|
I currently have one site that has a mix of Windows NT and Solaris web
servers (currently 4 in total) providing located in Toronto, Canada. In the
works is another site that will likely be located somewhere in the USA (for
strategic business reasons??) and will likely have a similar mix of web
servers. The Toronto site and the US site will not be connected via the
same ISP so I must assume the I will be dealing with separate subnets. When
the US site comes online, later this year, I would like to provide load
balancing between the two geographically separated sites.
Will the users have a satisfactory response time/uptime going through the
director to these machines elsewhere?
Currently we host our machines at an ISP with a T3 into the net - I think
they are connected to Sprint. We would expect to find an ISP to host our
proposed US site with at least the same bandwidth (or more). I'm hoping
that the LVS solution will be stable and not require many, if any, kicks.
Approximately 10% of the cpu cycles will be https, most will be http with
java apps that will require a persistent connection back to a servlet on a
particular web server.
Simplicity/maintainability is a priority.
Ed
-----Original Message-----
From: Joseph Mack [mailto:mack@xxxxxxxxxxx]
Sent: Monday, August 07, 2000 3:56 PM
To: Ed Wiseman
Cc: Lvs-Users (E-mail)
Subject: Re: FW: Multi-site LVS solution
On Mon, 7 Aug 2000, Ed Wiseman wrote:
> I want to load balance HTTP/HTTPS. I don't want to go the two tiered
route
> if I don't have to. According to the documentation and the diagram
> http://www.linuxvirtualserver.org/VS-DRouting.html, VS-DR needs to have
the
> Real Servers on a local subnet. I have two subnets, one can be local to
the
> VS but the other is physically and logically elsewhere on the planet. Can
> VS-DR handle this between the two sites?
Can you tell me more about your restrictions? At the moment it appear that
you have 2 machines on a local net and 2 machines somewhere else on the
planet.
Will the users have a satisfactory response time/uptime going through the
director to these machines elsewhere? Should they be a part of the
solution at all (you're going to have to call someone up to kick them if
they hang etc)? What fraction of your CPU cycles are going to be https?
If this is a production system you want simplicity/maintainability before
you want performance (no one complains that much if the response time is
slow, but people really care if it's down).
Joe
--
Joseph Mack mack@xxxxxxxxxxx
|
