Ever-expanding data volume continues to challenge many IT organizations, and storage administrators are increasingly unsure if 100 percent of their data could be recovered in the event of a loss. The Dell™ PowerVault™ DL2100 Powered by CommVault® Simpana® software offers a comprehensive backup-to-disk backup and recovery appliance that includes built-in deduplication and a platform designed to support the growth of an organization. CommVault Simpana software offers a single platform for unified data management using a single interface. This discussion focuses on deduplication and other features such as integration with virtual environments and replication. Darin Camp, senior technical alliance manager at CommVault, is guest moderator for this discussion.

Technical Community - Background Reading

Chat Transcript

Dell-JeffS First, this is an informal chat. Feel free to jump in with your questions at any time. The chat transcript will be posted in a couple of days, so no need to take notes. Today's topic is centered on the Dell PowerVault DL2100 Powered by CommVault. We have a couple of guest moderators from CommVault, who have spent some time over here in the lab recently. Darinc_cv and Don_foster are here to take all the hard questions, and I'll take the easy ones. Okay, that being said, we have a number of things we can talk about: deduplication, backup to disk, and some inside information on a new offering, Snap Backup for EqualLogic.
bradfair Ooh, I'm loving the sound of Snap Backup for EqualLogic
Dell-JeffS Heck yeah! Anyone doing backup to disk? Or anyone have a PowerVault DL2100 (or PowerVault DL2000)?
erson We're doing disk to disk to tape with Microsoft System Center Data Protection Manager (DPM) 2007 and a Sony Ait-3 tape loader
ceri We're currently doing disk to disk to tape with EMC NetWorker and a Sun X4500 as a storage node. Data goes to tape as fulls after a week, if I recall correctly
erson We're using a PowerVault MD3000i with nearline SAS drives as the back end for DPM
ceri Could we slot the PowerVault DL2100 in as part of the NetWorker solution?
Don_Foster How long are you typically retaining data on disk in DPM and NetWorker?
Dell-JeffS We managed to convince the product group marketing team to give us a PowerVault DL2100. Then Darin and company volunteered to come over and show us how to make it purr
ceri Currently, we can keep it there for a month or so. We're licensed for 10 TB of disk storage
erson Don_foster, it depends on the data, but between 14–60 days
Dell-JeffS Did a few tests with our fileservers and VMware
David_Paquette How often are your backups running with DPM or NetWorker? Nightly?
ceri Yep
Don_Foster What is your typical size of a full backup, Ceri?
erson Also depends on the data, but for the documents that change most frequently every 15 minutes. I'm interested in doing host-level backups of guest VMs in Microsoft Hyper-V R2. Is that something that the PowerVault DL2100 and CommVault can handle?
ceri Don_foster, ah, there's the rub. ;) We are a Russell Group university and have no typical application. Full backups range from ~20 GB up to 2 TB per server, but generally about 50–70 GB per server
Dell-JeffS Good question, Erson
Dell-JeffS @ceri, sounds like a prime target for dedupe
Don_Foster Erson, we can provide disk-level protection of your Hyper-V R2 environment to disk, leveraging the PowerVault DL2100 Appliance and dedupe the data on the backend disk store. Restores from these data protection operations will allow single object and file recovery as well
DarinC_CV Ceri, with our file-aware deduplication we do very well in environments that have a large mix of file types
JasonPowell I still need something that can dedupe lots of video files :-)
David_Paquette Does the dedupe use fixed-length or variable-length data segments?
ceri Dell-jeffs, Yep. We are planning an EMC Avamar proof of concept real soon now. If there are other options such as this product, I'd like to look at those too
Don_Foster David, we are content aware
erson Don_foster, so even if I do host-level backups of guest VMs (pretty much VHD files then) CommVault Simpana can dedupe the contents of the VHD files?
