Accelerator for SAN client

You can use accelerator with SAN client, However I question your use of SAN client where accelerator would benefit you.

Juan, yes you can as they told you.

One thing to consider is that you can NOT use source dedup with SAN Client, then you need to disable source dedup if you like to use accelerator with a SAN client.

I agree with Andrew.

Clientside dedupe is suited to FS backups with many files , accelerator improves on that by reducing the amount of files to scan. So effectively you have a small list of which you might only send a few bytes (unique new segments). This is what gives you the extreme logical performance figures you see.

By moving to SAN Client you change to target dedupe, effectively removing the performance benefit. It mght be faster transport but will it be faster overall?

Anybody have scenario to share where it's beneficial?

Why do you think that you're removing the perfomance benefit? it isn´t.

Accelerator reduces the time that you spend scanning the FS, but also allows you to send only files that have changed at the same time that you're doing full backups daily. 

Imagine that you have a FS with several files, and you're using accelerator and source dedup, you're reducing the amount of files that are going to be scanned, and also, not all the blocks that have changed are going to be send to your Media Server or Appliance becasue you have source dedup.

In the case of target dedup, you're reducing the scanned files, and you're only sending the blocks that have changed of these files to your Media Server or Appliance, and there the dedup will be done. But in this case you're sending the data through SAN.

Accelerator has drastical benefits on performance with or without source dedupe.

Hi, you said "In the case of target dedup, you're reducing the scanned files, and you're only sending the blocks that have changed of these files to your Media Server or Appliance, and there the dedup will be done. But in this case you're sending the data through SAN."

That's not correct, target dedupe sends all blocks to the media server (as indentified by accelerator) and the media server performs the deduplication.

Or am i missing something?

Riaan, you are correct. Accelerator by its self has no idea what has been sent and what the MSDP has in the way of fingerprint. So if there is a change it sends the change.

I De Pedro you also are correct in that the data sent to the MSDP would move faster over the SAN once it gets there however in a file system backup that takes more time than going over the LAN.

However the additional cost of infrastructure required of a SAN client negates the value that the SAN client brings to a file system backup with the presence of Accelerator and the deduplicating client. If you are spending $1000 a port on the client and $1000 a port (cabling included) on a switch the speed of backup over the SAN does not make fiscal sense. The use of a SAN client only makes sense if you are backing up data that can be streamed like database backups or large files such as video or database dump files.

In a regular file system backup the throughput on a SAN client just does not warrant the extra cost. So while accelerator works with a SAN client there is no  benefit to using a SAN client in a situation where backups would benefit from the use of accelerator. Remove the SAN client from the equation and the you can add in client side deduplication and blow the doors off of SAN client. Provided your LAN infrastructure is up to it.

Riian, Andrew,

let me explain one thing. First, Andrew, you're right in terms of cost of infraestructure, but in some customers all these infraestructure is now present, then they don´t need to pay anything more. They're re-utilization some infraestructure.

Effectively, accelerator doesn´t know anything about fingerprints, but not moves all the data, then it doensn´t send all the data to the Media Server, only the changed blocks, check the admin guide and you can found this:

How accelerator works

■ If the client has no previous backup, NetBackup performs a full backup and

creates a track log. The track log contains information about the client's data,

for comparison at the next backup.

■ At the next backup, NetBackup identifies data that has changed since the

previous backup. To do so, it compares information from the track log against

information from the file system for each file. For NTFS and ReFS file systems,

it also uses the Windows change journal to help identify the data that has

changed since the last backup.

Accelerator uses the Windows change journal in two ways: To check for changes

in the file system metadata, and to help detect which files have changed since

the last backup.

See “Accelerator and the Windows change journal” on page 694.

■ The NetBackup client sends to the media server a tar backup stream that consists

of the following: The client's changed blocks, and the previous backup ID and

data extents (block offset and size) of the unchanged blocks.

■ The media server receives the client's changed blocks and the backup ID and

data extents of the unchanged blocks. From the backup ID and file system

descriptors, the media server locates the rest of the client's data in existing

backups.

■ The media server directs the storage server to write the changed blocks and

the unchanged blocks in a new full image.

The difference with the dedup, is that if it's source you can dedupe the changed blocks for coincident blocks, and with target, you check only for storage. But you're sending a lot of less data to the Media Server (depending on your change ratio).

Please clarify because the quoted text (which I've read before) is a bit unclear.

"The NetBackup client sends to the media server a tar backup stream that consists of the following: The client's changed blocks, and the previous backup ID and data extents (block offset and size) of the unchanged blocks."

Is the text referring to bpbkar (the client) in the text above?

Is the text referring to the media server (bptm) or the deduplication plugin?

If Accelerator is providing only the changed blocks are we talking about segments here which is the same as clientside deduplication?

Where does accelerator fit into the following sequence?

Client Side Dedupe (SOURCE)

• The Backup/Archive Manager (bpbkar) generates the backup images and moves them to the client nbostpxy process by shared memory.

  The Backup/Archive Manager also sends the information about files within the image to the Backup/Restore Manager (bpbrm). The Backup/Restore Manager sends the file information to the bpdbm process on the master server for the NetBackup database.

• The client nbostpxy process moves the data to the deduplication plug-in.

• The deduplication plug-in separates the files in the backup image into segments.

• The deduplication plug-in buffers the segments and then sends batches of them to the Deduplication Multi-Threaded Agent. Multiple threads and shared memory are used for the data transfer.

• The NetBackup Deduplication Multi-Threaded Agent processes the data segments in parallel using multiple threads to improve throughput performance. The agent then sends only the unique data segments to the NetBackupDeduplication Engine.

• The NetBackup Deduplication Engine writes the data to the Media Server Deduplication Pool.

Media Server Deduplication (TARGET)

• The Backup/Archive Manager (bpbkar) on the client generates the backup images and moves them to the media server bptm process.

  The Backup/Archive Manager also sends the information about files within the image to the Backup/Restore Manager (bpbrm). The Backup/Restore Manager sends the file information to the bpdbm process on the master server for the NetBackup database.

• The bptm process moves the data to the deduplication plug-in.

• The deduplication plug-in retrieves a list of IDs of the container files from the NetBackup Deduplication Engine. Those container files contain the fingerprints from the last full backup for the client. The list is used as a cache so the plug-in does not have to request each fingerprint from the engine.

• The deduplication plug-in separates the files in the backup image into segments.

• The deduplication plug-in buffers the segments and then sends batches of them to the Deduplication Multi-Threaded Agent. Multiple threads and shared memory are used for the data transfer.

• The NetBackup Deduplication Multi-Threaded Agent processes the data segments in parallel using multiple threads to improve throughput performance. The agent then sends only the unique data segments to the NetBackup Deduplication Engine.

  If the host is a load balancing server, the Deduplication Engine is on a different host, the storage server.

• The NetBackup Deduplication Engine writes the data to the Media Server Deduplication Pool.

Riian,

Really this is for create another thread, but I will try to answer your questions.

Is the text referring to bpbkar (the client) in the text above?

Is the text referring to the media server (bptm) or the deduplication plugin?

If Accelerator is providing only the changed blocks are we talking about segments here which is the same as clientside deduplication?

In the same manual where I copied the text you can find the attached files, I think that these images explains and answer your questions, also I recommend this topic from connect:

https://www-secure.symantec.com/connect/blogs/frequently-asked-questions-netbackup-accelerator

And the second part:

https://www-secure.symantec.com/connect/blogs/frequently-asked-questions-netbackup-accelerator-part-ii

Accelerator and Dedupe are different features, with similar behaviour but no exactly the same, that can co-exist.