Gluster users - Jan 2009 - Server-Side AFR+Client HA vs Client-Side AFR

What are the pros and cons using Server-Side AFR+Client HA. To me it sounds
like a winner combo?
What about comparing serverside afr+client ha to client-side afr? What are
the main differences there?
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20090126/3e049969/attachment.html>

At 10:40 AM 1/26/2009, Karri Niemel? wrote:
>What are the pros and cons using Server-Side 
>AFR+Client HA. To me it sounds like a winner combo?
>What about comparing serverside afr+client ha to 
>client-side afr? What are the main differences there?
client side AFR has advantages in that you get HA for the clients.
if the client handles the AFR and one of the 
servers is down, then it will continue operating from the other server.

with Server Side AFR, you have a benefit of 
optimized network operations, but this depends on your configuration.

Typically servers are connected via a 
higher-speed network than clients are, so a 
client updates a file to a server, the server 
then updates quickly to the replica.

You also have a reduced liklihood of 
misconfiguration with server-side replication.
if you have a bunch of clients and one of them is 
misconfigured, you will have unpredictable 
results and files may not be replicated to the correct servers.
this is especially likely when you grow/change 
your configuration--updating a couple servers is 
trivial, while updating dozens of clients all at 
the same time would be a challenge.

So, really it just depends on your specific situation to know which is best.

At 10:40 AM 1/26/2009, Karri Niemel? wrote:
>What are the pros and cons using Server-Side 
>AFR+Client HA. To me it sounds like a winner combo?
>What about comparing serverside afr+client ha to 
>client-side afr? What are the main differences there?
client side AFR has advantages in that you get HA for the clients.
if the client handles the AFR and one of the 
servers is down, then it will continue operating from the other server.

with Server Side AFR, you have a benefit of 
optimized network operations, but this depends on your configuration.

Typically servers are connected via a 
higher-speed network than clients are, so a 
client updates a file to a server, the server 
then updates quickly to the replica.

You also have a reduced liklihood of 
misconfiguration with server-side replication.
if you have a bunch of clients and one of them is 
misconfigured, you will have unpredictable 
results and files may not be replicated to the correct servers.
this is especially likely when you grow/change 
your configuration--updating a couple servers is 
trivial, while updating dozens of clients all at 
the same time would be a challenge.

So, really it just depends on your specific situation to know which is best.