Dell-ScottH Hey guys! VCP class is talking about how to create a VM <yawn>. Thought I'd drop in…
erson It would be very advantageous to dedupe the OS in every guest VM—a lot of redundant information there
DarinC_CV Ceri, I would definitely check out the PowerVault DL2100 for backup to disk or tape with embedded deduplication
ceri Erson, yeah, you'd think bit-level dedupe would be most efficient ;-)
Don_Foster Unlike the typical target-based dedupe solution, where the dedupe appliance must deconstruct and identify unique and non-unique segments on the fly, we are able to eliminate known non-unique data sets from the dedupe hash process to ensure you are only storing unique data
erson Scotth, I'm looking forward to you telling me how to use the Dell repository manager and make Dell OpenManage Server Update Utility (SUU) updates that I can use with the Dell Unified Server Configurator (USC). It didn't work at all for me, but at least I got a pretty error message
David_Paquette Don, I think you are saying you provide inline dedupe during the backup from a hash catalog?
Don_Foster Erson, that is correct
ceri Sounds interesting. So how is the PowerVault DL2100 presented? As tape?
DarinC_CV Ceri, agents are installed on the host machines and communicate with the media agents through our own protocols. The PowerVault DL2100 acts as both a media agent and the CommServe
ceri Darinc_cv, okay, so the hash catalog lives on the client as well as the PowerVault DL2100 for performance reasons?
Don_Foster David, as we compile our data stream on the client the hashes are only executed on live data
DarinC_CV So you can have Windows, Linux, Unix, Netware hosts all backing up to a dedupe data store on the PowerVault DL2100
ceri Darinc_cv, okay. I couldn't just drop this in under NetWorker then it seems? (Not a complete barrier, but need to know)
Don_Foster Data stream tag headers, index, and control information are automatically excluding this from the hash process
David_Paquette If I dump a deduped backup onto tape, does the tape contain all of the metadata necessary to reconstitute itself during a recovery?
erson /me corrected the URL on the PowerVault DL2100 spec sheet: www.delltechcenter.com/page/01-12-10+dell+powervault+dl2100+powered+by+commvault
DarinC_CV Ceri, correct. This is not a dedupe appliance but a complete backup and archive solution that includes a dedupe option
Dell-JeffS Thanks, @erson. Show off. :)
ceri Darinc_cv, okay, thanks
erson Jeffs, yeah, I should start doing that every chat :)
Don_Foster We then are able to compare segments of files, databases, VMs without the overhead of trying to strip out the metatags, and so on from the data stream
ceri Any significant overhead from reconstruction when doing restores?
Don_Foster David, our dedupe to tape essentially is a backup of your dedupe store
erson Are any features purchased separately?
Don_Foster If you decide to move your dedupe data to tape, all of the information required for a restore is available as a part of your CommVault infrastructure (PowerVault DL2000, etc.) and the tape media itself
Don_Foster Ceri, great question!
David_Paquette I guess the question I am really asking is if I carry the tape to a different PowerVault DL2000 will the tape have everything it needs to restore?
Don_Foster Our dedupe process is managed in a way where our dedupe database is check in only
DarinC_CV Ceri, nothing significant. We do not use the dedupe database on recovery so our recoveries are not hindered by having to go through the dedupe database, which makes us much faster
Don_Foster Thanks, Darin
Dell-JeffS @erson, I'll see what I can find out about separately purchased features
ceri Interesting, thanks
Don_Foster David, yes, in a dedupe-to-tape scenario, you would have everything you need to restore the data on it, however data that has been moved to dedupe tape must first be propagated back to a disk target prior to recovering
ceri Okay. Licensed based on data volumes, or can I throw as much at it as I have disk available?
