We have a D2D 4106FC, MSL 2024 and Data Protector 6 server connected to a San switch. I am looking for a document which shows about how to integrate D2D with Data Protector and configure backup jobs first to D2D and after that move to MSL 2024 Physcial Tape library.
IMO you should read at least the "D2D Best Practices for VTL, NAS and Replication implementations" paper to get an impression of how to best deal with the new D2D backup appliances. Heaving read that, you may question your idea of using the D2D as a VTL (which is implied by buying the FC model), but there may be reasons for that (a requirement of SAN based backup mostly). There is also a guide for integration with DP but it's for NAS mode. It's called "D2D NAS Integration with HP Data Protector 6.11". I don't know of a guide for DP integration with D2D appliances in VTL mode, but maybe there are other VTL guides available that would apply.
Finding the docs in HP's maze of web sites is a challenge, for instance they are not available from the easily navigated "Enterprise Library". I navigated thusly: Use the "Support & Drivers" entry point, give it your product name (d2d 4106fc), when there, from the left column click on manuals. That's where the downloadable documentation resides, and the documents should be there.
Regarding D2D integration, I can only speak for NAS integration into DP as a File Library, as I consider that the superior operating mode. IMO VTL is just for use cases where you *must* use an emulated tape. One other advantage of VTL is in replication scenarios, where media from multiple source VTLs can merge into a single destination VTL, something that cannot be done with NAS.
The general way to implement things in DP (whether by VTL or NAS-based File Library) is to create the D2D backend device, make it available to the OS below the DP installation (mount the NFS/CIFS share, connect the iSCSI VTL or make sure the FC VTL is visible to the OS), attach to it from within DP (create File Library or Tape Library, in the latter case, choose VTL licensing mode) and prepare pools and media as usual (not much to do here, as media is almost automatic with both). Then target the backup specs in question to File Library writers or VTL drives as required (create lots of writers/drives in the first place, so you gain parallelism and flexibility) and chain a post-backup copy to tape to those backup specs.
I dunno anything about VLS, but assume the difference is that D2D appliances come with builtin deduplication and offer NAS integration options, opening them for use cases beyond VTL.
Hi. Why use NAS mode for D2D if HP says in mentioned document that In VTL mode backup perfomance better? For example, we have D2D 4112 and want copy VMware over SAN to it. As i understand i must use VTL mode for D2D if i want copy from VMware ESX to D2D over SAN? I want to understand what are best practicies for D2D in VTL mode - what emulation type, count of drives and slots, DataProtector device parameters are better for maximum perfomance?
Igor V Abzalov wrote: Hi. Why use NAS mode for D2D if HP says in mentioned document that In VTL mode backup perfomance better?
Well, backup performance is a multi-dimensional issue. Just saying "VTLs are always faster" is simplifying things too much. Thus every best practices guide has to be taken with some grains of salt, especially if the environment it focuses on is different from yours in certain key aspects. For instance, I always bump into people who bluntly consider SAN to be faster than LAN. That is often a false assumption when applied to the backup world, as SAN backup serializes what could otherwise be multiplexed. As in most real cases, bottlenecks are neither at the destination nor at the network, but actually at the source of backup data, enforced serialization makes things worse. But, of course, you might have the right environment to feed SAN based backup at good performance, so that may still be the best solution for you. It depends.
I prefer NAS mode as that gives me more flexibility. I could, for instance, plug more than one DP File Library into one and the same share, increasing dedup capacity (paying for that with performance penalties). I can have way more active media in a File Library than I can in even the most generic VTL emulation. That may seem paranoid with 1024 slots, but given you have a never-append media policy with lots of parallelism, they are eaten faster than I felt comfortable with. If the stats are right, I've got 794 active media on my File Library with just 3 weeks of protection, a 4106i class D2D and at approx. 70% backend storage fill state. So I wasn't that paranoid after all it seems.
For example, we have D2D 4112 and want copy VMware over SAN to it. As i understand i must use VTL mode for D2D if i want copy from VMware ESX to D2D over SAN?
Yep, as I said - direct SAN feed into the D2D appliance is one reason to buy the FC box and consequently use VTL mode. And it actually makes sense in a VM environment that has fat machine images laying on SAN-attached storage and can arbitrate a block-level snapshot access to them for fast backups. Works well as long as storage is fast and serialization doesn't underfeed the D2D. The D2D allows for parallelization without multiplexing by simply using lots of independent destination "drives", ideally the source is capable of using that (I remember VCB proxy could only access one VM at a time, that would again have a potentially large serialization cost). But then, there isn't that much choice of how to perform VM backups otherwise, so the architecture may still be the best solution for the problem at hand.
I want to understand what are best practicies for D2D in VTL mode - what emulation type, count of drives and slots, DataProtector device parameters are better for maximum perfomance?
