NetApp - Dave's Blog

Will Flash Replace Disk?

People often ask me when disk drives will be replaced by some newer technology, like bubble memory or holographic memory or flash memory. Last time I got this question, the person aimed me at this article: Flash for Laptops: Computers that use flash chips instead of a magnetic hard disk for memory are coming.

So will flash replace disk drives? Two answers:

I don't think so. For high-capacity storage systems, I don't think flash or any other new technology will replace disk drives any time soon.

I don't care. From NetApp's perspective, I don't think it really matters, because I don't think it would have much effect on our technology or our business.

Now I'll dig more into each answer.

I don't think so.

It makes great sense to replace disks with flash in small portable devices. Flash has already replaced disks in the iPod, and I've been craving a laptop made with flash for a while - lighter, more shock resistant, less power, no spin-up time. I want one!

But for high-capacity storage, flash is just too expensive. On the web, I can get a 250 GB SATA drive for $160, or $0.64 per GB, while 1 GB of flash costs $20 - thirty times as much. Even more for high-speed flash. Of course, flash is getting cheaper pretty fast, but disks have an amazing track record of getting bigger and cheaper. Technologies like perpendicular recording, Heat Assisted Magnetic Recording (HAMR) and Bit Patterned-Media (BPM) promise to continue this trend well into the future.

Of course, there are other potential competitors to disk, like holographic storage, but they keep not shipping. I'm not saying that nothing will ever replace disks, but I'd be very surprised if any new technology were shipping in sufficient volume to replace disk drives five years from now. And it wouldn't surprise me at all if disk drives were still the high capacity medium of choice even ten or fifteen years out.

I don't care.

Having said all that, I don't really care very much one way or the other, because I doubt that a new storage technology that replaced disk drives would have very much effect on NetApp or our competitors.

My guess is that most of our storage and data management capabilities will still be required no matter what this new technology is. For instance, I bet that it won't be 100% reliable, so we'll still need RAID-like technologies to protect against failure. We'll still need remote replication to protect against disasters. Pretty much any capability you can name - backup, snapshotting, encryption, compliance, thin provisioning, application integration - will still be necessary, so what difference will it make whether we buy disk drives or some futuristic new storage device?

If the new technology doesn't emulate disk interfaces, then maybe a few device drivers will change. If it isn't a spinning media, then some optimizations we do to deal with seek times will no longer apply. Maybe some new optimizations will be needed instead. But other than that, I just don't think it'll be that big a deal to replace disks as the underlying storage technology.

Storage Haiku

I really enjoy writing, and of course I've devoted a good bit of my life to thinking about storage, so imagine my excitement at discovering that Mike Karp is running a contest to identify the best storage haiku. (Hurry! Contest ends April 30th.)

To get your creative juices flowing, let me share a few of my own masterpieces. First I thought I'd toot my own horn:

Life is too complex
Unix/Windows do not help
Invent appliance

Then I decided to focus on some of NetApp's capabilities:

So many users
Not enough disks in array
Thin provisioning

Tape falls off of truck
Twenty million credit cards
Should have encrypted

ATA drives suck
And yet so inexpensive
Better have RAID-6

I have a hunch that touting your own products is not a winning strategy, so I thought I'd better move on to broader industry topics:

Summer storms are here
Data below sea level
Remote site saves job

ILM is hot
But what does it really mean?
I am still confused

Virtually painless
Says doc with three inch needle
Virtual means "not"

Twin towers fall
No D2D for e-mail
Terrorists have won

In the end, though, I think what matters most is the experience of the end user:

Digital photos
The sound of one head crashing
A childhood is lost

If you aren't wiping a tear from your eye, then you have no soul.

Calling all poets: Give me your best shot!

NetApp Makes Fortune 1000

Sometimes I just feel like bragging. NetApp just made the Fortune 1000 list for the first time, coming in at #929 with $1,598m in revenue for calendar 2005.

Since we have such a high growth rate, I've always found it more useful, when doing market analysis, to look at our annualized run rate, rather than our trailing full year. In our last quarter, we did $537m, which annualizes to $2,148m. That would have taken us up to #772.

