Centiq Blog

A short description about your blog

Mar 04
2010

Does the LeftHand know what it's doing?

Posted by: Steven Calvert in Infrastructure

Tagged in: Storage , HP , Hardware

Steven Calvert

( 3 Votes )

Storage at it's most fundamental level is a bucket, where you store useful stuff that you may or may not want again later. Of course what you put that stuff into can vary in design; it might not be a bucket but a box instead, or a sieve, or a safe, or the dog's basket because he never sleeps in the damned thing anyway. Likewise with disk based storage there are various methods to implement a storage subsystem design, all slightly different, all have their own way of doing things.

I recently spent a couple of days going back to school to learn about HP's up and coming storage product, acquired by HP after purchasing LeftHand Networks some eighteen months ago. This product has since been rebadged from "LeftHand" to the more snazzy, forward thinking name of "P4000". If I had a range of storage called "Eva" I'd call any additional product ranges something along the lines of "Bob" or "Sue". It's probably this urge that's the reason why I'm more technical and not marketing focused, however I thought that I'd share a few initial thoughts about the product with you.

HP LeftHand P4000 storage has been getting a bit of a buzz around it thanks to its more unique approach to scaling storage. Hardware wise it's very basic in choice, there's only a couple of units available,  the P4300 G2 and P4500 G2. These two units have either eight or twelve disks in a 2U drawer format, and each have a couple of gigabit iSCSI ports in the back. (With an option to upgrade to 10Gbit with the addition of a couple of PCI adapters.) You can also pick the disk capacity, however all bays must be populated from purchase with the same size disk. There are no expansion drawers like EVA or IBM's DS mid-range, if you want more storage you by yet another header unit and spread your volumes across multiple nodes. While this seems like crazy talk it does provide some interesting benefits. The first is linear growth, the more storage that you add the more interfaces, cache, and processor units you also gain. The second is in resilience, a host volume spread across multiple P4000 units can be mirrored using a concept called "Network RAID". That means if a single P4000 dies, the data is still accessible via one or more of the other P4000 units should you wish, be it on the same site or at a remote location.

 

 

2x P4300 Units.

That's not all, there's also the concept of a Virtual SAN Appliance (VSA). This allows you to use a VMware ESX virtual machine instead of a P4000 physical unit to manage storage, but keep all the functionality. VSA is therefore aimed at remote offices running VMware environments, to allow them to replicate volumes back to a primary site for backup. (VSA also has a 10TB limit, before you think of doing away with P4000 hardware altogether.)

Finally LeftHand has all the functionality you'd expect from storage these days, such as snapshots, mirroring, and thin provisioning, but with none of the additional licence costs! This makes the somewhat economical P4000 even more attractive for mid-range solutions.

So what are the advantages and disadvantages provided by the P4000? The biggest has to be the flexibility of the storage itself, it allows for the hot add and removal of storage units, along with the automatic redistribution of data across the available nodes, without impacting host connectivity. This is very useful for small dynamic environments which aren't planned, such as test environments or governmental outfits, where there's a lot of applications either in production or test that are being added in on an ad-hoc basis, with no predictable growth or performance.

It's not all sunshine, daisies, and crack cocaine with the P4000 though, there are some key points to be aware of. Performance will be limited with this device so if you're looking at high performance databases you may find more traditional approaches better suited. iSCSI is more limited in bandwidth compared to Fibre Channel, and with other systems sharing those same two Ethernet ports (per node), things will prove congested with heavy loads.

While Ethernet switches and iSCSI initiators are cheaper than fibre channel switches and HBAs, you're still looking at a dedicated storage network; This is SAN product and not NAS. There's a lot of chatter between P4000 units and not just with hosts, and so it won't play nice with your corporate LAN. Therefore you need to use this with dedicated storage networks only.

The platform is also Windows focused. While other platforms (including our personal favourite AIX) are supported for connection via iSCSI, it's only the Windows platform which supports the GUI management and the "Device Specific Module" which automagically directs I/O to the correct P4000 node. Other platforms send out I/O packets to nodes on a more random basis, meaning that P4000 nodes may need to re-route I/O operations between themselves to write data to the correct section of disk.(Hence the chatter between units.)

Finally if you're using nodes of different disk sizes, be aware of "stranded capacity". Like using odd sized disks within a RAID array, any nodes within a "cluster" default down to the lowest capacity node in the cluster. You'll probably notice this most when adding newer, larger capacity nodes to an existing cluster. As a result you won't get all of the capacity of your new node immediately, however if it's part of a hardware migration, the remaining space will become available once the smaller node(s) are removed. Again, all this is handled dynamically and automatically by the P4000s.

So in summary; cheap, feature rich, but can be wasteful in capacity if you're not careful. Also the performance may not be ideal for high I/O and throughput applications. It does things a little differently, enough making it worthwhile taking a look at. If you're currently using P4000 units I'd be interested in hearing from you, it'd be nice to know how these things cope in the field and if they're proving to be robust.

Hits: 892
Trackback(0)
Comments (0)Add Comment

Write comment

security code
Enter the displayed characters


busy

Bloggers

Alastair WilliamsAlastair Williams:
Data Management

Rebecca PritchardRebecca Pritchard:
Project Management

Robin WebsterRobin Webster:
UNIX

Steven CalvertSteven Calvert:
Storage

Steve StringerSteve Stringer:
Blade and SAP BWA

Glyn HeathGlyn Heath:
IT Industry

Emily MalbonEmily Malbon:
Helpdesk and Support

Tags

tecniq site follow the Centiq twitter Centiq on LinkedIn Centiq fanpage on Facebook