Printer Friendly

Avamar defuses data explosion: company tackles data growth disease, not symptoms.

Avamar Technologies made a splash a few years back when it unveiled its commonality factoring technology that enabled unprecedented reductions in backup data loads. The company's Axion backup and disaster recovery products are now widely available through the reseller channel and partners such as Hewlett-Packard, IBM Global Services and StorageTek, which offers the service through Avamar partner Arsenal Digital. In June Avamar named former CNT executive Edward Walsh as its new chief executive officer. In this discussion with CTR, Walsh explains why Avamar is the only vendor that is tackling the disease rather than the symptoms of explosive data growth, the company's approach to customers and partners and why ATA drives are not a significant part of Avamar's disk-to-disk backup technology.

Q. What was the attraction that brought you to Avamar?

A. I was looking for a new job. I had grown CNT and actively merged them with McData, so I was actively looking for a new opportunity. I looked at about 20 different companies, but in one of the my trips I got cornered by this guy named Bill Gurley (an Avamar board member and a general partner at Benchmark Capital, one of Avamar's investors) and he said I've got an idea for you but I can't tell you what it is, but let me white board if for you. And about a minute into it I said "backup software?" The world needs more backup software like it needs a hole in the head. And that's how the thing started. But he made me talk to Rory Bolt, who is our CTO. Once you see what Avamar has, it's truly compelling.

It's not like the other technologies that are incremental. It's all ILM this or archive or CAS products, which are nice to have but you don't need them. The more I got into Avamar, the more I met with the top 15 clients, our partners and our employees, I saw that the market is now and people need what we have. And I'll tell you why they need it.

Data is exploding. Everyone's been talking about the way data's exploding but it's been going on for a while. The way we've been managing data, protecting it through backup and recovery and replication hasn't changed in quite a long time. We're the only solution that I found that attacks the disease and not the symptom. It goes right at the data at the source. Instead of doing full backups and now incrementals, we look at it differently and use technology with all the MIPS we have available now, toss a little compute power at it, and put in some algorithms and now look at the data and are able to truly break it up into lower components and only send across the wire what's truly changed, and a map of how to rebuild it. So instead of doing backups at 200 percent of the full data set, we can do it at 5 percent. As the data explodes, we're the only ones that are tackling the real problem, or the disease, not the symptoms, which is where all the ILMers come in. The tape library is slow so let's put a disk pool in front of it, or VTL or backup to disk. Your network isn't big enough so let's increase it. Your disaster recovery's not good enough so let's add new products like disk mirroring, which are really expensive.

That's why I think this opportunity is so exciting. The market's here, our technology is right in the sweet spot, we're the only one tackling the root problem, which is you have to move less data. That's we why we use the term "move the data not the media".

We're starting to get clients. Also, we're starting to see the partnerships take off. We're seeing our regional VARs pick it up. We have regional VARs actually dedicating resources to Avamar. All they sell is Avamar. Why would you want to do that? We're the only ones to tackle the real problem. So it's a way for VARs to open doors and expand their business. Which is exactly what VARs are looking to do to get the angle, the wedge in the account and have that "aha" conversation. The scale is that you're moving such a small amount of data you now only have to scale the environment for a fraction of what you needed previously.

It's a natural progression from what people are going to do and the data explodes, you have to move less data. And you have to do it at the source, which is at the server level. You can't do it in the back office with some VTL. You have to do it where it hits so you get better wall clock time. Backups aren't taking the full weekend; they're going down to 15 minutes. The network bandwidth is being reduced so you can replicate all over the world wherever you want to do it.

Our message of integrity, speed and scale is resonating. Because we've looked at the system and the way we are able to do our back end--integrity, the integrity checks backup to disk you greatly reduce the need for tape other than for archiving. It's really the speed of disk backup. The scale is that you're moving such a small amount of data you now only have to scale the environments for a fraction of what you needed previously.

In the end I went after Avamar pretty hard. Out of 20 different technology firms, they're the only one that in my opinion was a game changer. So I went from I'm not interested to I'm in a week's time.

Q. You mentioned your strategic partnerships with companies like HP and IBM. It seems to be a bit of a paradox working with these server companies who are also major storage suppliers when your technology allows people to get the job done with significantly fewer hardware resources.

A. It is a little bit of a paradox but for different reasons. They're hearing from their customers how do I manage this? From IBM Global Services, they're hearing how do I do this? How do I get better recovery point or time objectives? How do I do this when my budget is less? They're looking at a compelling and competitive tool. So IBM GS is looking at this as a way to grow their business, because the clients can't continue to do what they're doing. So he leapfrogs the clients, because if they're just doing the tape rotations and take the tape offsite to Iron Mountain, they're nowhere near the recovery point or recovery time objectives and their costs are through the roof.

