How To Guy

Q&A with Ed Lee on RAID, Virtualization-Aware Storage

Lee looks back at his work with the Berkeley RAID team, and how that impacts his work as the lead architect at virtualization-aware storage maker Tintri.

• By David Davis
• 08/15/2012

I often wonder what it would be like to one day be dubbed the creator of "X", where X is some amazing technology used by IT pros around the world. I was honored to be sitting in on a recent product briefing with Ed Lee, one of the creators of RAID storage. Today, Ed is lead architect at virtualization-aware storage maker Tintri, and he was gracious enough to grant me an interview. Read on to learn more about Ed and how his work with the RAID team at Berkeley has made an impact years later on his work with Tintri.

David Davis: What is it like, today, to be identified as one of the creators of RAID storage? How was it that you were involved in that?

Ed Lee: At the time, storage was not a "sexy" research area -- perhaps it still isn't! :-) But I saw the great team that professors Randy Katz and David Patterson were putting together for the Berkeley RAID project and I said to myself that I had to be part of that team.

In fact, I just attended our 20-year RAID project reunion at Berkeley, and looking back, it is hard to believe that it's already been 20 years and how much RAID is the basis of the storage industry as we know it today. I feel extremely fortunate and privileged to have been a part of it.

How does that intimate knowledge of RAID help you in your role at Tintri?

Tintri certainly uses RAID technology, but what has helped me more in addition to the knowledge of RAID is the experience I gained working with the other members of the RAID team, and seeing first hand how the development of a technology like RAID can have a profound effect on how people use storage.

With the "big name" storage vendors out there, why should the common admin consider Tintri when the lease on the old SAN comes up?

Well, first of all, they shouldn't wait for the lease to expire. :-) You can get Tintri now to simplify your life and speed up your VMs and reuse the legacy SAN for the non-virtualized applications that it was designed for.

The main advantage of Tintri is its simplicity and VM-aware storage management features. Traditional storage was designed to manage storage for physical infrastructures, not virtual. A virtual infrastructure is a software-defined infrastructure that is distinct from the underlying physical infrastructure. To a legacy storage system designed for managing storage in the physical realm, IO streams generated by virtual machines are blended together by the hypervisor into an unintelligible mishmash of requests. It's like trying to make sense of what's happening in a computer system when all the communication is encrypted. You lose visibility and control of the virtual environment.

Tintri, however, understands virtual machines. When it sees an IO request, it knows which VM and which vdisk generated the IO request and furthermore, you can tune and manage the storage on a per-VM basis. It sometimes reminds me of the movie, "The Matrix." To understand and have control over the virtual realm, you must become "aware" of the virtual realm and develop tools to control it.

What is VM-aware storage?

VM-aware storage is storage designed to manage storage in virtual environments. It is designed and optimized for abstractions such as virtual machines and virtual disks rather and LUNs and volumes. It provides visibility and control of storage in virtual environments.

How does monitoring storage stats with regular storage differ from monitoring storage stats with Tintri?

When a regular storage system gets an IO request, it has no idea which VM generated the IO request. As a result, it cannot report per-VM stats about what happens inside the storage system. With Tintri, we know for each IO request which VM and vdisk generated the request. We can directly measure all relevant stats on a per-VM and -vdisk basis. Now you can use third-party tools with legacy storage systems to try and deduce storage stats on a per-VM basis, but that is like using velocity and trajectory measurements and "dead reckoning" in order to estimate your current position. Using Tintri is like using a GPS. You just look and you know where you are.

Tell me about Tintri and snapshots and cloning...

Because Tintri is VM-aware you can snapshot, clone and -- in the next version of the product -- replicate on a per-VM basis. You don't need to spend a lot of time grouping VMs with similar snapshot, performance, replication, etc., requirements into the same LUN or volume because your legacy storage system only allows snapshoting and replication of LUNs and volumes. Why replicate or restore an entire LUN or volume when you only need one VM?

Additionally, Tintri snapshots and clones are created almost instantaneously; they consume no additional space when created and perform as well as the original. They can also be further snapshotted and cloned. In other words, there are no disadvantages or limitations to using Tintri snapshots and clones.

What are the common use cases for Tintri's cloning technology?

Tintri cloning is useful whenever you want to deploy a copy of an existing VM or create a modified version of an existing VM. For example, you may want to create dev and test copies of a database, or deploy additional instances of a server, or create many copies of a desktop.

With our VAAI plugin, any clones created from the vSphere GUI will automatically use the Tintri cloning feature. This means that you can clone a TB database in under three seconds! Also by enabling the View Composer Array Integration in View 5.1, any linked clones created by View Composer will automatically use Tintri clones. This gives you the manageability of View combined with the performance of Tintri clones.

What's the future of the Tintri product line (replication, etc.)?

Replication is the next major software feature for Tintri. More generally, we want to make managing storage in virtual environments as easy as managing a CPU is today. Ultimately, we want users to manage VMs, not storage. Storage management should be an automatic, integrated component of managing VMs.

When you aren't talking about storage, what's your favorite hobby?

There are a variety of activities that I enjoy and find relaxing like going for long walks or listening to talented amateur singers, but my favorite hobby is thinking about how technology affects people's lives. When I was younger, I was very interested in utopian literature -- how a few simple ideas can dramatically alter society, for better or worse. I find it incredibly interesting that today, technology can serve as a catalyst for such changes.