NAT and TCP Load Distribution [7:62088]
Can anyone give me a real world example of why you would need to consider using TCP Load Distribution. I am summarising slightly but TCP Load Distribution seems to be a method of using a single IP address (the Global Inside Address)inbound; which is handed off to different devices on the inside. OK. If this is a fair description I can see that this would be useful for load sharing amongst internal servers. IE maybe if an increasing number of Internet customers were accessing your resources - on line shopping whatever - that you might want to spread (balance) the load among several identical servers? So is 'loosely' like HSRP (not to do with redundancy so much) but conceptually in that there is a VIRTUAL entity that supports multiple physical entities (servers) to enable the load distribution amongst these 'real' devices. Therefore is the case that the real devices dont need to be network devices - they would most likely be UNIX (typically Internet facing) boxes of some sort? Any response to clarify my muddle thinking much appreciated! Apologies for dumb question. Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=62088&t=62088 -- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
Re: NAT and TCP Load Distribution [7:62088]
hi it is nothing like HSRP, even if looked @ from the virtual IP point of view, for the simple reason that HSRP elects one active router and only router is active @ any time, (as opossed to TCP load Balacing, that uses all the ip's of the servers to forward data to) and the real ip of the active HSRP router is transparent to the user's of course but to the packets them selfs as well. TCP load balancing NOT sharing, (because the router will distribute the flows amongst the TCP server's, one flow for each server in a round robin fashion, therefore balancing, because traffic is equally balanced based on flows (unless if u look at it in terms of bandwith, in which case, sharing would the term to classifie this, i think) . also the real IP's of these TCP servers, are not transparent to the packets, they are, only to the user and the router must rebuild the packet fields and frame fields, then load balances to the servers. the obvious limitation, is that the above can only be done to TCP traffic. hope the above helps regards ""Peter P"" a icrit dans le message de news: [EMAIL PROTECTED] > Can anyone give me a real world example of why you would need to consider > using TCP Load Distribution. I am summarising slightly but > TCP Load Distribution seems to be a method of using a single IP address (the > Global Inside Address)inbound; which is handed off to different devices on > the inside. OK. > > If this is a fair description I can see that this would be useful for load > sharing amongst internal servers. IE maybe if an increasing number of > Internet customers were accessing your resources - on line shopping whatever > - that you might want to spread (balance) the load among several identical > servers? So is 'loosely' like HSRP (not to do with redundancy so much) but > conceptually in that there is a VIRTUAL entity that supports multiple > physical entities (servers) to enable the load distribution amongst these > 'real' devices. > > Therefore is the case that the real devices don't need to be network devices > - they would most likely be UNIX (typically Internet facing) boxes of some > sort? Any response to clarify my muddle thinking much appreciated! Apologies > for dumb question. Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=62094&t=62088 -- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
Re: NAT and TCP Load Distribution [7:62088]
Thanks for that. Is my 'real world' example right in broad conception ? Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=62096&t=62088 -- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
Re: NAT and TCP Load Distribution [7:62088]
Web server farm for ur ebiz site, u assign private address to the servers, and use only one public ip to access them all. ""Peter P"" a icrit dans le message de news: [EMAIL PROTECTED] > Thanks for that. Is my 'real world' example right in broad conception ? Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=62102&t=62088 -- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
Re: NAT and TCP Load Distribution [7:62088]
sorry i didn't enphasis on a point that from i've written could be misleading, (the router must rebuild the packet fields and frame fields, then load balances to the servers.) the router will actually, nat the ip, forward the packet in which case a rewrite happens to the frame's mac. regards, ""Juntao"" a icrit dans le message de news: [EMAIL PROTECTED] > Web server farm for ur ebiz site, > u assign private address to the servers, and use only one public ip to > access them all. > > ""Peter P"" a icrit dans le message de news: > [EMAIL PROTECTED] > > Thanks for that. Is my 'real world' example right in broad conception ? Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=62106&t=62088 -- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
RE: NAT and TCP Load Distribution [7:62088]
Right. OK. Therefore - lets say we have 6 servers in our farm each issued with RFC 1918 private address numbers. These get mapped to a singular global inside address. When our customers start hitting the servers - they get connected to one of the 6 servers in a round robin sequential fashion. This seems to imply that the same resources need to exist on each server ? Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=62242&t=62088 -- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
RE: NAT and TCP Load Distribution [7:62088]
Yes "Peter P" @groupstudy.com em 31/01/2003 09:56:20 Favor responder a "Peter P" Enviado Por: [EMAIL PROTECTED] Para: [EMAIL PROTECTED] cc: Assunto:RE: NAT and TCP Load Distribution [7:62088] Right. OK. Therefore - lets say we have 6 servers in our farm each issued with RFC 1918 private address numbers. These get mapped to a singular global inside address. When our customers start hitting the servers - they get connected to one of the 6 servers in a round robin sequential fashion. This seems to imply that the same resources need to exist on each server ? Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=62244&t=62088 -- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
