[ZODB-Dev] ReplicatedFileStorage (Re: BerkleyStorage)

[Toby Dickenson]

On Sun, 29 Jul 2001 18:50:33 -0300 (BRT), Christian Robottom Reis
<kiko@async.com.br> wrote:

>On Fri, 27 Jul 2001, Toby Dickenson wrote:
>
>> Yes, that was me
>> http://www.zope.org/Members/htrd/ReplicatedFileStorage
>
>Toby, does this work seamlessly under the ZEO?

Yes.

>And would you estimate what
>the performance hit is?

On writes, there is an obvious cost of transferring the new bytes over
the network to the replicas. You also have to wait for two network
round-trips at each transaction commit. Additional overheads *should*
be minimal.

The extra cost of reads is negligible; all reads are satisfied
locally.

Im currently using ReplicatedFileStorage on one of our high-write
intranet sites. It is replicating over a LAN into a low-end machine,
and you cant feel a performance difference.

Another question is the performance impact on the machines holding the
replica... This too should be tiny.



Toby Dickenson
tdickenson@geminidataloggers.com