[Info-vax] SCS over IP

[Stephen Hoffman]

On 2013-10-30 12:54:37 +0000, 11brvo at gmail.com said:

> Is there any advantage to using SCS over IP as opposed to hard-wiring a 
> two node Integrity cluster together for SCS traffic?
> 
> The two nodes in question are rx2800s and are geographically co-located 
> with minimal distance between the two.  However, future plans might 
> entail the addition of a third node into the cluster with this 
> additional node being geographically distant from the other two. (DR 
> purposes.)  Both rx2800s run v8.4

Any advantage?  Here?   I don't see one.

SCS over IP is effectively a software replacement for Ethernet bridge hardware.

Native SCS is faster and lower overhead, and will be the preferred path 
even if SCS over IP is configured.

If and when that "future plans might..." status becomes "immanent real 
physical hardware" status (and not some variation on "the end is nigh" 
status, or fodder for the latest episode of "Port That Code!!") and if 
and when you do not have an Ethernet bridge available between the two 
sites, then you can look at adding SCS via IP.

For the remote connection, you're going to need a network connection 
commensurate with the volume of likely-shadowed data, and the rest of 
the DT baggage that gets discussed, too.  For various folks, shipping 
shadowset clones or tapes or such is more cost-effective.

Could somebody justify SCS over IP here?  Probably.  There are some 
other details that may or may not be of interest, such as SCS over IP 
supposedly also encrypting, though I don't know of any third-party 
studies of the particular implementation.  But on an Ethernet, native, 
classic, unencrypted SCS will be preferred when that's available.  Or 
somebody will suggest not having to reboot the cluster when adding the 
DT host, but in my experience that's fairly rarely an absolute 
constraint for sites running two bottom-tier servers.




-- 
Pure Personal Opinion | HoffmanLabs LLC