Storage Monkeys

Curtis / Greg / Marc:

The title of your podcast episode #8 concerned me as our goal at Dell is to be direct and put our customer’s first. While the bulk of the discussion regarding marketing promotion around deduplication was focused on CommVault, you did mention Dell briefly, and I thought I might try to help to clear up any misconceptions you may have gotten from our recent announcement.

•The PowerVault DL2000 Powered by CommVault (http://www.dell.com/dl2000)is a back-up to disk solution with optional integrated deduplication.
- It is a full back-up to disk solution including hardware, back-up software – which is CommVault Simpana 8.0, thus the call-out in the product name – and services.
- It also can include an integrated tape library for archive to tape & is one of the first disk solutions that offers deduplication to disk and tape.
- Block-level dedupe is one feature of the solution. Data is compressed at the host and sent to the media server, where deduplication occurs. We do not call this out as source-based dedupe – please let me know if you see that somewhere and I will work to get it corrected.
•There are pros and cons to both an approach that integrates deduplication into existing applications, like in the ISV as CommVault has done with Simpana, and one that utilizes a dedicated deduplication appliance approach, a la Data Domain, EMC DL line, Quantum Dxi, etc. (see below)
•One of the key call-outs in our deduplication announcement was that you first need to understand your environment. Deduplication is one approach to addressing data management challenges, but not the only one. Dell has added services that help to determine if a user would benefit from deduplication, and if so, recommend the best way to implement it. Dell can provide solutions utilizing several different approaches, unlike some vendors who only offer one way to solve a problem.

Our goal is to help customers solve their data management challenges, not contribute to the confusion, so please let me know if there are any additional things you think aren’t clear or anything else you’d like to discuss about our point of view on dedupe. Unfortunately, I don’t know anything about the acquisition speculation, and even if I did, couldn’t comment on it – but I did find that segment of your discussion entertaining .

Greg:

Good feedback and thank you for clearing this up - your explanation is exactly my understanding of the solution. I don't think this was as much a Dell issue as much as it is a CommVault marketing issue.

Here is one example of my concern:

From Chris Mellor's article in The Register(http://www.theregister.co.uk/2...ce_dedupe/):

****
"Instead of seeing these software products backup data to deduplicating appliances, such as those from EMC, Data Domain or Dell's DL2000 CommVault Simpana box...."
****

Either Mellor doesn't understand the solution or he was misled. The problem is that there are numerous references like this that confuse the DL2000 with Data Domain. Again, I don't think this was a Dell problem, I think it's a CommVault marketing problem.

Hey - you want to be a guest on Infosmack?????

Apples and oranges:

"The appliance has a data ingestion rate of 1.5TB/hour, which is faster than the 1TB/hour ingestion rate of Data Domain's backup appliance, a CommVault spokeswoman said."

-ComputerWorld - http://www.computerworld.com/a...Id=9134105

Please read this article and tell me how Dell is not referring to what they have as source dedupe:
http://www.infostor.com/index/...ht_on.html

I have also seen a Dell PPT depicting backing up fileservers and laptops NOT connected to a local media agent saying that "only changed blocks" will be sent from those clients. How will that happen if you are not doing source dedupe?

The PPT also shows how "smaller pipes needed" to back up clients using your solution. _I_ know that this is because you're compressing data on the clients prior to backup (which I'm still not convinced is a good idea), others are looking at this thinking that you're deduping at the client, which you're not.

Another phrase that hints at source dedupe is Dell's referring to what they're doing as "end-to-end dedupe." It's not quite end-to-end is it if it's not deduping at the client?

In addition, both Dell and CommVault are referring to what they are doing as "block-level dedupe." They are not deduping at the block level. They are deduping at the chunk/segment level. A segment/chunk is a series of blocks, typically 64K in size with the CommVault offering. 64K is hardly a block.

To add some fuel to this... From David West's blog (http://news.commvault.com/DavidWest/ 000016_The_Broken_Backup_Foundation_beneath_the_House_t
hat_Data_Domain_Built.asp):

"So, is it time for some of the dedupe appliance hype to be replaced with a more realistic picture of what you can and can't do with a standalone hardware-based approach? "

This is a little deceptive as well. The CommVault/Dell backup appliance is a software-based backup and dedupe, not hardware based at all. It's backup software on a server appliance. Geeze.

The same blog also contains these statements (emphasis mine):

"As a point solution, dedupe appliances have done a good job of helping customers slow their back-end data growth, but they can't alleviate broken backup and archiving challenges. Therefore, they can't address ever-increasing amounts of front-end data and this is where the fundamental problem lies.

To do that, you need to go to the source–where data is moved, managed, protected, replicated and archived."

While dedupe is an absolutely necessary technology in reducing data to lower storage costs, HOW you go about doing it can make a big difference in terms of how much space you actually save. A smarter approach embeds deduplication in the data movement software–the "source" you might call it–as backup and archive software vendors control the "source copy" and therefore have the benefit of being both "content" and "application aware."

"Software-based deduplication is better optimized for both data ingestion and recovery, yielding up to 50 percent faster restores than its appliance-based predecessors. That's because data reduction begins at the source."

So how is someone to read this and NOT think that they're doing source dedupe?

Its all about load, bottle necks and balanced use of infrastructure in my view..
Where can you afford or is it best to have the work loads done? on the client, on a proxy or Media Agent or at the target ie the disk..
What will each approach cost you in terms of load on that system, $$$ to procure and manage and what is the impact or benefit of each approach?
There is work that can be done on the client.. perhaps compression (some clients), hashing, comprehension of whats changed and whats not, be that at file level, block level or bit level.. These things are appropriate on the source in my view but best approach differently, based on the type of client and the environment its within.. Are we looking at laptops / desk tops, SMB servers, Enterprise class systems, Virtual Environments, 24x7 operational environments?
There are other tasks where loading up each client is not always ideal, such as compression (some clients), the analysis of data that is being moved in relation to the rest of the data in the environment and checking it against dedupe databases for a decision on whats retained and whats already stored.. In my view this process needs or should be global.. This is variable though depending on the size and geo disparate nature of the systems under management.
Should each client check each client for uniqueness of data, does a distributed checking methodology work better for your business or should a central point manage that task and store data accordingly?
If you comprehend the data associated with each system or source in each location you can manage that data, transfer it, protect it, store it intelligently and cost effectively across your entire environment and base that approach on the criteria defined by your business policies.

The discussion is not really about what value their solution has. I don't doubt it has value for certain environments.

What we're talking about here is deceptive marketing practices.

I mistakenly said it was a Chris Mellor story that referred to CommVault, but it was NOT. it was an InfoStor story from Kevin Komiega:

http://www.infostor.com/index/...ht_on.html

Mr. T is awesome.

Storage is pretty cool too but not as cool as Mt. T.

End of line

I think some of this confusion when it comes to Dell may be in those not realizing when we (and others) are talking about our point of view on dedupe and when talking about a specific product like the DL2000 that was part of our recent announcement.
Our point of view, among other things, includes that you need to understand your environment and needs to determine whether dedupe is good fit and, if so, what will be the best way to implement it, and that integrating dedupe into existing processes, like ISV software, can provide greater benefits than implementing a stand-alone appliance in the long run. One size does not fit all, and we can provide solutions that are source-based, ISV-based and target-based.
This is a confusing topic to many, including those in the industry due to nuances of implementation and terminology. The challenge for us all in talking about dedupe is to be simple enough to be understood above "the weeds" yet convey key technology concepts and value propositions that are helpful to all. I am enjoying the discussion, and hopefully this helps readers understand all the different aspects of our recent announcement, and that it continues to grow and evolve as Dell and others continue to build dedupe as a feature into more and more places in the IT realm.

After reviewing all the articles, I am taking back my comments about CommVault. Dave West's blog is misleading, but he's basically guilty of using a poorly chosen word when he said you need to "go back to the source." He did not mean to imply that they do source-dedupe, as he knows the difference between that and what they do.

Dell, however, gets no such pardon, as you have specifically said that your product is a source-dedupe product. Let me cite two examples:

While being interviewed about the Dell/CommVault/EMC announcement, Brett Roscoe, senior manager, product marketing, Dell Enterprise Storage, said:

"There are different approaches to deduplication for customers to choose from and, internally at Dell, there is a market for both source-based and target-based dedupe," says Roscoe.

This (and other comments he must have said) cause the report to then say this:

"On the source side, Dell today unveiled a new member of its PowerVault disk-based backup platform based on the latest version of CommVault's Simpana software, version 8.0."

(The interviewer understood Roscoe's comments to mean that CommVault does source dedupe.)

"Target-based deduplication has its place, but we are taking a stronger stance on source-side deduplication because it is happening faster than we originally thought it would. We didn't expect ISVs would get it integrated into their products this quickly," says Roscoe.

So while being interviewed about the Dell/CommVault/EMC deal, he emphasizes the need for both source and target dedupe, and that he was surprised the ISVs got source dedupe out so fast. What else is the interviewer and a reader (who is "above the weeds") going to think but that CommVault has source dedupe -- which they don't?

My other example is your post: "One size does not fit all, and we can provide solutions that are source-based, ISV-based and target-based."

In a comment to an podcast where we specifically said that Dell is being misleading by talking about source dedupe, you say that you can provide source dedupe . You don't clarify that are (you MUST be) talking about Puredisk or Avamar. I know that because I know that the CV product isn't source dedupe, but the casual reader is going to think that you mean to say that CV does source dedupe.

If you want to say that they can get target dedupe from Dell (by buying the EMC/Quantum product), what you're calling 'ISV-dedupe" (by buying the CV product), or source-dedupe (by buying Puredisk or Avamar), then you should say that. What you're currently doing is confusing and drawing the ire of people like me.

Seems like both vendors were blurring the lines of what their solution was capable of.

Going back and looking through all the comments and articles once again, I can see what you are talking about Curtis. There is confusion about what we announced and what we see as the future, and it was most certainly unintended.
The DL2000 powered by CommVault is not source deduplication, but instead compression at the host, deduplication on the media server, disk storage and integrated tape support. We do believe that deduplication is going to spread to more places, beyond just as a dedicated appliance, i.e. at the source, in the ISV, as well as other places, but did not intend to imply that the DL2000 is a source-based solution.
I do disagree though with your last comment, "you say that you can provide source dedupe.....but the casual reader is going to think that you mean to say that CV does source dedupe" if you are referring to this from my last comment, "One size does not fit all, and we can provide solutions that are source-based, ISV-based and target-based." No product or solution is called out there and no one product is all of those things. Dell, just like many others, provides solutions that leverage integration with products from their partners, and you are correct, we provide solutions including Avamar, PureDisk, Simpana, EMC and Quantum, and these are called out in our solutions we create with customers. Instead of listing out all the different ways that Dell can help with deduplication, including through our partners, I think it is more important highlight that users should first understand what the problem is that they are trying to solve, then think about the options available to address it.

One could always, and many often do, install the CV dedupe code on the host and deliver the dedupe at source if the host warranted it... its only a license and a small local DDB cache that's required. Not that different to others really... perhaps add a local and remote NAS target, facilitating a local disk based backup with remote DR copy also, with both being deduped....want it faster? buy quicker disk and or faster servers and use 10gig...

@gregorydwhite

My biggest goal was to get both CommVault and Dell officially on the record that what they're doing is not source dedupe (yet). That goal has now been accomplished. So the next time I hear a Dell/CV sales rep telling a customer that it IS source dedupe, I'll correct him/her and point them to this podcast and its comments.

As to my other comment, I was just using it as an example of the kind of thing I was talking about. Marketing people have to be very careful the way they say things as to not mislead people -- unless that's their intention, of course. My belief was that the announcement and all our conversations on this blog have been about the CV offering. I felt that for you to just say "we do offer source dedupe" without clarifying that it was coming from something other than CV continued to muddy the waters. I do get your point that a Dell marketing person probably doesn't focus on individual solutions, rather they try to communicate the overall vision. It's just that this conversation WAS about one particular part of the solution and what it does or does not do, so I would think it would be appropriate to call out that source dedupe, while available, does come from another solution. Something like this would have been nice, "While the DL2000 does not do source dedupe, we do offer it via products from other partners." Just a thought, not a complaint, per se.

@jeremy

This is correct, but I'm assuming (although haven't asked) that a dedupe media agent license is significantly more expensive than a standard client. So while what you say is technically valid, I don't believe it's financially feasible for the backups of, say, hundreds of laptops, which is what a source dedupe product is great for.