ChristianZ
Beanieboy's concerns are not if the product works, it is regarding the use of two across a 1Gb link between sites.
From what I read, cmanucy configuration was at the single location through a 3COM 4500G switch, results could vary for a site to site configuration.
From what I looked at on the Web, the product looks impressive, I still think it would be wise to perform testing on the intended solution before committing $$.
As the Author of "Open unofficial storage performance thread" do I dare question your logic? 
(attempted humor)
best regards
Bernie.
ChristianZ: NSM2120 actually is HP
beanieboy:The 1Gbit link between sites is one half of theoretical maximum throughput you can get out of those boxes. The more your I/O traffic will contain random operations (database-like), the less throughput will it require. There is another reason for two links, however - redundancy. You have perhaps considered it already, just mentioning it here for safe.
And to make you comfortable, there are plenty of SAN/IQ installments running VMware, just ask LeftHand for references if you like. I work for a LeftHand reseller in Czechia, have deployed 3 SAN/IQ+VMware installments with Dell hardware with no troubles.
--
I think there is some confusion about LeftHand's solution.
LeftHand does network raid. 2 or more NSMs should be deployed for redundancy. You need 3 units to do 3 way replication, etc. Adding more NSMs increases your storage as well as your I/O. A virtual management IP address is created (quorum is created in LHN terms). This becomes your Discovery target. Every NSM can respond to this IP address so it's fully redundant.
This NSM cluster is comparable to a single unit. I don't know if I would be comfortable with a single NSM solution, but that's a personal preference.
Mirrorview is comparable to RemoteCopy from Lefthand. RemoteCopy is asynchronous only. It uses LUN snapshots to replicate delta changes from a LUN at one site of NSMs to another site of NSMs. It's really no different than most of the iSCSI vendor replication options.
I think the network raid design of LHN got confused with Mirrorview/RemoteCopy. The 2-way, 3-way replication is a configuration option in the network raid. Data blocks are written across NSM units in a cluster...2 way, data is written to two NSMs in a cluster. 3 way, 3 NSMs, etc. The higher the replication, the more data protection you have. It also increases your storage (1 block x 3 = 3 blocks of storage on 3 NSMs).
As for the 1Gig link between sites....I don't know. Is this DR site next door? 
If he means that the connections to the switches are 1Gig, yes, this is almost a requirement in iSCSI deployments these days. You can of course etherchannel the links for failover and higher throughput.
I heard from an LHN rep a few months ago they are now upto 7 way replication in the network raid (don't confuse their 2 way, 3 way, 4 way ... replication with other vendor's replication features) and now support this replication (including fail over prioritization) across MAN links. I don't know what b/w and latency requirements are needed for this to be successful.
I've run LHN in production. The performance was on par with every other iSCSI solution I've used. Your mileage of course varies based on your environment and how much $$ you can spend. I really liked their solution. The management interface seems sort of 'low-end' but it was simple and easy to use. One negative was it required you to be on the iSCSI network or have it routable from your data network, but I usually isolate my iSCSI networks from my data networks. The product scales as you expand the number of nodes, very cool product but definitely different from the other major iSCSI vendors.
They released a couple different virtual appliances of their SAN/IQ software if you want to play with the solution more. I use the laptop vsa in my laptop ESX Cluster.