NFS in the cloud

so I have several questions for the folks who have been using the cloud:

1. is anyone else running into issues when taking the default API URL when running service mauinfs configure on a redhat with public IP? It seems I can't reach the API host (compute.emccis.com - which is the default URL). Anyone with a clue?

2. does anyone have a recommendation what to do for Windows host in terms of NFS client setup?

I am tempted to use a redhat small image to mount NFS storage and expose it as CIFS - except samba reliability is not worth the bytes it burns on disk.

Other solutions are axeNFS which I test drove and wouldn't spend $0.01 on it (pretty much unusable product), and Microsoft Unix Services - for which I seem to need a dedicated WIndows 2003 Server box and AD domain to map users.

I wish there was a new Cellera device fronting all this storage - so you can hit it using Unix NFS or Windows CIFS or access block device say via iSCSI (the new Cellera devices are both SAN and NAS devices all in one with high availability guarantees)...

3. has anyone used the NFS server for the database storage? Before anyone jumps at my throat for running database off of a NAS, I have EMC core paper in hand that shows superior Oracle performance off of NFS NAS (yes, it was a Cellera device). My experience is telling me since last century that running database off of a NAS is just a raw deal (you get 1/10 of the performance when same workload profile is ran against a database configured with SAN storage). However, it appears that a few NFS tuning tricks here and there, and NFS NAS is a decent choice for a database. The $64,000 question is if anyone has tried to run database against the Atmos NFS service - I figured I'd ask before trying it myself.

4. has anyone tried running an active-active database (e.g. Oracle RAC) in the cloud? I wrote up extensively on this, and in general Oracle RAC, besides the heartburn with NAS heads, has additional heartburn because of the drifting clock in vmware environment. Cliff notes: it's by design as vmware virtualized the clock and on Unix you count passage of time by counting ticks (interrupts) and by design you are bound to miss a few ticks and your clock will drift, causing the ntp to create a drift file that automatically adjusts the clock - but then you overcompensate as your actual scheduling varies under vmware just by the nature of the workload, and you end up with an erratic clock that switches back and forth between running too fast and running too slow, causing complete chaos with Oracle clusterware (and it would in any clusterware). I figured an NFS NAS head might resolve this problem, but haven't thought it through. So again, figured I'd ask before trying it out. And yes, I know Oracle does not support RAC in vmware environment.