Ask The Expert Summary: Deployment and implementation of Data Domain Virtual Tape Library (VTL)
This article summaries 2012 Chinese ATE activity: "Deployment and implementation of Data Domain Virtual Tape Library". The original thread is: https://community.emc.com/message/661694#661694.
How are the customers working with data domain? Is it by means of VTL/OST or NFS/CIFS sharing?
Across our global customers, all the methods you spoke of are used. The methods they chose depend on their requirements and environment. Technically, the VTL is based on fiber protocol while the others are based on network protocol. If you already have a SAN environment, and the backup server has a spare HBA card, you can consider using VTL to achieve the desired backup speed. Of course, you can also consider using the network (OST / CIFS / NFS). We recommend, and it is best, to Setup a dedicated backup network to avoid the production line from being impacted.
Finally it all depends on your backup environment as well as the amount of backup data. CIFS is for the Windows backup server, and can be combined with the AD domain. NFS Share can be deployed on a UNIX platform while OST can be deployed on all the platforms and supports 10G networking.
Is OST used very often in China? Is there anything we should pay attention to if we use OST? Does EMC have a professional team responsible for the implementation of OST?
We are seeing a few Chinese customers now using OST as a backup protocol, and the number is still growing. Now Netbackup, Backup Exec and Networker are supported by the Data Domain (DD) OST backup protocol and Oracle RMAN back up is supported by the Data Domain OST device. With that said, customers don’t need to rely on the backup software for such a backup method.
In addition, the DDBOOST backup method can achieve quite a good performance as the media server can only send out the unique data to data domain, so it can reduce network utilization. EMC has a professional team for implementation if required.
Can you provide more details about RMAN+OST backup?
From a DBA’s perspective, implementing this setup would be easier to manage. The DD BOOST plug-in has to be installed on the Oracle server and the server connects to the Data Domain via a LAN connection. The backup index will be recorded in the RMAN catalog, and we even can copy the data to the DD in the DR site. The RMAN server will record these as replicated data.
Is the data decompressed during the recovery period? Is the amount of transferred data equal to the amount of data which has not been de-duplicated before being backed up?
As DD is architected with the de-duplication technology, all the data is only copied once into DD. It is a tree-structured file system. During the recovery period, the data will be retrieved according to the index as if it has never been de-duplicated.
If we deploy double backup domains, how can the replicated data be visible to the backup program?
If you are using the OST, then all the replicated data will be recorded in the catalog of the backup software, the backup server will record all the indexes. This is also one of the reasons why many customers select the OST.
If you are using Virtual Tape Library (VTL), it requires two domains configured as a VTL environment. VTL data replication is cassette tape-based unit, so a file is taken as a tape. If you want the backup software to identify the copied data, you need to import the tape into the tape library, and then the backup software will recognize it.
If you're using NFS or CIFS, these data on the destination DD is a share, the backup server on the destination is allowed to access the share.
For OST, is there only one backup server? Is there anything special about NW and NBU configuration?
With OST, the media server just needs to load one more plug-in for it to work. It should be a client / server mode of communication. NW has integrated the OST plug-in, without requiring an installation.
Does virtual tape library support only EMC's own backup software? Is there license limitation when increasing the channel of the virtual tape library?
Our VTL can be applied to any backup software. About increasing the channel, there is no license limitation, as far as I know. It is still best to consult the software provider.
Does the VTL come up to tape damage and robotic damage like the PTL?
Usually this issue is a SCSI problem in the communication process, so it is just a logical error. If you configure it properly, this problem can be avoided.
How has DD boost been deployed in China? I heard with DD boost, that the backup throughput could be increased and the network bandwidth could be reduced, so how can this be true? Is there any special care to take during DD boost implementation?
DD boost is a well proven technology, and originated from the OST protocol developed by Symantec. DD Boost replies on the data de-duplication algorithm, and the clients only send out the unique data segments to the Data Domain box, which means the data will be compressed when it arrives. As of the moment, we only support EMC networker, Avamarand Symantec backup software.
If the available capacity is 9T on DD, how can I plan to use it when creating a virtual tape library? Can I create 22 tapes if a single virtual tape is 400GB?
Your algorithm is based on the duplication technology. DD is based on the de-duplication technology, so you can create more virtual tapes. As for the sizing, first you need figure out the capacity to protect the data in the information life cycle, and then divide it by the data compression rate.
Does DD (or in combination with other tools, hardware or software) provide the function to monitor the state of the virtual tapes? Such as the completion of the backup, work state of the physical disks and alarms?
EMC DPA software meets all the requirements mentioned above.
Why do we emphasize that one host fiber channel port should be dedicated to one DD fiber channel port?
This is not emphasized by DD but instead by the manufacturers. Sharing the host HBA with multiple storage arrays is not recommended. If it is shared, the I/O requests might not be processed well in the FC communication and may cause some timeout issue. In that case, once the host doesn’t timely receive the confirmation package, it will trigger a reset, such as a target reset and a LUN reset, which will cause a backup failure and host error, such as device offline.
Author: Kevin Liu
Please click here for for all contents shared by us.