Whereas, he can do it as a service, I can get you replicated offsite with recovery point and recovery time objectives. They've got a great new opportunity with no competition. STK--through Arsenal Digital--sees it as an equally good opportunity. They know tape's going to get a little bit eaten so who better to help them? They're trying to do it as a service first, which is a good way to go about it. So you're going to continue to see the momentum.

Q. Will we continue to see Avamar go direct to customers and work through service providers such as Arsenal?

A. Our model is to engage directly with the clients. We're not going to change our model, but do it intelligently. You've got some clients who will just take and make no value change, they do their own scripting, their own support. Then you've got others who want to buy the service for a price per protected terabyte. It allows clients to consume it in a way they want to. So to me it's the way to have a whole product. You'll see us building out professional services, you'll see us increasing our support, you'll see us doing more customized packaging and scripting, but what you won't see is us competing with our partners, because that ecosystem is going to make us very successful. More important, it gives the clients what they want. So what we can do now is go to the client and say, here's this great technology look what it does for you and you can have it in-house or as a service offering.

Q. And Avamar will be supporting your channel partners with joint sales calls and other activities?

A. Absolutely. One of our great value adds is having that light bulb moment with the partner there. It's even better than training in the classroom. Most of them get it right away. If you have a VAR that knows backup, they light up and say "wow, I see what I can do with this." In the last six weeks I've had four meetings, and three out the four were the "ahas." And the fourth was probably not the right VAR for us.

We had our customer advisory council last week, Cisco, MCI, Adaptec, Morgan Stanley, Qualcomm. We talked them through and asked them, what did you buy it for? What they all brought up was the wall clock time. They all expected the reduction in data with the commonality factoring. They all said, we didn't quite believe you, but you did deliver. They were all pleasantly surprised there. They said the wall clock time. Because you take these incremental backups from 2 to 4 hours down to 15 minutes. Their weekend full is only another 15 minutes, compared to a 6 to 12 hour backup for tape systems. And that buyback was the light bulb moment for them. There was one customer with 60 hours for a full backup once a week and 6 hours of incrementals every day, and that came down to 20 minutes every day for full backups. Literally 50 percent of their IT infrastructure was shut down for data protection.

Q. It seems that the data reduction numbers that you're talking about are so ridiculous compared to what customers are familiar with that they're saying that it can't be real?

A. Exactly. The numbers are so off the charts from what they're expecting. Think about it. For example, to protect 50 terabytes, I do a full backup, then I do 5 terabytes a night for incrementals for six days, so that's 80 TB of data I have to move to protect 50 TB. With Avamar, we move less than a terabyte of data to get the equivalent of a full. So I move 3.5 terabytes a week. Do you want to move 80 terabytes a week which gives you one full backup and six incrementals or 3.5 terabytes which gives you seven full backups? So that's where the game changing comes in. We're the ones that are attacking the real disease. We're not addressing the symptoms, such as "well that tape drive's slow so let's put a disk buffer." As one of our customers said "that's just a BandAid on gangrene." There are other vendors that are doing data de-duplication but they're doing it at the server. It's almost after the fact. So behind the backup server it's like moving an entire avalanche of snow down to the bottom of the hill instead of just carrying a snowball down. They're doing a disk pool behind a Veritas server and it's really masking the problem.

Q. So Avamar's approach is in the data path at the point of creation?

A. It's on a server like Veritas or Legato you have an agent in the client's production environment on the client's application servers. At that point we do the data de-duplication before it goes on the LAN to the backup server, which in our case is an Axion disk backend. The other thing that's counter-intuitive is you'd think with all the data crunching it would consume CPU cycles on the application server when you're doing back-up that's 4 to 12 hours pinging the CPU. Backup consumes memory and bandwidth so you have options. You can pay for 15 minutes or you can pay for 4 hours. So what we do is we lock the file systems and just like incrementals we look for what changes, and then we do some algorithms on that particular information to look for redundant data. We just don't send the redundant data. This is our commonality factoring that allows us to only send the sub-file content that is truly unique, and metadata about how to rebuild it.

It's really not rocket science. It's the implementation that's patented technology. What we're doing is bringing the compute power of Moore's law to take on this challenge. You could never have done this with the compute power that was available or on the back end how we're doing all the error correction, if you didn't have Moore's law working for you. Ten years ago you didn't have the computer power to do what we do, but now it's readily available and it's really cheap. So we're applying that power in an intelligent way.

