On Apr 7, 12:09=A0pm, "GlenB" <batchelg.remov...@[EMAIL PROTECTED]
> wrote:
> Mark,
>
> =A0 What about Visage.DRS? I'm surprised Ross hasn't jumped in here
alread=
y.
> I've never test driven it, so I can't sup****t it either way.
>
> GlenB
>
> "Mark Brown" <Mark_Br...@[EMAIL PROTECTED]
> wrote in message
>
> news:OZ6dnZdPL_Zi0WzanZ2dnUVZ_hKdnZ2d@[EMAIL PROTECTED]
>
>
>
> > "JJCSR" <JCro...@[EMAIL PROTECTED]
> wrote in message
> >news:190f7226-0670-459f-8e79-8cd3ac7675d1@[EMAIL PROTECTED]
> >> Is anyone out there experienced with "hot-backup", using D3/Linux?
> >> If so, can you re****t pros and cons? =A0 Has anyone who might be
using
> >> hot-backup found any reason to use transaction-logger in conjunction
> >> with hot-backup?
>
> >> Thanks,
>
> >> Jim Cronin
> >> Director MIS
> >> Kittery Trading Post
>
> > Jim,
>
> > Hot backup IS transaction logger. =A0In a nutshell, here's what
happens.=
>
> > You do a full save of your system. =A0You start logging transactions.
>
> > You take your filesave to a second machine, one with at least as much
di=
sk
> > and i/o speed.
>
> > You do a full restore on this backup machine. =A0You define a slave
mast=
er
> > relation****p between the two boxes via the software provided. =A0Then
st=
art
> > hot-backup.
>
> > Hot backup takes the transaction log from the master machine and
> > distributes it to the slave machine(s), keeping the critical files
> > updated.
>
> > With reasonably fast lan speeds and reasonably fast hardware, the two
> > boxes are almost virtual twins. =A0Which is what you want. =A0If the
mas=
ter
> > box goes down, you switch over to box two and your back up and
running.
> > The only data it is possible to lose is whatever is in the transaction
> > logger pipeline, so the idea is to keep it to a minimum by "tuning"
syst=
em
> > parameters.
>
> > Some fairly large players use hot backup. =A0Auto Trader Magazine pops
t=
o
> > mind.
>
> > It can take hours to restore a large system and additional hours to
> > restore the transaction logs.
>
> > If it is reasonably monitored (slave and/or master phantoms die
sometime=
s)
> > and don't let the logger queue get too big, I've never heard of anyone
> > with a better solution.
>
> > --
> > Mark Brown
> > Sr. Software Engineer
> > Drexel Management Svc Inc.
>
>=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0484...=A0=A0=A0=A0=A0=A0=A0(cel=
l)- Hide quoted text -
>
> - Show quoted text -
Too busy at present for detailed response (20hr days & all that), but
you are right, Visage.DRS was designed to overcome failings in hot
backup that were picked up by major listed company by an external IT
audit, the biggest being trans****t failure, auto-recovery and ability
to replicate to multiple backup servers.
They now have a clean bill of health :-)
No application changes are required, and replication is provided at
the account level. We have tested with D3/Linux & NT, and have
recently had some interest from U2 sites, which is surprising given
the fact I thought the IBM offering this regard was wuite good,
but .... stay tuned
|
