Hu has a post up about HDS's ProtecTier VTL offering which caught my eye. Lets watch..
I recently saw an article by Beth Pariseau about reasons why people are not buying deduplication products. The main reason given was the slow speed of backup compared to non deduplication VTL solutions. Compared to tape libraries and VTLs which top out around 500GB/s, the Diligent ProtecTier solution is cited at 220 MB/s. While that is better than the 100MB/s that is cited for Data Domain, it is judged to be too slow.
That might be because for Enterprise customers it is too slow. I'm not saying that it's too slow for all customers, just customers who have large environments and require a higher transfer rate than what's effectively performing on par with four LTO 3 drives. (LTO 3 having a native drive transfer rate of 80MB/s.)
I have another take on this. The speed of backup was a concern when backup was done during backup windows. Today most users will create a snapshot or clone copy and do their backup in the background without the need for backup windows.
Backup windows haven't gone away, they're alive and well and still causing people pain. Indeed VTLs evolved from the need to help customers get their backups done within their window by leveraging the speed of backup to disk but without having to drastically overhaul their existing backup environment.
As is mentioned further down while the backup is fast the restore is just so much faster as it's coming off disk and you're not dealing with multiplexing or mount operations. The other issue is that customers might not have chosen or just can't deploy Array based instant copy (Snap/Clone/BCV/Whatever) technologies across all the systems they backup.
For some folks it's a mission critical thing only, and involves taking numerous significant point in time backups during the day to ensure a fast restart. Ideally customers doing that should be looking at CDP, but that's another discussion. There's also a lot of DAS and NAS out there which needs to be backed up, the DAS pouring across the LAN while the NAS devices might be streaming off via NDMP over FC. The NAS devices might be using snapshots, they might not. So while backups do happen in the background a lot of backups don't, and people are still playing beat the clock with their backup window.
The main concern today is the recovery speed of backups. For recovery, we achieve 400 to 500 MB/s using one ProtecTier in front of an AMS 1000 modular array. So while the backup time does take longer today, it can be done using a shadow image copy, without impact to the application, and the recovery times are still comparable to other VTLs. The major advantage of Diligent’s deduplication is to reduce the amount of data backed up by a factor of 25 to 1, and that is money.
Spot on with the money thing but it's my understanding that ProtecTier doesn't reduce the amount of data backed up, it reduces the amount of backup data retained. Yes I'm nit-picking and that isn't what Hu meant I know, but it's an important point to mention about VTL de-dup. Everything still gets backed up even if only the unique pieces are kept. Avamar on the other hand does reduce the amount of data backed up as it de-dups globally, both at the client during backup and at the storage target. So you could see a de-dup ratio in the order of hundreds to one before it leaves the client, as it's only sending subfile changes and not the entire file containing the change, and up to 30:1 when all data from all the clients lands on the back end. So you've reduced your storage requirement when the data is at rest as well as drastically reducing the bandwidth requirement when the data is in motion.
By writing this post I haven't set out to bash anyone's solution. I know that there are people out there quite happy with ProtecTier, but I would like to point out that the performance levels we're seeing when de-dup happens at the VTL could be significantly better than they are now. We're starting to see a number of different approaches spring up in the market, Sepaton with DeltaStor, FalconStor with their Single Instance Repository, Quantum have the Rocksoft Blocklets technology now shipping in their DXi series, Data Domain have their new DDX array, Diligent/HDS are going to add clustering, and so on, but when I look at the EMC DL 4000 series and see that the throughput of something like the DL 4400 is 2200MB/s you kind of realize that for some approaches there's a very long road to get to that level of performance, and all the while customers keep asking for even faster VTLs.
The question I suppose this raises is where is EMC's VTL de-dup offering? I'd imagine EMC will answer that question soon enough.
Making predictions would take too much of my time away from Marvel Ultimate Alliance on the XBox 360, and right now The Avengers are in the middle of battling Loki for Asgard.
It's important work.
As EMC exits 2006 I'm confident about EMC in 2007. This is down to the fact that internally everyone who should be speaking to one another now probably is speaking to one another.
EMC's internal communication channels are alive and humming with people from all parts of the business. Be they people who have been there for years or people who have just joined us from acquisitions, and new ideas get kicked around all the time.
It hasn't all been smooth sailing as you can imagine, some companies found their place in the grand design quicker than others, and I remember when some of EMC's earlier purchases might as well have been located on Neptune for the level of interaction I saw coming out of them. Thankfully that's gotten a lot better as time went on, and these days people are not hesitant to go looking for opportunities to work with other groups. The more successes they have together the more successes they want, the more they want the more people they have to speak to, the more people they speak to the more ideas they generate. The more ideas they generate the better it is for EMC customers.
