VMware Cloud Community
whiskyman201110
Contributor
Contributor

about to buy first SAN - HP Lefthand or DELL EQualogic?

Hello everyone,

we are currently in the market for our first SAN. Very close to ordering in fact.

We have narrowed it down to HP P4500 multi-site SAN, or the DELL equalogic ps4100x.

Not having worked in the storage area much before, we're unsure which one to go for.

Heres what our structure will be within the next few months:

2 sites

2 esx hosts one site, 1  esx 3.5 host and 1 esxi 5.0 host.

2 esx hosts other site.1 ESX 4.1, 1 esxi 5.

NO Vcentre currently in place but I plan to deploy this/next quarter.

HP have proposed 28.8tb multi site SAN p4500.

Dell have proposed 2 X 7.2tb Ps4100X's (one for each site)

I have about 3TB of data spread between both sites.

Dell option is alot cheaper, plus that includes 4 X 10G switches.

HP option includes the VSA licences though.

At first glance it would appear the HP multi site SAN has some very nice HA options utlising "network Raid", however we have roughly 4mb connecting WAN sites. Is it do-able over this kind if bandwidth in the first place?

Also does anybody have experiance managing these systems, or any input Ie pros and cons of both these SANS?

Any advice greatly appreciated!

Cheers

Tags (3)
Reply
0 Kudos
9 Replies
proden20
Hot Shot
Hot Shot

I can't comment on the Lefthand or cross-site functionality of either.  What I can say is that I managed a 100TB EMC Unified shop and the 30TB Equalogic we used for iSCSI was 100x more reliable and easier to work with.  Ultimately, we used the Equalogic whenever we had the opportunity, and the EMC's sat there doing nothing.

Equalogic is a solid SMB platform.

Reply
0 Kudos
Vannus
Enthusiast
Enthusiast

While the multisite side of things would probably cope with a 4Mb link, you could find yourself loading that very quickly, I normally recommend at least 10Mb for such a solution, but since you have so few hosts it may well be fine.

It really comes down to how fast and easy you want your recovery from one site to another to be.

The Lefthand can give you a WAN HA solution if you split the VLANS between the sites (ie the same network IP range for the VM's). In that setup both sites are in a single cluster and vm's can fail from 1 site to the other. This is Synchronous replication so quite bandwidth intensive. you do have to be careful with your networking so its worth having a proper network design workshop before you start.

Hardware HA relies on a bonded pair of nics on the lefthand SAN (the box is basically a HP server with a specialist version of linux running on it).

The thing I dont like about lefthand is that they dont usually have the ability to have a separate management interface for the Controllers so they have to be managed from the SAN interfaces,also, the network raid is set up is also quite wasteful of disk capacity in my opinion. (28.8 is raw data, so thats 14.4 each site before its raided on the local SAN box. then you have to halve that to allow replica space and take some more off for snapshots etc. the capacity soon goes)

EqualLogic is definately an 'easier' solution to manage, especially if your not used to SAN's. Replication is Asyncronous for EQL so you cant do a multi-site cluster as you can with Lefthand.

To recover, you have to manually promote the lun and import the VM's. Although you could use VMware SRM to automate this for you.

This solution is also far more forgiving from a network point of view, you dont need split VLANS and separate management interfaces are possible for the controllers.

The EqualLogic has two physical controller modules in an active/passive setup so is much more robust from  a Hardware HA point of view.

Without knowing your environment its difficult to say which would work out best, it really comes down to how much you want to spend on DR capability. But my gut feeling is that the EqualLogic may be the better choice.

Reply
0 Kudos
Josh26
Virtuoso
Virtuoso

What is your server platform?

If you have a series of HP servers I wouldn't consider a Dell SAN, and vice versa.

Reply
0 Kudos
whiskyman201110
Contributor
Contributor

Thanks very much for the input Proden, Vannus.

My server platform is dell right now. I have a couple of old G4 HP380s but they will be moving onto a DELL ESX host this year.

I am personally keener on the dell option as I will also be able to use openmanage for a bit of monitoring. I'm not sure I could use openmanage to look after an HP box.

Also I aggree that the Dell option looks easier to manage. We had a demo of some of the snapshot tools that integrate with exchange and it looked very nice.

Josh just wondering why you wouldnt mix an HP san with Dell servers or vice versa. Is it bascially so you only have one platform to manage or is there specific technical difficulties that could come up in the future perhaps?

I'm not totally confident our link could cope with a true multi-site sync cluster either so async replication would perhaps be the way to go.

This input has been really helpful guys thanks.  We will be pulling the trigger on it very soon so I think I'm beginning to come to a conclusion.

Cheers.

Reply
0 Kudos
whiskyman201110
Contributor
Contributor

I'm still trying to get in my head exactly why the HP site has to be 4 nodes 2 each site, while the Dell one is only 2 nodes, 1 each site.

Is it bascially for extra space? Whats the advantage (if there is one ) of it being 4 boxes instead of 2.

Is this a reasonable summary below:?

With the HP 4 node lefthand, your network and the multi-site hp cluster handle the HA/Failover. And it would be seamless.

With the 2 node Dell solution, you would be replicating only (no cluster), so you would rely on vmware (srm or similar) to handle the failover?

why does the HP soloution need 4 nodes? Is it because of they way the cluster would function, (network Raid) that to get the same space you basically need more trays of disks?

I appreciate all your advice this is the first time weve looked at this technology.

cheers!

Reply
0 Kudos
Vannus
Enthusiast
Enthusiast

Those statements are accurate, the near instant failover is what you are paying the extra money for with the HP solution.

Two nodes at each site will give you extra resilliency.

With the lefthand there are 2 types of protection going on. The first is the Raid level inside a single node (this is standard raid accross the disks in that node, usually Raid1 or Raid5)

When you add a second node on the same site, you can setup network raid. this mirrors a lun to both nodes, so if a Node fails, the lun can continue to be presented to the servers.

when you add the remote site nodes into the network raid, that lun is also mirrored to the remote site nodes, so you effectively end up with 4 copies of the data.

example1.jpg

For the EQL box, whilst there is only 1 node each site, those node are more resilient at the hardware level. the disks in the node will be raided, and their are 2 physical controllers in each node. If a controller fails the node will switch to the other one.

If you add a second node on the local site with EqualLogic, and add it to the Group all luns are stripped between the 2 nodes, increasing performance.

Reply
0 Kudos
whiskyman201110
Contributor
Contributor

Thanks Vannus that clears it up for me nicely!

I guess my descision shall be based on how important seamless failover on the storage side is to tthe business then.

Thanks Again!

Reply
0 Kudos
Vannus
Enthusiast
Enthusiast

BTW, the instant failover technology is dependent on VMware HA, so until you have vCenter in place there will still be a manual process to recover the VM's in either case.

Reply
0 Kudos
Josh26
Virtuoso
Virtuoso

whiskyman201110141 wrote:

Josh just wondering why you wouldnt mix an HP san with Dell servers or vice versa. Is it bascially so you only have one platform to manage or is there specific technical difficulties that could come up in the future perhaps?

Cheers.

I had a HDD fail in a Dell SAN a while back and Dell refused to look at it because a DL380 was connected to it. In the end we just lied about what was connected until they shipped a replacement.

Technically problems like this are rubbish and should never happen but the reality is that they do, and I wouldn't expect anything better from HP.

There is also a management benefit. Openmanage as you suggest will manage a Dell SAN and a server, but is unlikely to be effective on an HP SAN. Vice versa.

Disclaimer: I'm now an all HP shop.

Reply
0 Kudos