Quick NVRAM questions

[EricLee]

I just wana ask several questions about NVRAM:
1.We all Know NVRAM will log the write IO,But I wonder if the log including the original Data?If NVRAM has to cache the original data,I think It's too small for the data stream incoming.
2.If the original data is cached in the RAM or NVRAM,does it mirror to the partner node?If not,when the partner node takeover the local node,then the original data cached in the RAM will lost?because the partner node can not get the original data(not log) by each other.

Hope someone can give me the answer...

[bbjholcomb]

Recently we had this explaiined to us from NetAApp. When a write request comes it's cached in system memory, the system that requested the write operations is told is was completed, a copy of this write request is written to NVRAM then this is replicated to it's partner NVRAM. At some point the data is written to disk from the main memory. If a failure over takes place the partner is complete the write request. I hope this help.

[saranraj456]

NVRAM just have the journals/logs not the actual data .

half the capacity will be usable, rest half will be used for mirroring.

[aborzenkov]

---

NVRAM just have the journals/logs not the actual data

---

Sorry, this is wrong. NVRAM contains actualt client requests as received from client.

[saranraj456]

That is weird for me.

http://www.netapp.com/uk/media/wp-7107.pdf

this WP says NVRAM caching uses 120B header for the incoming data.

is there any recent change for this or i am sound different ?

Thanks

Saran

[aborzenkov]

---

this WP says NVRAM caching uses 120B header for the incoming data.

---

please read again.

---

Consider a simple 8KB NFS write request. ... Data ONTAP merely has to log the 8KB of data

---