[Date Next]
[Thread Prev]
[Thread Next]
Re: scheduled power outage for NPB (fwd)
- To: "Carl Malamud [IMS]" <carl@radio.com>
- Subject: Re: scheduled power outage for NPB (fwd)
- From: Yoichi Shinoda <shinoda@jaist.ac.jp>
- Date: Mon, 13 May 1996 23:36:50 JST
- Cc: not-park@park.org
- In-reply-to: Your message of "Thu, 09 May 1996 14:40:44 -0400"
- References: <199605091840.OAA02828@trystero.radio.com>
Carl wrote:
>A manual switch of the primary place is the first step. Next
>is to put in the engineering we had discussed previously that
>allows a hot swap between the two. The only things that
>appear tricky are:
>
> 1) DNS (point to the right place and don't point to a down machine)
> 2) Guestbook and updates (town is currently the master on this)
> 3) Proper redirects at the bandwidth.cgi level
>
>I think getting this to work properly would be a very good
>enhancement to the park. In an ideal world, we could even
>have Japan and Town loadbalance for each other.
For the guestboot and updates, I might want to propose a brand new system based on the method we used in our first Internet disaster drill to distribute a safety record of a person received at one service site to other sites for later retrieval at any of the service sites: the NetNews servers cross feeding each other (we used a complete mesh feed in the drill).
The advantage of this approach is that there's no 'master' site. We don't have to worry about a master site going down. We even can tolerate multiple sites going down simultaneously. Once a new guest record is inside the news system, it will eventually be delivered to other sites unless something goes extremely wrong.
Bad news is that we have to install the news system, and implement a new guest book system that interfaces to the news system.
--- shinoda
References:
- Re: scheduled power outage for NPB (fwd)
- From: "Carl Malamud [IMS]" <carl>
[Date Prev]
[Date Next]
[Thread Prev]
[Thread Next]