P4300 Clustering Query

[rianui]

Hi Gurus,

I have a query about the P4300 storage. I have a client who got a new one. It is in a set of two and each has eight fully populated hard drive slots with 300GB SAS disks. He wants to move one storage box to a DR site so that data one one can be stored on the other as well. Is this the way it works? Are the two boxes supposed to be seperated as opposed to being in a cluster? If he does this he will get less usable storage than he was expecting and that is forcing the vendor to exchange the disks with higher capacity drives.

Regards,

Rianui

[bryan_hadzik]

What network raid are the volumes set as? 

You are not allowed to change the drives out with the P4000 license

[Emilo · Accepted solution]

Hello,

It depends on what he is trying to achieve.

They are marketed in 2 nodes so that a client can set them in a 2 node cluster with the minimum recommended 2=way replication. This setup provides for a good balance of performance and redundancy.  It will make life with the P4000’s much easier as maintenance and unplanned for outages as all the data is always replicated on the other node. The P4000 was designed for this. If you client wants to separate them into 2 different sites for disaster recovery they can certainly accommodate this but you loose all the benefits of what the product was designed for. People concern themselves to much with disk space and not enough about data availability and recovery. If its mere storage space they want perhaps a box with just a bunch of disks with no management software would be a better fit.

Do not let them get bigger disks it defeats the purpose of more spindles equals better performance. The customer may really be looking for a better job of design. Maybe what they could do and you could propose is to buy another p4000 with bigger drives it is main purpose in life is offline storage?

Its difficult to really make a recommendation about anything without knowing what the goals is. When a person buys the P4000 they are really buying the software management that comes with the hardware. With the setup they purpose they are severely limiting many of the features that they are paying for.

[rianui]

Hi Bryan,

The RAID level that the client wants to use is 1+0. This slashes the capacity to half or less the raw capacity.

Do you mean that we cannot upgrade disks to 600GB drives because of license limitations? That is surely a point to note.

Regards,

Rianui

[rianui]

Hi Emilio,

He is basically trying to get more usable storage capacity than he was provided and replicate from one storage box to the other. Buying another P4000 storage is out of question due to budget constraints. The vendor quoted the storage based on a budget the client gave him but didn't get to the details :-(

I guess the focus here is wrong. However concinving the client to make changes at this point will be next to impossible. As Bryan has mentioned, we may need to get a license that will allow for 600GB disks as opposed to the existing 300Gb drives, right?

Rianui

[oikjn]

this all depends on the level of redundancy he is looking for.  If you look at the default packages, the multi-site starter package includes FOUR nodes so that there is a pair of nodes at each site.  This provides you node redundancy at each site and then full site redundancy.  IF the customer decides they don't care for node redundancy, then you can run a single node at each site, but then things get complicated....  IF the sites have enough bandwidth and low enough latency that you can run the two sites in a synchronous configuration, you can then keep network raid10 running and maintain node redundancy, but this will almost certanly come at the cost of performance since you effectively will ALWAYS be pulling data from the remote node and the latency and bandwidth to the remote node becomes the performance limitation (for this reason this is almost never done).  The other option is to run one node on each site and us asynchronous replication. The problem with that is you are now forced to use network raid 0 and you lose 90% of the benefits of the VSA system.  Forgetting about the storage size issue of each node, running one node on each site is typically a mistake.

If they currently have enough capacity on their system by using both systems with the default network raid10 setup, I would suggest they better defind exactly what they are trying to achieve for DR.  Most likely, they should have some VSA licenses that came with the package they bought.  Using those licenses, they should be able to build some cheaper (lower performing) storage nodes that they could place at their DR site.  Those software VSAs would then be put in their own management group and you would have to do asynchronous replication between the two management groups.

[rianui]

Thanks Oikjn,

The client has a database and applications that need to run on RAID 10 as requested by the DB vendors. He requested for storage with this in mind. His main points for getting the storage were to get both boxes to have the same data and in two sites in case one site becomes inaccessible. Clustering them on the same site isn't what he is looking at at the moment. The network bandwidth will be a big factor as you said but he is willing to overlook that as long as data on the primary site ends on the DR site eventually.

I guess the main concern now is the license. Will adding 600GB SAS drives need the license to be upgraded?

Regards,

Rianui

[oikjn]

the VSA SAN system is very different than most traditional SAN systems in that it is a network node based system.  This means there are TWO raid systems you have to take into consideration.  the first is the raid level of the disks on each node and the second is the NETWORK raid level (IE node redundancy).

The default disk raid is raid5 on the nodes and network raid10.  98% of the time, this doesn't need to be changed and I would not suggest changing it now.  It sounds like your client doesn't require realtime replication to a DR site and isn't looking for a "warm standbye" site as much as a way to get the data offsite in case the building burns down the data isn't totally gone. 

I don't think there is any cost effective way to upgrade your nodes to larger disks and I don't think this is the way to go for this client.  IF they can live with just offsite replication using a scheduled snapshot, then I bet they already have the HP license required to do this and they only need to buy a cheaper server, stick a bunch of disks into it and load up hyper-v or ESXi to run a VSA instance.  They should see if their starter bundle they purchased included some VSA licenses and if it did, they should go that way, if not, they should scream at their current VAR and then find another and see about pricing for a VSA license... IF they don't have the license already, I think you can activate a 60 day fully functional trial and they can see if that option works for them without having to buy the license.