Date: Wed, 06 Feb 2036 22:28:23 -0800 (PST) From: Simon Shapiro <shimon@simon-shapiro.org> To: Tom <tom@sdf.com> Cc: hackers@FreeBSD.ORG, Capriotti <capriotti@geocities.com> Subject: Re: SFT Message-ID: <XFMail.360206222823.shimon@simon-shapiro.org> In-Reply-To: <Pine.BSF.3.95q.980106113011.22841B-100000@misery.sdf.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 06-Jan-98 Tom wrote: > > On Tue, 6 Jan 1998, Capriotti wrote: > >> Does anyone have any news about System Fault Tolerance under Free ? >> >> Like what Novell has, from mirrowed disks to mirrowed servers ? > > Mirrored disks can be done with ccd. > > Mirrored servers are basically what Unix types call a cluster. On > Novell this is easy, because Novell boxes are basically just file > servers, > but a Unix box could be doing many different things. Migrating tasks > from the failed system to the working system, and assumption > of the IP traffic, is difficult. I am working on such a system. Full time. Our solution os a bit different from Novell's. We are aiming at true non-stop, full utilization model (no standby). Simople mirroring is an expensive and non-scalable solution to data persistance, but has very little to do with high availability. To throw a monkey wrench at this discussion, please consider that the most common (by far) cause of system service failure (crashes are severe failure to service, right?) is software bugs. Far behind it lag cables and connectors, yet again far behind these are disks and memory. It is my hunch that DRAM fails as much as hard disks do (per storage unit per time) or even more. This flies in the face of software solutions to high availability. [ Obviously this is exactly one half of the story. You come up with the other half :-) ] ---------- Sincerely Yours, Simon Shapiro Shimon@Simon-Shapiro.ORG Voice: 503.799.2313
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?XFMail.360206222823.shimon>