[ZODB-Dev] Using ZEO to solve problem with slow connection

[sean.upton@uniontrib.com]

DirectoryStorage will work for one-way replication, but I think this =
assumes
that the replica is offline until needed (i.e. you can use for hot-ba=
ckup
using resource takeover strategy - i.e. have heartbeat resource scrip=
ts stop
replication cron jobs, run checkds.py on the replica, start your seco=
ndary
ZSS process, takeover the IP of the primary - this also requires, gen=
erally,
that the hot-backup likely be on the same LAN unless you put the
responsibility of the switch on the ZEO client - which is certainly p=
ossible
in ZEO 2).  I think for replication that works one-way, but requires =
the
replica to be active/online may require replication of transactions a=
bove
the storage level, which I think Zope Corp's commercial ZRS product c=
an do.

Sean

-----Original Message-----
=46rom: Steve Alexander [mailto:steve@cat-box.net]
Sent: Friday, April 04, 2003 5:16 AM
To: Roch=E9 Compaan
Cc: ZODB-Dev
Subject: Re: [ZODB-Dev] Using ZEO to solve problem with slow connecti=
on



> First, would it make sense to zip marshalled objects before they ar=
e
> transferred between client and server?

You can use a ssh for this. The -C option will compress the data, and=
=20
you can specify a NULL cipher if you don't need to encrypt the data.


> Secondly, how difficult will it be to modify ZEO to not delay clien=
t
> reads if the record exists in the client cache. The server then pus=
hes
> modifications to the client when they occur and the client updates =
the
> record in its cache as soon as it is available locally.
>=20
> Any other ideas will be appreciated.

If you only need one-way replication, look at Directory Storage.

--
Steve Alexander




_______________________________________________
For more information about ZODB, see the ZODB Wiki:
http://www.zope.org/Wikis/ZODB/

ZODB-Dev mailing list  -  ZODB-Dev@zope.org
http://mail.zope.org/mailman/listinfo/zodb-dev