Wow, this has been a fun topic!   I’m really enjoying debating it, and enjoy even more that not everyone agrees with me.

I do not think that de-dupe is the only deciding factor in a purchase, but at this point, I do believe it is a show-stopper feature if you plan to use your IDT (intelligent disk target) as your PRIMARY storage device for backups (as opposed to only using it for staging).  De-dupe makes the device cost 10 times less.  That’s kind of huge.  De-dupe is the only feature making disk affordable enough to be used as a replacement for tape (at least onsite).

You mention replication.  I like replication, and I like the idea of replicating my backups off-site, and we’ve got some customers who have done it, and some more that are working on it right now.  BUT I’d say that unless you’re talking LAN replication, or you’re talking a significantly small amount of data, accomplishing replication without de-dupe is impossible.  It’s just math.  A) The device that people already can’t afford (a device big enough to hold all backups, not just be a staging device) now has to be doubled in size (one onsite and one offsite)  B) The occasional fulls and full-file incremental backups are going to create a whole lot of data that needs to be replicated.  So without de-dupe, replication becomes prohibitively expensive.

You mention performance.  I’m all about performance.  I do not, however, agree with your assertion that all de-dupe vendors have performance issues at a certain level.  I agree that many of them do have these issues.  Once you get to the performance ceiling of a given device, you have to buy another one and they don’t share de-dupe information.  However, that’s not the way all of them are.  If you find yourself in need of MB/s in the thousand(s) range, there is more than one vendor that can give you that within a single de-dupe setup (meaning it will all get de-duped together).

I completely agree that this is all new.  So you have to deal with that.  That doesn’t change the fact that de-dupe makes the replication idea much more feasible for many, many customers.  

If you don’t mind paying 10 times more (at least) for the hardware, and needing 20 times more bandwidth to replicate your backups, then feel free to stick with the more established products, and I truly mean that. It’s just that most customers I’ve talked to just can’t ignore those numbers.  They’re seeing de-dupe as making the not-affordable affordable and the impossible possible.

Curtis - Although I agree with the other responses you have given out with respect to the tape vs. disk cost I am not sure about your statements below.

 

Going back for a second to the cost of tape vs. disk... if you do an analysis make sure to take all things into account when you backup to tape. This is why most people don't get a proper cost associated with tape backup i.e:

1. SAN ports

2. Tape drives -> fixing them, lost time, shoe-shining

3. media cost -> fixing media, media failure cost(cost of not being able to do a restore)

4. off siting -> the cycles/dollars lost in handling that internally, the cost of dealing with Recall/Iron Mountain (or whoever), the cost associated with the delay in waiting for a tape to be recalled...

5. library maintenance cost

6. restore duration cost (i.e. if i have 100 people waiting for a Tier 1 server to be restored...)

Anyways the list of "invisible costs" associated with tapes go on...

 

As for your EMC CDL comments... First I believe they are now called EDL (EMC Disk Libraries) because they take into account their new Symmetrix backend devices. Although I agree with you that de-dup is important to the future of backups you make it seem that it should be the only deciding factor in a purchase! If you push de-dup aside for a second what do most customers want? My guess is performance, availability, stability, integration with backup application. This has been my thought process and these de-dup companies you speak about such as Sepaton, Diligent, Data Domain all at one point or another have HUGE performance hits (i.e. we have tape drives that go faster then some of these), little capability to scale (without combining multiple devices together), or have un-explainable single points of failures.

I also agree that replication is important and if you can minimize the amount you replicate then great. Here is my dilemma: Most of the de-dup vendors out there  (i.e. I am thinking of Sepaton) that can perform de-dup have only been in the replication business for a year (probably less) and have very little maturity in that space! That scares me a bit... 

 

As for backup integration I personally like the fact that with EMC I can have a built in media server on top of my VTL and control everything from what I am familiar with... no other vendor offers that!

 

Anyways just my two cents... Bottom line is that I agree that de-dup is important but if you can push that aside and look at the other technical merit (assuming that all vendors will have de-dup sooner than later) suddenly the list of enterprise level candidates drops significantly from what I am seeing.

 

-Nicholas

 

 

---

From: veritas-bu-bounces AT mailman.eng.auburn DOT edu [mailto:veritas-bu-bounces AT mailman.eng.auburn DOT edu] On Behalf Of Curtis Preston
Sent: Friday, September 21, 2007 1:13 PM
To: Kevin Whittaker; Jeff Lightner; veritas-bu AT mailman.eng.auburn DOT edu
Subject: Re: [Veritas-bu] Tapeless backup environments?

 

The only issue there is that the EMC CDL does not support de-duplication, and it doesn’t look like they’ll be doing it any time soon.  I know they’re working on it, but they haven’t announced anything public, so who knows.  Compare that to the other de-dupe vendors that announced probably a year before they were ready, and you’ve got some sense of my opinion of when EMC de-dupe will actually be GA – if not later.

 

Your design would work great if you had de-dupe. Without de-dupe, you are going to be replicated 20 times more data (or more), requiring a significantly larger pipe.

 

---

W. Curtis Preston

Backup Blog @ www.backupcentral.com

VP Data Protection, GlassHouse Technologies

---

From: veritas-bu-bounces AT mailman.eng.auburn DOT edu [mailto:veritas-bu-bounces AT mailman.eng.auburn DOT edu] On Behalf Of Kevin Whittaker
Sent: Friday, September 21, 2007 7:48 AM
To: Jeff Lightner; veritas-bu AT mailman.eng.auburn DOT edu
Subject: Re: [Veritas-bu] Tapeless backup environments?

 

We have it on our plan.  We will be using tape for only long term retention of data.

 

Our plan is to purchase another EMC CDL, and mirror our existing EMC CDL to the EMC CDL at our DR site.  Our master server already is duplicated, and this will allow us to start restores of stuff that is not tier 1 applications that already are mirrored to the DR site.

 

I would prefer not to save the long term on tape, but we don't have a solution for any other way to do it at this time.

 

Kevin

 

---

From: veritas-bu-bounces AT mailman.eng.auburn DOT edu [mailto:veritas-bu-bounces AT mailman.eng.auburn DOT edu] On Behalf Of Jeff Lightner
Sent: Friday, September 21, 2007 9:44 AM
To: veritas-bu AT mailman.eng.auburn DOT edu
Subject: [Veritas-bu] Tapeless backup environments?

Yesterday our director said that he doesn’t intend to ever upgrade existing STK L700 because eventually we’ll go tapeless as that is what the industry is doing.   The idea being we’d have our disk backup devices here (e.g. Data Domain) and transfer to offsite storage to another disk device so as to eliminate the need for ever transporting tapes.

It made me wonder if anyone was actually doing the above already or was planning to do so?

_______________________________________________
Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu