Dell PowerVault DR6000 - Manual (Page 19)

Streams vs. Connections
This topic describes the differences between data streams and application connections.
Streams can be likened to the number of files written at the same time to a DR Series system. The DR Series system
tracks the number of files being written and assembles the data into 4MB chunks before processing that section of the
data. If the stream count is exceeded, the data is processed out of order and overall deduplication savings can be
affected. For details on maximum stream count, see the
Dell DR Series System Interoperability Guide .
Connections are created by applications; within a single connection, there can be multiple streams depending on the
application and how many backup jobs are running in parallel over that single connection. Replication can use up to 16
streams over a single port using one connection.
For example, suppose you are running backups using Backup Exec and using DR4100 and the CIFS protocol. If you have:
• One Backup Exec server connected to the DR4100 over CIFS and one backup running, you have one connection and
one stream.
• One Backup Exec server connected to the DR4100 over CIFS with 10 concurrent backups running, you have one
connection and ten streams. This means that Backup Exec is writing ten different files to the DR4100.
Replication
Replication is the process by which the same key data is saved from multiple storage locations, with the goal being to
maintain consistency between redundant resources in data storage environments. Data replication improves the level of
fault-tolerance, which improves the reliability of maintaining saved data and permits accessibility to the same stored
data. The DR Series system uses an active form of replication that lets you configure a primary-backup scheme. During
replication, the system processes data storage requests from a specified source to a specified replica target, which
acts as a replica of the original source data. This replica can then be cascaded optionally to a third location called a
Cascaded replica for an additional copy.
NOTE: The DR Series system software includes version checking that limits replication only between other DR
Series systems that run the same system software release version. If versions are incompatible, the administrator
will be notified by an event.
Replicas/Cascaded replicas are read-only and are updated with new or unique data during scheduled or manual
replications. The DR Series system can be considered to act as a form of a storage replication process in which the
backup and deduplication data is replicated in real-time or via a scheduled window in a network environment. In a
replication relationship between two or three DR Series systems, this means that a relationship exists between a
number of systems. One system acts as the source and the other as a replica, with an optional third cascaded replica if
you have chosen to keep two instances of replicated data in your backup workflow.
Replication is done at the container level and is one directional from SCR to Replica to Optional Cascaded Replica;
however, since replication is done at the container level you can set up various containers to meet your specific
replication requirements for your specific workflow. This form of replication is supported for the CIFS, NFS, Rapid CIFS,
and Rapid NFS protocols and is fully handled by the DR Series system.
Unlike NFS, CIFS, Rapid NFS or Rapid CIFS containers, RDA with OST, RDA with NetVault Backup, and RDA with vRanger
container replication is handled by Data Management Applications (DMAs) media servers.
The DR Series system supports the 64:1 replication of data (32:1 if on DR4X00 and 8:1 on DR2000v), whereby up to 64
source DR Series systems can write data to different individual containers on a single, target DR Series system. This
supports the use case where branch or regional offices can each write their own data to a separate, distinct container
on a main corporate DR Series system.
NOTE: Be aware that the storage capacity of the target DR Series system is directly affected by the number of
source systems writing to its containers, and by the amount being written by each of the source systems.
19