matter in China anyway.
I know cisco is near the $1B mark, and I'm sure there are others that I can't think of right now ;)
Does anyone else see that flicker on the friendlysolutions site with an LCD? ouch.
I would argue that computer proliferation to the masses means that your average computer user knows LESS "these days" than the average computer user a few years ago. Laptops outsell desktops in the consumer market and the few make it easier for the many. The more people on the network, the easier it gets to find help and guides to do just about anything. More work is done for you and you don't have to be on your own. Think about maybe your grandmother using the emailz on the interweb, it's possible now but probably not likely 10 years ago. The iPod argument is also part of my point - my little sister can use an iPod because it's designed to be easy and intuitive. These days just means more people are using your application or system. I still don't think that means the systems running your application or service have changed that much in terms of what is supposed to be monitored, sure there may be more parts. Is it up? - is still paramount, but obviously I think it's a good idea to get more info. I've used nagios and it's just not for me, but that's cool. I believe these days is just overused.
I love when people say "these days". So back in 1995 things were different? When did they change? Why the time reference?
Not everything is about home internet usage. There is much more to networking and the internet in general, than the home user.
You are mostly right and partly wrong. Store and forward vs. cut through isn't really an issue anymore as it is very rare to see store and forward (unless it's a broadcast frame). These are buffering issues, where CEF is a propriatary forwarding technology whereby the entire routing table (FIB) is put into memory in a quickly accessable hash. Fast switching and CEF are both routing enhancements. You sound as though you are explaining the original multi-layer switching model: netflow switching. The first packet of a stream was inspected at process level and then a netflow entry was built. Subsequent packets in that stream hit the netflow shortcut and required no process lookup. This is specific to multi-layer switches. I would say the majority of latency on the megaweb is not from switches - it's probably an mostly distance and then some is router latency.
Dude, they surely tried this. Don't assume you fixed the problem when in fact you weren't there! Depending on the platform and functions applied to this particular device, it could have been much more complicated. Usually they are running traffic tunnelled through the FWSM module and it forgets to take into account the .1q tag or they are using an encryption module which had calculated on pre encryption sizes.
So they can do what redhat does... enterprise support on an OSS product.
agreed, the number of modules is astounding, seriously. The time it would take to find that kind of market to make all those little modules for a computer would be asinine. Besides, the future isn't in 7200 series routers, it's CRS, GSR, 6500(7600), etc. You would also be hard pressed to do better than cisco overnight. cisco has an army of programmers and a solid foundation around which endless features are added. I like the idea though!
When the rights of the unpopular are abridged, everyone's rights are in danger.
Well put.
Solid comment... a little proofreading never hurt eh? Another note - Americans, as in North Americans? I think that's pretty ignorant to assume the U.S. comprises all Americans.
Dude, because someone sells a product, that does not mean they are the ones using and operating it. For instance, take the same company: cisco makes networking products, but they are not an ISP - interesting concept eh?
What happened in china? Oh yea, the free market of the world allows a company sell products to another country. Cool. Does your company use netscreen? That's just as bad, but it's not illegal. What China decides to do with it is not up to cisco. There needs to be an uprising of the people, new legislature, etc -- hardware is not going to stop china. They make everything anyway! If they really wanted it, they would just skim it from the factory.
"If you arent a criminal and arent commiting crimes what do you have to worry about?" - spoken like a true neo-conservative. You totally missed the point man.
I understand the connectionless small packet idea... but I was referring to a reference Ogun made: "UDP is done in full software mode, you can't use CEF etc on UDP." I have never heard of this, just wondering what he meant (I work in TAC ;))
And that's going to happen with the automatic updates or what? Network software upgrades take a long time to go through...
Welcome to the real world, where supply chain management is not perfect. That is why it is so hard to produce hardware and not just anyone can do it. The good news here is that you can keep track of these issues. When you talk to your sales person come contract renwal time, I am sure they will compensate you for this and other issues like this.
UDP? I think you mean IPX maybe? CEF applies to IP routing and UDP is IP. You also forgot to mention the GSR and CRS. The 6500 may not be what these guys are competing against though, I see them trying to compete with the 3600's and ISR's at this point. Either way cisco spends a great deal of time optimizing software algorithims since it is a core component of networking. Some cases hardware helps, but there are quite a few memory models throughout the different lines, and to say most is the same hardware is just not true. AIM encryption module, FWSM, 6k, 4k, 3550, 3750/3560, VPNSM, etc are all examples of hardware accelleration. Heck, even the 2950 does QoS in hardware.
Hmm, sounds just September 11, 2001... if you don't know what I am talking about, educate yourself.
I was about to make that comment, kudos. More bandwidth is just brute force, multicast can git 'er done.
That's my school! By the way if spending money on an education seems like a bad investment, maybe school isn't your thing. I made tuition back not even a year after graduation :)