Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Wed, Apr 05, 2006 at 12:32:56PM -0700, Gary W. Smith wrote:
> One of the client sites that we setup we configured a matched set of
> VMWare servers. They had a need for a clustered Exchange server and a
> clustered SQL server. They also had a limited budget for what they
> wanted. Here is what we did.
>
> We setup VMWare on the two servers and created 3 virtual servers each.
> Node one was the linux-HA running DRBD and iscsi-target. Node two was
> Exchange and node three was the SQL server. Duplicated this on second
> environment.
>
> It works like a champ. Cost ~$5k for the two name brand servers, dual
> Xeons, 4gb ram and 250gb raid drives...
>
> So yes, it can be done on the same box and it can work great. I would
> recommend that you have the iscsi server running on the node that's not
> running everything else (otherwise disk performance becomes your issue).
Are you sure about this? I'd expect the require round-trip on the
network to cause a more significant performance decrease or delay
increase:
Exchange VM1 --- iSCSI, network ---> DRBD VM2
|
+--> Disk write
|
DRBD VM1 --- DRBD protocol, network, data ---+
|
+ Disk write
|
+ --- DRBD protocol, network, ACK ---+
|
Exchange VM1 --- iSCSI, network, ACK ---+
While if the iSCSI server runs on the same physical,
Exchange->DRBD/iSCSI would be faster (virtual network interface), and
the block data would only be sent once over the real network interface.
--
lfr
0/0