The end result truly changes the game for the end user about how they keep up with data growth, how they increase the recovery and how they drop their price per protected environment: TCO.

When you replicate environments and now you're doing something that's much better than tape for about the same price as tape. But you get better wall clock time, better integrity, very efficient replication to remote locations.

Q. The first generation of D2D addressed the backup window issue to some extent, but didn't really create any significant cost savings compared to tape. Is the Avamar solution the missing link that redefines the cost structure for disk-based backup and truly relegates tape to an archival medium?

A. Relegated to archiving is a good way to put. And the other part to the economic discussion is we were talking before about throwing disk buffers and disk stages at the current solution. When you look at the TCO for that, all you've done is added costs to the equation. You haven't eliminated any costs. So by definition disk is more expensive because I've added it into my tape architecture. With this solution I don't want to say it's a complete replacement, but it displaces tape from the operational backup. One's a BandAid and one's penicillin.

Q. Is it safe to say, then, that the key enablers of this technology are the processing power of the current generation of Intel and AMD chips in combination with low-cost ATA drives?

A. ATA drives haven't been a driver here. Most of our customers are using SCSI and Fibre Channel. The first and most important thing is that disks have dropped in price dramatically over the last five years.

Q. That's surprising, because when you talk D2D in any context the assumption is generally that it's ATA-based.

A. Our customers have such buying power and comfort level with their existing drives that they just don't have a comfort level with the ATA technologies yet. We also move a lot less storage, so the fact of the matter is we don't need as much storage so they can spend on the drives that they're comfortable with. Linchpin decision #1 was that disk is going to become cheaper over time. Decision #2 was if I approach backup with disk as the primary target as opposed to tape, then I can get new economies of speed, scale and data integrity. Why is that? Well, tape is an offline linear device so if I write something I don't know if I've written it, so I write it again, and again and again. And by the way, it uses a flimsy media that tends to fail 35 percent of the time. Disk is a spinning media. I know what I've written so I don't need to write it again. Therefore, I get economies of scale not only across time but across operating environments, platforms, applications, etc.

Q. One other issue, in terms of what's driving the business: how significant has the compliance opportunity been for Avamar?

A. From the compliance standpoint, it might not be for a specific regulation, but we allow them to protect environments, replicate remotely and we're able encrypt the data. There are a lot of different benefits, but people are using it to just have compliance about protecting their environment. It's not necessarily tied to a particular compliance issue, but we see people applying it to HIPAA, we see people applying it to the SEC regulations, SOX. It's more of a horizontal issue: how do I fulfill my fiduciary responsibility, how do I get it back?

What's happened is that the data has grown so quick, how do you continue to back that up, project it and replicate it, and what's happened is they weren't. So all this discussion about SOX, HIPAA and compliance brought this front and center. It's a contributor, but it's really bigger and more horizontal than that.

Q. Where is this company going? What's the future for Avamar?

A. We really revolutionized the way you do backup, replication and restore. We're attacking the real disease at the source. From here, there are a lot of things we can do with the technology, but that alone is a huge market space for us. It gives us a lot of things we can do for our clients and build out. Now once you have an efficiently stored environment, it's replicated for protection and it's a catalog of your data over time, that's a really interesting foundation for an enterprise content manager. It's basically what you get when you have our environment.

In the short term, we're making sure we're growing our customer base. We're listening to our clients very intently understanding what they're looking for. From a strategy perspective that might seem pretty ho-hum, but it's really about what our clients are looking for. So the key is how we grow our customers, how we grow our partners, how we listen to them and do what they need. It's really customer-driven features rather than a major upgrade to the core technology is what they need.

Our job is just to do business fundamentals. The market is there. Our technology is a clear fit. We're alone in that race. What we have to do is build the business based on that.
COPYRIGHT 2005 West World Productions, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2005, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Avamar Technologies; Edward Walsh
Author:Murphy, Dan
Publication:Computer Technology Review
Article Type:Interview
Geographic Code:1USA
Date:Aug 1, 2005
Previous Article:Preparing for encryption: new threats, legal requirements boost need for encrypted data.
Next Article:Getting more intelligent about storage management.

Related Articles
Content-addressable storage. (Storage as I See it).
How women interpret gynecological symptoms: qualitative research shows symptoms affect behavior and sense of well-being.
Pervasive data volume reduction: a high-value alternative to aging tape-based backup technologies; A paradigm shift in backup technology can cut...
Study to seek workplace factors that cause depression in women.
TCO analysis: where D2D fits--part 2.
The year in storage: data protection led innovations.
Leadership perspectives 2005.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters