On Sat, Feb 23, 2008 at 6:55 PM, Shawn Everett <[EMAIL PROTECTED]> wrote:
>  I'm looking to buy 1 or more cable and DSL connections, plug them into this
>  box and have some level of speed/redundancy created as a result.

  The "right" way to do this is with traditional IP routing.  Done
right, traffic can go out either pipe, and come back though either
pipe, and will work for everything.  You get an IP network assigned to
you, and then have both pipes routing to and from that network.  If
both pipes are from the same ISP, this can be done fairly easily
(assuming the ISP is cooperative and competent).  If both pipes are
from different ISPs, you need to get an AS (Autonomous System) number
and run BGP (Border Gateway Protocol).  Running your own AS tends to
be infeasible for most smaller organizations.

  There are other ways, but they have limitations.  The fundamental
issue is that if you have two pipes that are not cooperating on
routing, the rest of the world has no way of knowing those two
different pipes go to your single location.

  So you can, for example, get a device that balances your web
browsing traffic between the two pipes.  It will do that by sending
one request out one pipe, and the next request out the second pipe,
and the third request out the first pipe, and so on.  (Maybe it
manages the ratio more intelligently, but the basic concept remains
the same.)  But no single page request will be any faster than the
single pipe it is on.

  Incoming traffic -- say, to a web-server you host in-house -- is
even harder to manage, because again, the rest of the world doesn't
know you have these two pipes.  You can use round robin DNS, but it
won't react at all to unbalanced loads or failures in one of the
pipes.  You can use dynamic DNS update, but that has issues with
caching.  You can try and use a short TTL to override the caching, but
that doesn't always work in practice.  It can be better than nothing,
but be prepared to have it not work for some
clients/customers/people/whatever.

  Some protocols have their own failover mechanisms built-in.  For
example, email.  Set up multiple MX records and SMTP can automatically
handle routing incoming mail if one pipe is down.  Some VPN clients
have similar functionality.  Another option for VPNs is to present the
user with a choice of VPN gateway, and give them instructions that if
one does not work, try the other.

-- Ben

~ Upgrade to Next Generation Antispam/Antivirus with Ninja!    ~
~ <http://www.sunbelt-software.com/SunbeltMessagingNinja.cfm>  ~

Reply via email to