HP Lefthand vs EVA 4400 and SVSP.?

[MAAH]

Hi all,

is there an easy way to compare these too concepts.. when to use it or not.?

Will P4000 units support the path's the way as the HP SVSP will do..? = nerver lose them.?

With HP svsp you can skip SRM on the ESX Hosts
and soo..

[René Loser]

Hello,

Yes, both are "always on" solutions. So ESX doesn't recognize if one site fails. No SRM is needed for both solutions.

Big difference below: FC SAN vs. iSCSI

best reagrds,
reNe

HP PreSales Storage

[MAAH]

Ok.. So basically the biggest differncis is between the systems is FC and ISCSI..?

No other funtionallity..snapshot, busniess copy and soo..?

the price..? for a solution with two EVA 4400 and SVSP and 20 TB storage.. is the
P4000 solution mutch cheaper.?

No FC, No Cache, How many units do you need
to reache the same perfomance as the EVA Systems with FC..?

[MAAH]

One other thing.. Is there something else to think about if you have an mixed enviroment with ESX and normal OSTD Windows Clusters..?

[teledata]

Remember that the performance requirements are tied closely to the # of disks (typically with any SAN solution).

You are still looking at (# of physical disks x IOP/disk) - RAID overhead.

There is caching on the P4000 series just like the EVA.

Also although a single FC connection is more than 1Gb, the P4000 series scales as you add more modules, so # of modules x (2) GbE (minus network overhead) = total throughput capacity.

If I remember correctly in the EVA series you choose x number of disks per LUN (essentially create multiple RAID arrays), the P4000 always stripes across all disks across all nodes in a cluster.. There are advantages/disadvantages to both methods depending on your utilization.

[MAAH]

Ok.. Don't you have the possiblity to add
10Gb interface to the P4000.? But it has to
more than that.. 8Gb FC on EVA's and also the protocol it self must be different than running with ISCSI and a normal network swtich.?
Also on the EVA's dont't you have all the FC-Drives to one diskgroup and then add Virtual Disks to it..?

[Mark...]

Hi,
To try to clarify some of your points above.
Even the latest EVA's are only 4Gb host ports. Fabrics can be 8Gb.
The LHN product is 1 or upgrade to 10Gbe.
EVA's put disks in disk groups and then you create a virtual disk from a disk group - not as described above.
The LHN stripes across nodes none/two/three/four multiple writes of the same block for protection.
The EVA does snapshots, snapclones, mirrorclones, containers CA and al disks are full provisioned but you can use something called DCM to dynamically extend/shrink LUNS for some OS's.
The LHN does snapshots, smartclones and remote copy - the way that it does this IS NOT the same as an EVA!
You need licences for everything for an EVA.
LHN comes with all lice as part or package.
There are many more similarites / differences between the products....
There are advantages and disadvatanges to both you need to decide more closely on what your requirements really are and how much money you have !! IF money is the deciding factor then LHN is probably the way to go but it may not be....
Mark...

[MAAH]

Sure.. BUT my biggest issue is if the P4500 will work in the same way as SVSP. Because an EVA installtion with 2 EVA's with busniess copy without SVSP will not be an "always on" system if you look to ESX and MS Cluster (without Cluster extension)..?

Right.?

Will P4500 multisite handels the
path's the same way as SVSP for EVA.?

Because if it looses one path over the
ISCSI to the VMware ESX host it will
lose the disks to thoose VM's and LUN's
working over that path., or..???
And then you need to reconfig/scan/reboot the ESX host so they will use the other path instead..?

[kghammond]

We are planning to use LH across sites, we don't have this all setup yet. We have 10G between sites. We are not planning to use multi-site. Although Multi-site gets you local usage of your LH's at that site, it does not support MPIO round-robin currently in vSphere.

Without MPIO, you cannot scale an individual VM workload beyond one (NIC) and one LeftHand node. This is due to an iSCSI target session being persistent on a NIC. MPIO round-robin resolves this problem, but as I stated MPIO round-robin does not work with Mulit-Site.

Your aggregate ESX workloads will span LH nodes so as a whole you can scale, but individual workloads for a individula LUN cannot scale without MPIO.

Our plan is to just trunk our VLAN's over to our DR site and then let the LH's think they are on the same LAN. The downside to this is we will only be able to scale to about 6-8 clusters of LH's before we start filling up the 10G pipe.

We are hoping that LeftHand will release a MPIO stack for vSphere that will allow localized MPIO traffic in a multi-site cluster, but 3rd party MPIO providers require vSphere Enterprise Plus licenses...

I hope this helps a little in your decision process.

[MAAH]

Thanks.. Finally we are getting some where.. ;-)

This means that SVSP is the king! When
using SVSP I will be able to use two
EVA's or any FC SAN soulution and this will
be presented as one system to the hosts even if there are on different sites failover will no longer be an issue for me.. :-)
this of course inside one SVSP/LSI domain.