Don_Foster Dedupe to tape provides a phenomenal way to reduce the amount of tape and storage you may be using for your long-term data retention copies
DarinC_CV Erson, yes, there are two main offerings: an entry-level offering that does not include dedupe by default and an advanced offering that includes dedupe and revved up hardware
David_Paquette If I had two PowerVault DL2000 appliances, would I be able to perform global deduplication between the two?
erson Can I attach any PowerVault MD1000, PowerVault MD1120, PowerVault MD1200, PowerVault MD1220, and PowerVault MD3000 direct attached storage (DAS)?
ceri Dedupe to tape seems sensible to me. Many products seem to un-dedupe when going to tape, which is a bit annoying (although, I can see why they do it)
Don_Foster Yes, but it depends on what your interpretation of global dedupe is
ceri What's a typical block length, and what algorithms are you using to detect collisions, please?
Dell-JeffS @erson, today no PowerVault MD series support. Stay tuned though…more details in February
David_Paquette I guess in this case global dedupe means that both boxes would share data and hashes
Don_Foster Ceri, dedupe to tape is great if you are not expecting frequent and fast recoveries from tape
erson Darinc_cv, thanks
erson Jeffs, so only PowerVault MD1000 support today?
Dell-JeffS Yes
Don_Foster So, yes, we could configure multiple PowerVault DL appliances together in a CommVault grid (a feature we call Gridstor) to allow the storage to be leveraged and deduped across both targets
erson Any plans to support disk storage over Internet SCSI (iSCSI) instead of DAS?
Ceri Don_foster, yeah. I expect fast recovery to be required for disaster recovery purposes, so in theory it should still be on disk anyway in that circumstance
Don_Foster Block sizes are defined by storage policy. And up to you, the end user, to choose
Dell-JeffS @erson, you are going to get us in trouble...again, February may have an answer for that question. :)
Don_Foster As small as 32 KB to as large as 512 KB—this enables you to choose what level of performance versus deduplication storage savings you need
Dell-JeffS Excellent questions
David_Paquette Is the dedupe based on disk blocks or other file-based data segments?
Don_Foster Ceri, exactly, and you could replicate one PowerVault DL2000 to another without re-hydrating the data to allow you to have a disaster recovery (DR) available copy
Don_Foster David, it is segments, so we break the objects and files into segments and hash from there
Don_Foster Quick question to turn the tables on your guys :) : Do you have any needs to encrypt your backup data, or parts of your backup data sets?
erson I reckon disaster recovery almost always means that you want the latest full backup, so you should never have to go to tape for that
Don_Foster Erson, you are correct. Dedupe to tape is not the best solution for DR
erson Don_foster, we have been discussing encryption for a small subset of the data we back up
Don_Foster Depending on infrastructure, bandwidth, etc., we can put a solution together that may work, but the sweet spot for dedupe to tape is condensing the long-term MF and YF backups
Don_Foster Erson, how do you think dedupe will work with those encryption algorithms?
ceri Don_foster, we don't need encryption in the main. I can see that a requirement could grow for it, but can't comment any further
erson It would probably depend on the performance hit as well. If it was an elegant solution I could see us encrypting more than the absolute must
Don_Foster By having embedded dedupe in our software, we can compress and encrypt data and still provide the same levels of dedupe ratios on the backend
David_Paquette Preferably you want to encrypt after you dedupe
erson I agree with David
Dell-JeffS Is now a good time to bring up the Snap Backup with EqualLogic? FYI, we weren't sure we could talk about this until just the other day. Don, care to elaborate on what's coming?
Don_Foster David, CommVault can allow just that as well
Don_Foster Anybody leverage or want to leverage snapshots in a way that they become more strategic in your environment, other than a pit recovery copy?
JasonPowell Bring us more EqualLogic snap goodies!
Don_Foster Snap Backup is CommVault Snapshot enabler that integrates shortly into the Dell EqualLogic array without any scripts
ceri Don_foster, we currently use Novell Netware for file servers, which allows users to recover their own deleted files; so if there's anything that could help there with, say, Windows...
David_Paquette I'm curious about how that would work since the normal EqualLogic snapshots are block based
techanalyst Are these backups performed through scripts at the CP, or is there the GUI option as well? Can someone address automation of backups and verification?
Don_Foster This allows you to schedule our snapshots internal to your data protection suite, and then leverage them for standard data protection operations
techanalyst In this environment that is...