Mostly what the D2D best practices paper says. IMO the only sane VTL emulation choice is the generic one with 1024 slots and up to 63 drives (maybe 64 with FC, IIRC 63 was an iSCSI limitation). You need as much drives as possible for decoupling sessions and still maintaining parallelization, that eats into the total number of drives really quickly (four backup sessions to 4 drives each will lock up 16 drives for decoupling, if they are later copied from original drives, you are at 32...). You want as much slots as possible due to the media allocation strategy of VTLs as I understand them (I may be wrong, though). AFAIK, a VTL has exactly as many media as there are slots. There is no automatic way to deal with an external pool of media and request virtual cartridges move in and out of the library, so no way to implement a virtually unlimited reservoir of virtual media. As a result, everything in the entire protection cycle of the VTL has to fit into the available slots. Correct me if I'm wrong...
As for device parameters, you best chose the same (high) block size your backend tape drives have, so as to make copies more effortless.
But you are right, a paper about how to best attach a VTL (specifically D2D in VTL mode) to DP which ponders more deeply on these issues would be very welcome.
What i don't understand in document is why HP recommend to make virtual tape Non-appendable? If we limited in number of tapes = number of slots why close tape after each backup session? Why not copy all virtual machines for example every day to only one virtual LTO5 1600 GB tape, and switch to second tape only when first is filled?
Igor V Abzalov wrote: What i don't understand in document is why HP recommend to make virtual tape Non-appendable? If we limited in number of tapes = number of slots why close tape after each backup session? Why not copy all virtual machines for example every day to only one virtual LTO5 1600 GB tape, and switch to second tape only when first is filled?
From what I concluded (reading between the lines of the paper) an append operation is costly and maybe inefficient on files in the dedup virtual filesystem the D2D boxes create internally. They warn to never use append in the NAS application case, either.
There may be another reason: Do emulated tapes actually ever fill? It reads as if the emulated media type (LTO class) is mostly irrelevant and virtual tapes could perhaps be written forever. That would clearly make append mode unusable.
Even if I interpret things completely wrong, there still remains the operational advantage of single object media never being locked from expiration due to other independent objects with longer protection on the same medium.
I currently have a D2D4324 in use as a VTL and am planning on more. From my reading the recommendation to use "do not append" helps performance when remote replication is being used. With a "tape" that is appended to when using remote replication, the entire tape has to be processed and reviewed to find changes from the copy at the remote site. So with a "do not append" tape it can be processed and replicated once instead of being processed every time a new backup is appended.
In addition to the "Best Practices" manuals there is also a "HP Data Protector software interaction with deduplication enabled replication" manual (white paper) that may be of interest and is oriented for use with the D2Ds.
With my D2D backup performance is great but I do have a performance issue when coping backups from the D2D to a tape to send off-site. We are still working toward doing the remote replication and eliminating tapes.
Sure you could, but this reduces the effectiveness of housekeeping and deduplication in the D2D as well.
Always keep in mind, that a single non-appendable media is free immediately as soon as the protection of the objects is gone. If you have multiple objects on the same media this could take much longer depending on protection used. From a storage perspective in a D2D, there is no big difference between 500 non-appendable or 100 appendable virtual tapes. Both will occupy nearly the same capacity but the non-appendable library will be much more efficient.
--- Please assign a KUDO to this post, if you find it useful.
If you need more capacity or slots, just add more slots/tapes to the VTL using D2D WebUI and expand the repository using Data Protector GUI. I recommend initializing the media manually. I use strict policy on all libraries in Devices & Media. Depending on backup type, schedule and protection the amount of required media will vary and you need to find the correct amount. But this will not hurt. Feel free to alter protection and schedule of your jobs to match business RPO requirements and change the sessions already stored in the VTL to find the right balance quickly. If you change data protection to existing objects in IDB, this can help to free up sessions you ran with the wrong protection. Expired media will be reallocated for backup. Using a free pool with a defined notification on it can help to reduce out of media conditions.
--- Please assign a KUDO to this post, if you find it useful.
Well, i'll change policy to Strict. What about number of media pools with D2D virtual tapes? Should i make different pools for File servers, MSSQL, Oracle and Exchange or use one pool for all D2D backups? We use D2Dbs emulation for D2D4112, is it good for use with DP? And we need some info about one tape size. We set it to 1600 because our Level 2 device is MSL2024 with LTO5 tapes. We want to make Object copy to tape after 2 weeks.
This is not a question related to Data Protector directly. As you might know, each share or VTL on a D2D system is an isolated dedupe store. To follow best practices, you should create seperate VTLs for different data types (if possible). This will ensure seperation of data - something you would do using pools on your MSL tape library.
Lets assume you have 500GB Oracle, 1500TB file server and 20GB of Exchange data. It would be best to create three VTLs (Oracle, File, Exchange), but 2 VTLs will still do to the job if you put the small amount of Exchange data into one of the other VTLs. This will help to optimize deduplication ratio and reduce unneccesary overhead. Multiple Data Protector pools will not help. You could create different pools for the VTLs, but this is not a requirement.
I use larger cartridges (3200GB) as they consume exacly the same capacity as a 100GB tape on a D2D - about 3,5GB. As your media is non-appendable in Data Protector, large datasets will most likely fill a single cartridge which is very efficient compared to hundrets of small tapes. This is also good for housekeeping and replication.
--- Please assign a KUDO to this post, if you find it useful.