NetApp - On The Edge

Fascinations

My apologies. I promised to write a series of blog entries reflecting on WAFL. And I have been writing them - several in my head. Editing. Discussing them with a few people at NetApp.

I believe this is known as writer's block.

I have been reading Designing Type by Karen Cheng. I have to make a confession - I am totally fascinated by typography. As a kid, I bought a small Excelsior printing press with three drawers of fonts from a friend for $25 - it unfortunately is long gone. Early in high school I had the opportunity to go to the Newark Star Ledger (a local newspaper) and watch what I recall to be their Linotype machine in action. Later in life I took an interest in small edition books from presses like Trillium (now Electric Works) - books such as Sandow Birk's and Marcus Sander's modern reimagining of Dante's Divine Comedy.

I'm slowly reading Cheng's illuminating text on the "general issues of type design." Because it is quite an enjoyable read. Spare in exposition, dense in information, and a model of clarity. An inspiration shall I say for me to overcome my writer's block?

I got stuck on the letter 'O'. I never considered it carefully. To quote from Cheng's chapter:

Designing a capital O today involves both objective and subjective rules. First and foremost, the capital 'O' in a serif typeface should always be circular or oval. Rectangular, square, diamond, triangular or 'free' O forms are not legible, since readers of the Roman alphabet have long been conditioned to recognize an O by its symmetry or roundness.

It proceeds from there for several pages with examples from the major families of serif fonts that illustrate the variety within the design constraints of the letter. It is then onto the letter 'E':

The E is the logical letter to design after the O. The E sets several critical factors for the entire font: the proportional system, visual centre, vertical stem width and serif and bracket style.

As I made my way to 'D', I was struck by the near inevitability of the design of a typeface once you choose the starting point. And the conceptual integrity that arises as you innovate within the chosen constraints of the design rules.

It struck me that my attempt at distraction from my writer's block led me back to consideration of the basis for the original design of WAFL. Dave Hitz reflected back ten years later that Mike Malcolm laid out three critical components that WAFL had to enable:

• RAID
  
• Write Optimized Layout (reallocate on write)
  
• Delayed Write Allocation (battery backed RAM)
  

Within these constraints, Dave proceeded to design a new file system. Years later, looking back, I was struck by the inevitability of what Dave created. And the integrity of the design.

Is this how you slip past a writer's block? Let me curl up with the serif capital 'B'.

Mobile Me - Projekt Revolution

So, finally a software upgrade to Apple iPhone app for Typepad allows me to blog while out and about.

Now considering it, untethered context is sometimes unrelated to work.

Hanging out with a finance person from Cisco, a struggling artist, and research assistant studying diabetes at UCSF. Waiting on Jim Voll, one of the current WAFL architects.

Linkin Park headlines - listening to bands in the California sun.

Life is good.

Flexibility is Key

A technical paper describing the Flexible Volume architecture was presented at USENIX '08 in Boston in June. The paper describes FlexVols in ONTAP that support dynamic creation of right-sized data containers that can be resized (grown or shrunk), snapshotted, cloned (producing multiple writable shared unmodified data instances for efficient multi-use "copies" of data), and mirrored for disaster recovery. These features are part of our storage virtualization technology.

FlexVols evolved naturally in WAFL, from its original single volume, single RAID group design by Dave Hitz going back to the beginning of NetApp. Over the next few postings (I really do need to pick up the pace here on blogging) I want to explore some of the features and evolution of this core technology to our storage products.

From a technologist's viewpoint, what has struck me fundamentally about the design and implementation of WAFL has been its remarkable ability to evolve in an upward compatible way to take on new features and enable new capabilities since its inception. Prior to joining NetApp in January, 1994, I was at Sun Microsystems where I was a bit involved with discussions on file systems directions at Sun. I remember being struck by the impression that WAFL at that time had many of the things we were looking for in a new file system for Solaris.

Technology is interesting. You see the end result and sometimes marvel. But technology doesn't create itself. People create technology. Looking at the arc of WAFL technology over my time at NetApp I see three distinct phases from Dave's original design, to a multivolume, multiraid implementation led by Blake Lewis, to the FlexVol architecture led by John Edwards. After Dave's initial work, the development of WAFL grew to include many contributors.

The biggest reason I stay at NetApp is the pleasure of working with truly talented developers and shipping innovative product to our customers. It's a lot of fun!

We are going through some exciting times in the storage industry. The storage virtualization capabilities of ONTAP are proving to be a key enabler for re-architecting data centers around application server virtualization. The greatest gains in power efficiency and reducing data center footprint is by taking an end-to-end approach from the application to the storage holding the application and its data. As a technologist I find it interesting that WAFL and ONTAP continue to enable innovative and flexible data center storage architectures and enhance capabilities application provisioning and deployment.

Over the next few weeks I want to describe not only the what, but the why, of the architecture and design of WAFL and ONTAP and how it affected our product evolution.

How to Participate in the Linux Community

The Linux Foundation (of which NetApp is a member) has recently launched a resource for the Linux community - The Linux Developer Network. The site is currently in beta mode (like the perpetual Beta state of Google Mail? :-).

One very interesting resource is a small book entitled How to Participate in the Linux Community. A welcome practical guide for developers and engineering managers alike in how to participate in Linux kernel development.