Don_Foster The snaps are all GUI driven
Don_Foster Both Dell and CommVault have APIs that work together to provide a seamless process for executing and leveraging snaps
techanalyst So we would be looking at dashboards?
David_Paquette Would those be volume-based backups?
Don_Foster David, correct. Hardware snaps are all block based; however, after the snap is completed we would mount the snap to an MA (PowerVault DL2000) and then perform an index of the online snap
David_Paquette That makes sense
Don_Foster That will give us a catalog of the versions of files on EqualLogic for each snap copy, and based on your policy promote them to dedupe disk and/or tape for nightly backup. It is application aware for applications like Microsoft Exchange, SQL Server, Oracle, SAP, and so on
David_Paquette What is the processing impact of the inline dedupe?
Don_Foster For EqualLogic customers we can also automate this on Linux as well
techanalyst If so, can we selectively choose segments of the data to recover? Are the blocks searchable?
Don_Foster Techanalyst, since we are physically mounting the snapshot off host we are providing a filesystem context of the blocks
David_Paquette How is it application consistent? Are you using EqualLogic snapshots to call the Microsoft Volume Shadow Copy Service (VSS) writer on the source servers to quiesce the data?
DarinC_CV Ceri, with our Workstation Backup Agent (WBA) we can allow the end user to schedule backups and recover individual items for their laptop or desktop
Don_Foster So you can search multiple snaps for a single file or database and CommVault will restore just that file you chose from the correct snapshot?
techanalyst Okay!
ceri Darinc_cv, okay
Don_Foster David, CommVault has written its own VSS Software Provider that interacts with the different hardware platforms seamlessly
Don_Foster This allows us to quiesce Exchange, SQL Server, FileStream, Active Directory, and so on
David_Paquette Do you need agents on those servers, or is it done remotely?
Don_Foster It directly integrates with our application agents, so you can still perform log backups and initiate Pit restores. CommVault will replay logs and so on, on its own from the proper storage location. We do need agents to execute his process. For example, you could snap a SQL Server four times a day
techanalyst Are these virtual agents?
Don_Foster We would catalog the database instances at each Pit, and the fourth snap would then generate an automated off-host backup to the PowerVault DL2100 dedupe disk. Meanwhile, you can run 15-minute transaction log backups to disk
David_Paquette So the benefit of using Snap Backup is that it is offloading the backup process from the application server to the backup server?
Don_Foster And CommVault can restore to any point in time by leveraging the hardware for a revert of the most recent snap copy and then replaying the logs
Don_Foster David, that is just part of the benefit
techanalyst Would it not be cool if this environment would allow us to show graphics?
David_Paquette Sure, Don, just clarifying
Don_Foster It also enables you to leverage the hardware speed in reverting volumes and performing large database recoveries—all without a single script
Don_Foster Techanalyst, I agree ;)
Dell-JeffS @techanalyst, watch the Dell TechCenter site; we should have some goodies up soon
ceri Okay, so if we want to know more?
Don_Foster Oh and indexed snap copies; you don't have to find the needle in the haystack when you want file version X. You just search in CommVault
David_Paquette Great information. Thanks to the CommVault and Dell guys for putting this on
techanalyst I intend to, thanks. We should also have the option to be notified through SMS of interesting events
DarinC_CV Ceri, contact your local Dell-CommVault account team and they can hook you up
Dell-JeffS Thanks for all the good questions!
ceri Darinc_cv, alright. Thanks to the CommVault chaps for fielding all the questions
Dell-JeffS Okay folks. Excellent session. Transcript should be up in a day or two. Next week it’s 6 Gbps SAS
Don_Foster No problem!
Dell-JeffS Same bat channel
Don_Foster Thanks for your time!
DarinC_CV Thanks, everyone!
Dell-JeffS Thanks again and see you next week
techanalyst Have a great week everyone! Thanks for the warm welcome
erson Thanks to Don and Darin