Sizing guidelines for D-NAS
The sizing for D-NAS environments is based on your business requirements. Sizing depends on the storage array. It also depends on the characteristics of the NAS data that you protect. You can configure the D-NAS policies to use the NetBackup media server as a backup host. As a result, D-NAS backup processes may organically scale in terms of both performance and throughput.
When a D-NAS policy runs, NetBackup uses the nbcs (crawler) and the bpbkar processes on the backup host. It uses the bpbrm and bptm on the media server. Each running process consumes CPU cycles. The nbcs process uses CPU the most. The nbcs is a multi-threaded crawler. It traverses the NAS share during backup and index operations. Multiple nbcs processes handle concurrent jobs for backup and indexing operations of NAS volumes. One nbcs process corresponds to one NAS volume backup. If you use multiple backup hosts to back up a single NAS volume, each backup host uses a separate nbcs process.
The crawler process is multi-threaded. Multiple threads traverse one NAS share during D-NAS policy execution. This can lead to spikes in CPU use for the nbcs process. You can ease CPU use by decreasing the number of threads used by the nbcs process. You can set the MULTI_THREADED_CRAWLER_THREADS parameter in the bp.conf file. This changes the thread count used by each nbcs process. The default value is 20, and you can specify a value in the range of 1 to 200. You must set this parameter on the backup hosts used for NAS backups. This is applicable for the hosts on NetBackup version 10.4 and above.
The amount of memory used by a backup job of a single NAS share depends on the number of streams configured in the D-NAS policy. Each backup stream on a host uses one bpbrm, bptm, and bpbkar process. For example, if the policy is set to use 10 streams, then a single NAS share backup runs 10 instances of the bpbrm, bptm, and bpbkar processes. It runs only one nbcs process.
The amount of memory used by the bpbrm, bptm, and bpbkar processes is static and does not fluctuate much. The memory used by the nbcs process depends on certain data traits. These include file system hierarchy, number of files, and folders in a NAS share. If a NAS share has a very dense directory structure, then the nbcs process uses 200 MB of memory at its peak. If a NAS share has a flat hierarchy with millions of files in its directories, then nbcs uses 20% - 30% additional memory at its peak. Note that memory usage is not always at peak and decreases as the backup progresses. You may observe spikes in memory consumption depending on the data characteristics.
Memory usage is the same for all backup types. These include index tasks, first full, incremental, and accelerated full backups.
Table: Memory consumption for a single NAS backup with 5, 10, and 20 backup streams
NAS share memory consumption | Memory for 5 backup streams (MB) | Memory for 10 backup streams (MB) | Memory for 20 backup streams (MB) |
|---|---|---|---|
On backup host | 115 + 200 (crawler) | 230 + 200 (crawler) | 460 + 200 (crawler) |
On media server * | 780 | 1560 | 3120 |
Total memory consumption (approximate) | 1095 | 1990 | 3780 |
* For index from snapshot operations, only the media server memory consumption is applicable.
You can estimate the approximate memory needed for D-NAS backups using the table: Memory consumption for a single NAS backup with 5, 10, and 20 backup streams . Using this table, you can also estimate the number of backup hosts and media server hosts that you need to provision for D-NAS backups. This also helps to schedule the NAS backup jobs. For example:
During backup, a policy with 10 streams and 10 NAS shares uses about 20 GB of system memory.
During backup, a policy with 20 streams and 10 NAS shares uses about 38 GB of system memory.
Consider the following:
When a single host serves as both the media server and the backup, all memory usage occurs on that host.
When you use the multi-host feature for NAS share backups, the memory utilization of the crawler is distributed equally among all backup hosts.
Use multiple media servers if the overall number of backup streams for D-NAS backups exceeds 200.