[bind10-dev] Persistence vs. shared memory

[Michal 'vorner' Vaner]

Hello

On Mon, Sep 05, 2011 at 07:41:39PM +0200, Florian Weimer wrote:
> > It seems we already decided to go with the shared memory model, so
> > it might be off-topic, but I asked the friend I mentioned, about
> > persistence, just out of curiosity. He sent me some lecture notes
> > mentioning work of Sleator and Tarjan (but it doesn't contain exact
> > reference, unfortunately). After reading through them, I can think
> > of three ways of implementing the persistence for the DNS tree, so
> > I'm sharing it just in case someone might be interested.
> 
> I'm not sure what this is about exactly, but there's also RCU
> (a variant of persistence).

Sorry, I forgot to mention, we talked about how we will do updates in the
in-memory data source for storing authoritative data.

RCU in the strict sense copies everything, which is too expensive, considering
the data might be several GB.

In less strict sense, much everything described in the email could be considered
RCU, but the difference is how it is done. Maybe I just used the word
„persistence“ because I come from more theoretical background and we both think
the same.

> Are occasional pause times acceptable, or is there a stringent
> real-time requirement?

They are highly undesirable (people usually don't want their server to stop
answering queries). And while the expected query load is higher than update
load, it should be able to do few hundreds or thousands updates per second. We
probably can batch them together a little, though.

Thanks

With regards

-- 
echo '*' > 'rm -rf ~'; . *

Michal 'vorner' Vaner
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <https://lists.isc.org/pipermail/bind10-dev/attachments/20110905/3fe62733/attachment.bin>