2007 is going to be interesting, the company has been generating a lot of ideas and is working on a lot of stuff.
Lets see where it takes us. :)
Sangod's comments page throws a horseshoe while trying to leave a reply so I'll post it here instead.
Don't sync your clock inside the guest OS, sync the host clock and then enable time sync via the check box option in VMware tools instead. (Which you'll need to install into the guest OS.)
If you're running without graphics open the guest vmx config file on your host and set..
tools.syncTime = true
..or some such. Haven't poked around in there for a while just look for the syncTime option.
You'll lose basic VM time sync if the VM can't grab enough CPU cycles from the host for some reason, but if I recall VMware Tools will check the host clock once every 60 seconds and then sync the guest if it's fallen out of step for any reason.
And in response to this comment:
I believe it. I worked for EMC when the whole DataGeneral aquisition took place, I’ve seen what the Clariion was and what it’s turned into and I have no doubt that EMC has put a lot into it.
I’ll probably consider the upgrade once I have production off it - “if it ain’t broke, don’t fix it” is a great motto for a business that gets most of it’s traffic from the web.
Once the production clusters are off the Clariion, it will just be Celerra and Veritas owning the stores. That will be a little more worth the risk, though losing our NAS would be a painful prospect.
I’m thinking about moving from my current network backup to NDMP, have you, (or anyone else) used it and is it worth all the hype?
I don’t like the fact that I can’t do NDMP direct to disk, as our current backup environment is 100% disk->disk->tape and I don’t want to get away from that.
Heh. :) My first job at EMC was troubleshooting Symmetrix ESCON Directors to component level, you should have seen my reaction to the first FC4700 I ever took apart. "Where's the midplane? OH MY GOD SOMEONE MUST HAVE STOLEN THE MEMORY BOARDS!" I was told there weren't any and that you could administer it via a web browser, I thought they were joking. It's been a long way, the best part of a decade now, from there (which was 6%), to here. (Which is a commanding lead in the mid-range market.)
As for NDMP yes I have used it and no I don't like it but it's the fastest way of getting your data backed up from a NAS device. It's annoying that the NDMP standard was a product of Legato & NetApp and yet neither of them could come up with something which wasn't so bleechhh.
The core problem with NDMP though is that it works.
I use it, I don't like it but it works, and as such no one is in a big rush to replace it. NDMP isn't a tape only prospect for EMC NetWorker shops as EMC added NDMP B2D functionality back in NetWorker 7.2, so NW customers can B2D2T NDMP backups.
You were asking about the advantages of NW a while back, here's another one. ;-p
In case you haven't heard yet VMware Workstation 6.0 has moved into Beta bringing a ton (I nearly prefaced that with the F-word), of new features with it.
If your worthless family or no good kids happen to give you socks again this year for Christmas you don't have to ignore them for the rest of the day by vegging out in front of the TV and glaring at them when they attempt to speak to you. You can now slip on some headphones and turn your back to them while setting up your Virtual Machines to automatically act as VNC servers instead.
That's my plan for New Years Eve sorted, so what's yours?
Steve Duplessie wants EMC to fire 5000 employees. Merry Christmas to you too Steve, be aware that Santa's bringing you a lump of coal this year.
After EMC cut 1000 employees recently the reaction from the Storage Blogosphere was resoundingly negative, now I'd hope the Blogosphere realize that the street & analysts never get worked up about firing too many people, but are happy to chip in when they think you're not firing enough people.
Since Chuck gets in depth with the EMC/Microsoft alliance I thought it would be interesting to talk about some of the work the people who I work with have been doing at the engineering level. Josh mentioned this on his blog already but a good story deserves repeating.
What EMC/MS Exchange customers and the field are ecstatic about at the moment is EMC's contribution to the MS: Exchange Solution Reviewed Program. Microsoft's ESRP is the framework EMC uses to test and optimize it's solutions for Exchange so a lot of engineering time and effort was put in by Engineers in the Global Solutions Organization, Field Councils, and Microsoft's Exchange Division itself to apply the ESRP testing methodology across EMC's storage platforms. The results are acquired using Microsoft's testing methodology, validated, and published according to Microsoft's procedures. No mucking around or gaming the system, just reproducible results and expected performance characteristics tested by Storage & Exchange people for Storage & Exchange people.
