Messages in this thread | | | Subject | Re: PATCH: Raw device IO for 2.1.131 | From | "James H. Cloos Jr." <> | Date | 16 Dec 1998 17:26:55 -0600 |
| |
>>>>> "Martin" == Martin Tessun <martin.tessun@class.de> writes:
Martin> Perhaps this is not the right topic, but I want to say a few Martin> words to Oracle:
Martin> Oracle uses raw-devices, because they thought it was the Martin> fastest solution. In the meantime they have support for Martin> NetApp-Filers. So Oracle goes through the Net (may be private Martin> for being faster) and then runs into the Cache of the Martin> NetApp-Filer. So Oracle has no overhead for Filemanagement, Martin> and can be faster so, and the Net-App-Filer is optimized for Martin> Disk-IO and is thus faster.
Martin> The Bottleneck in this configuration is the net. But still Martin> Oracle support this solution and say that it is indeed faster Martin> than the Raw-Device solution if you have a fast net.
<advocate client=devil> The difference between using the toaster as the backing store vs Linux's page cache is that the former is `safe', in that, once the data hit the toaster and are acked, the dbm knows that they are not in danger of loss due to power failure. Were Linux's page cache transaction based and routed through battery- backed nvram, the comparison would be more valid.
In short, they want to be assured that the data made it to non- volatile storage. </advocate>
-JimC -- James H. Cloos, Jr. <http://www.jhcloos.com/cloos/pgp_public_key.txt> <cloos@jhcloos.com> E9E9 F828 61A4 6EA9 0F2B 63E7 997A 9F17 ED7D AEA6
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |