[linux-lvm] High level architecture (Long)

[Greg Freemyer]

Steve,

You are right about that being a long e-mail.  I have a Compaq (HP) background, and they call this a Disaster Tolerant SAN (DT-SAN).

Your also right about DT-SANs requiring deep pockets, and even with Linux I think it will take some fairly deep pockets.

HP has a consulting business specifically targeting DT-SANs. I don't know if they support Linux or not.  (It used to be VMS and Tru64 only.) 

http://h18000.www1.hp.com/products/storageworks/solutions/dtsan/

I have no idea where you will get a overall interest in your DT-SAN project 

Maybe someone at OSDL (http://www.osdl.org/) or on the linux-ha (heartbeat) list can give you a recommendation, or maybe you will have to start your own DT-SAN project

(A little more below)

 >>  Hi all,

 >>  1) A highly available, peer replicated (i.e. active-active),
 >>  geographically separated (n locations where n=2 and n>2), virtualized,
 >>  and Backed-up SAN solution.  That's a tall order.  Companies such as
 >>  FalconStor provide some of this, but at an exceptionally high price.

You do know how expensive that is from a communications perspective don't you.  The commercial offerings I have seen require a T3 or larger.  

I have never priced one, but I assume your talking the 10's of thousands per month if you are going between major cities.

I have priced a 100mbit connection across the north end of Atlanta.  That actually wasn't so bad.  Around $2K/month IIRC, but I think that is a rare situation with Atlanta having excess fiber in the ground.

 >>  2) SCSI over IP or something similar is desirable so that application
 >>  servers can use the exported virtualized storage (either locally on the
 >>  same LAN or remotely over a WAN).  Security here is a big ?.

I have tried the Intel iSCSI demo initiator/target pair and they seem to work well.  (I was just curious, so I did not do much testing.)

http://sourceforge.net/projects/intel-iscsi

The target software allows any Linux box to export a drive (or LV) to a Linux computer running the initiator.  

Since iSCSI was just formally approved, there should be a flurry of real targets, but I'm not sure thats what you want anyway.

Another competitive product is HyperSCSI, but I don't think HyperSCSI supports routers, so it is purely LAN.  (ie. It is a pure ethernet protocol, not a TCP/IP protocol).

 >>  3) A highly available virtualized application server pool that supports
 >>  automatic failover to a remote location.  A combination of Clusters,
 >>  LVS, user-mode-linux and replicated SANs can provide a potential
 >>  solution.

I follow clustering, and I don't know any split clustering solutions for Linux.  

One big issue is when the VIP moves, you have to update the routers to send traffic for the VIP to the new location.  This can be done, but it takes coordination with your ISP, or a lot of your own infrastructure.