If you're running Exchange on externally attached disk you should give them a look. If you're a fan of enough numbers to be dizzying you should probably also give them a look.
It takes a lot of time and effort by some very talented people to put these things together but the feedback has been so positive it's worth doing more of.
Since VMware Fusion has seen a level of interest from customers which is just head spinning, Mac customers will be interested to hear that VMware is actively seeking more Mac developers to join the rapidly growing Fusion team.
I came to MacOS X via NEXTSTEP, so watching Apple claw it's way back from the brink has been very satisfying, indeed in some cases I've been one of the internal noise makers for increased levels of Macintosh support by EMC but for various reasons it's always been difficult to make the numbers work and what with Apple's complete lack of focus on the enterprise market,
That being said one of the areas where EMC could be serving it's Macintosh customers better is with Retrospect, I'm a Retrospect user myself across all the systems here at home and it is somewhat galling to see the Windows only 7.5 Server versions just having so much more functionality than the Version 6 Mac offering. Yes I'll guess there are probably orders of magnitude more Windows Retrospect customers, but here comes that emotional argument again.
I don't speak about EMC's future plans, that's for EMC to speak to you about, but be assured that while I'm a Mac user I'll be on the case of everyone inside EMC providing a Mac product which I use. Big words from a guy who doesn't have a title or a corner office I'll admit, but you've never seen how much noise I can really make and I only have to win one argument to justify all the previous ones. ;)
Seriously, back in the boom when I did nothing but live in SSH sessions Solaris mattered. It mattered hell of a lot. Unless you were a Telco (Usually HP-UX) or a TruBlu shop (The term I use to refer to wall to wall IBM environments, so we're talking AIX in this case) any organization above a specific size was probably a Solaris shop. Solaris 2.6/7 and Windows NT being the foundation for many companies during the dot com era.
Post crash we have Sun hyping ZFS as well as considering slapping the GPL onto OpenSolaris, and yet I still see more people throwing out Solaris than adopting it. Indeed even I have stopped replacing Solaris instances when systems are retired, I just throw the workload into a Virtual Machine or onto a physical x86 host running Windows or Linux.
As for ZFS well some folks think it's the universe in a file system/volume manager combo, and I'm sure Symantec have been taking a long hard look at VxFS and planning a ZFS counter punch, but if you're a Windows or Linux shop chances are none of that will matter and you'll just run what's native. Be it NTFS or EXT3/EXT4
Perhaps ZFS is a bit too Star Trek for it's own good? For the most part people don't usually rush to jump from well understood technology to new technology. The higher the stakes the greater the resistance to change.
I'm not going to favor one product over the other in this post, though I would like to direct your attention the CERT metric attached to the advisory Decru are attempting to make hay with:
That's zero-point-six-four out of a severity maximum of 180. 0 being the least severe while 180 is doomsday. Since it's one of the first things I check when I read an advisory it was interesting to see that it's not included in the Decru's note to NeoScale customers and anyone else Decru think should know about it.
The fact that this is being used as a competitive talking point at all is bizarre, though I've seen Neoscale try and hype similar non-issues.
It is interesting to note that Decru appears to have taken a page directly out of the NetApp marketing play book by going straight to NeoScale customers with a "letter" and shaking the press tree looking for a few quick stories to fall out. NetApp have a habit of sending out error filled missives which are usually low on facts but high on slanted opinions and selective omissions when they think they can take a cheap shot at a competitor, Decru may have picked up a similar bad habit.
Such fuss over non-issues like these do make one wonder how much progress is really being made in the appliance based encryption market if minor bugs like this, resolved in the currently shipping firmware we're told, are all they have to throw at one another?
Two things I might need to do.
The first is register for the RSA Conference. Thousands of super paranoid security professionals all at the same event. The most frequent answer to the question "What do you do for a living?" being "Why do you want to know?!?" ;-p
Once again Gates and Ellison are keynoting. I like Larry Ellison he's the closest thing we have to a Bond villain in the I.T. industry.
Just a few months after RSA wrap up the US show we have the EMC Technology Summit, Software Developer Conference, and Documentum's Momentum show all coming together under the banner of EMC World. Or the sore feet and no voice show as I like to call it. By the end of the first day my feet are usually screaming from all the standing and walking around I end up doing, and I've blown my voice out from speaking to countless EMC customers.
If I attend EMC World this year, and I'm assuming no such thing as it's far far away and requires months of planning and effort, the second thing I need to do is buy more comfortable shoes and hire someone to speak on my behalf. Or just communicate using sock puppets.
Preferably sock puppets made from clean socks too.