What fun! Can you tell that I'm proud of my baby?

New Site Dedication in India

I am in India to dedicate our new site here. I wore a traditional Indian outfit—a tunic called a kurta decorated with an intricate pattern of gold and red thread, pyjama pants, and traditional embroidered shoes. I cut the ribbon and lit the flame and broke the coconut. Don't ask me to explain the flame or the coconut. I can't even explain why we cut ribbons at dedications, but that's the tradition.

When we first started a technology center in India, the reporters I talked with here were very interested to know how big we would get and how fast we would grow. I felt that they were asking the wrong question. To me, the size and speed were less important than building a solid foundation. At that point, I cared much more about how to establish strong roots, how to get a set of "old timers" in place who understood NetApp culture and had a track record of success.

Three years later, we have the foundation in place so we can talk about head count and growth rate. Today we have 270 people in India, but the new site that I dedicated has capacity for 750 people, and we have an expansion option that will take it to 1200. Over the next two years we expect to spend at least $150 million in our Indian operations. These are the details that the reporters were hoping to hear three years ago, and now I feel comfortable sharing them, because we have that solid foundation to build on.

NetApp is very lucky to be successful enough that we can grow our Indian operation in parallel with growth in the U.S. Many companies are cutting U.S jobs as they hire people in India or China, and I know that must be an incredible challenge for their cultures. For those companies, it must be so hard for existing employees to honestly welcome the new ones, when they know that old friends were fired to pay the salaries of the new employees overseas. So I think it is very healthy for NetApp that even as we expand into India, we are—right at the same time—continuing to hire in the U.S, both on the West Coast and the East Coast. I hope that helps our U.S employees offer a genuine welcome to the newcomers.

They say that what goes around comes around. Therefore, I find it fitting that NetApp is now tied with HP for top share in the Networked Storage market in India.

Why Encrypt Data on Disk?

I am at the Storage Networking World (SNW) conference, and in an IDC question and answer session, someone in the audience said:

I understand why you would encrypt data on tape, but why encrypt data on disk? It's going to be decrypted by the time it gets to the server anyway, so if you can get to the server, you'll be able to steal the data whether it's encrypted on disk or not.

The key is whether any people have access to the disks that do not also have access to the server. If so, then encrypting the disks may be useful, because it'll protect against those people.

We intuitively recognize that encrypting tape is useful because we know that lots of people might get access to the tape. If you send the tape to Iron Mountain, then everyone from the shipping clerk to the truck driver has access. As a result, encrypting tapes and nothing else is a common first step, just as the questioner suggests.

On the other hand, I would argue that in many environments, there are people who can access the storage who wouldn't necessarily be able to access the server. Especially with SAN, NAS or iSCSI, the server can be in a different location from the storage, and it probably has different access controls. The people who run backups or make remote copies of data, for instance, may not have login access to the application server. The janitorial staff may have physical access to the storage, and could walk out with a disk, but they probably can't login to the server. The questioner is correct that encrypting the data on disk won't protect against all attacks, but it can certainly create an extra layer of security that protects against some attacks.

One common use of disk encryption is in defense or intelligence organizations that want to consolidate storage for separate groups that have different security clearances. Mildly secret information is classified as "confidential", "secret", or "top-secret", but very secret information is "compartmentalized", which means that there are many different categories of secretness. Group A is not allowed to see group B's secret data, and visa versa. As a result, each group is forced to buy its own separate storage and keep that storage in its own secure area. Today, each group can have its own encryption appliance, with its own keys, but the storage can be consolidated into a shared, non-secure data center. This is especially useful for groups with large amounts of secret data, because often there's not enough space or cooling inside the secured areas.

Disk encryption can also be useful if you replicate data to a shared disaster recovery facility that is less secure than your primary site. Don't transfer the keys to the remote site until you actually need to use the data there.

My point isn't that everyone should encrypt all data on disk. Far from it. As I said above, the easiest and most common first step is simply to encrypt backup tapes. Over time, however, as privacy and security issues continue to grow, I expect that more and more disk storage will also be encrypted.