Join Newsletter
Trusted Business Advisors, Expert Technology Analysts

Taneja Blog

Taneja Blog / Data Center Systems

Storing the virtual infrastructure - Block or File?

Over the past year, we've noted a new generation of battle arising between block and file technologies, like never before. It is happening in the midst of the virtual infrastructure. Notably, it is drifting into the SME space, from the enterprise space where we first observed it years ago.

The crux of the battle is simplicity, in the face of a new compute technology that is bringing the server admin into closer day in and day out involvement with storage than ever before. Instead of standing up a server on top of local storage, now every virtual server setup practically screams for networked storage to make full use of all of the promises of server virtualization. And once external storage is leveraged, the benefits can be tremendous, but only if you can actually use the full capabilities of external storage. But the virtual server is still poorly equipped to orchestrate storage itself, and trying to make the virtual server and storage relationship work, and reap all of the promised benefits, across a multi-disciplinary team can be a nightmare. Sometimes, standing up a single repository where everything can be stored by everybody, without further interaction, while the storage team still marches on, just seems simpler. It isn't that simple. And that isn't a full picture of the costs either.

The virtual infrastructure is in good part about extending your business capabilities. How do you get there? By thinking about how you can better enable the applications that your business depends upon. Your choice of storage foundation will make a difference. A few things to consider:

Point 1: Can you manage all this contention? As you mash virtual systems together on top of ever bigger pieces of iron, and push more storage onto single arrays, the interactions from your virtual machines can create a huge IO blender. How do you separate the IO and make sure the right workloads get the right resources? How do you see what's going on to start with? If your block storage is efficient enough in the first place, you can segregate workloads and isolate storage resources without an increase in management effort, nor complexity. Not so much on a single mount point. Think about whether this is important to you, today, or tomorrow.

Point 2: Tiering and automated optimization. Once a platform is in place for recognizing the difference in storage demands between multiple apps, it can serve as a foundation for keeping things optimized. Either through volume redistribution, ala 3PAR's wide-striping, or automated tiering and scale-out with EqualLogic's PS series.

Point 3: Virtualization Integration. We're missing some of the huge leaps in storage integration capability that VMware recently unlocked with their vStorage APIs for Array Integration (VAAI). This makes an enormous difference in array efficiency. Can your storage system give you the IO efficiency you demand by taking advantage of various vStorage APIs, such as the VAAI SIOC, copy same, and similar features that can decrease over-the-wire traffic, reclaim wasted space, and better control who is consuming what from your storage system?

Sustainable scalability. Similarly, while you'd think that the promise of massive single points of attachment would be the nirvana for scalability, it isn't quite nirvana. What happens when you hit a limit? Is it really just a matter of storage motioning a VM? What if you need to retain snapshots across a storage motion? There some value to be had there from the scalability to be found in next generation block.

And I am quite confident in predicting the race will be on this year to see who out innovates who in the race for features and virtual infrastructure alignment.

This all goes to say that block isn't going to be displaced, there's just more than one alternative in the market. Sometimes simple is what simple gets. But it's not really any more complexity, to get a whole bunch more long term alignment with your application workloads. It just takes a bit more thinking about what you really need. Take another look at block, and you might be surprised, especially if you're not looking in the right place. Where is the right place? EqualLogic. 3PAR. Scale Computing. IBM Storewize V7000. If you're a Hyper-V customer, definitely add DataCore to the list. That's just to name a few. Meanwhile, you shouldn't rush to predict block's demise.

  • Premiered: 01/06/11
  • Author: Taneja Group
Topic(s): 3PAR EqualLogic Block iSCSI NFS NAS scale-out Array Virtualization VMWare VAAI Server Virtualization


There are no comments to display. Scroll down to leave your own!


Leave a Comment

You must be logged in to comment. Click here to log in or register if you don't have an account.