Storage Snapshot vs San backup speed

[Lagcat]

Hello

just a quick question of what speed i should be expecting between these types as i never really want to full back to network backup

at the moment:
i have a netapp fas 8020 16gb FC - currently on 5% disk util - so not doing much
2x cisco mds 9250i 16TB
HP Proliant DL 380 Gen9 with 2x 8gb fibre dual ports ports (2 ports for tape library) ( 2 ports for MDS failvoer)
4x c7000 Chassis using 8gb flex fabric

so i have a minimum of 8gb connection bottle neck which is still alot to have

if add the the netapp to veeam and do storage snapshots i get around 160mb processing rate on the backups

if i do not add the netapp to veeam and do a san storage backup i get around 270mb processing rate on backups

just wondering if anyone else has noticed anything like this? the backup window is not a problem so there is no imediate effect


also the same with a 3par f400 - i see a large difference between storage snapshots and san backup - 90mb -> 230mb - on its local proxy

i have not considered this being a proxy disk issue as the it does process at 270mb when needed

[Gostev]

Processing rate is not the best counter to look at I think as backup method may potentially impact the way it is counted. What about wall clock method, or actual throughput on proxy?

Anyway 90MB/s on 3PAR is an issue for sure, as in our lab 3PAR dishes out over 400 MB/s of actual throughput on v9. NetApp on the other hand is very picky as to its configuration, and the numbers can be all over the place.

[ssimakov]

Jake, there are several factors that may contribute to this. Major being:

• CPG configuration - disks type and RAID configuration
• number of concurrent tasks on proxy and repository
• job content, specifically how many datastores VMs are spread across and VMs sizes/distribution in general

Getting best performance from backup from storage snapshots requires some careful planning and understanding how it works but Anton is right, your numbers are lower than they could be. To give you some reference points - 200 MB/s in direct SAN mode are easily achievable from NL SAS disks with zero optimization effort. 2 GB/s in direct SAN mode are achievable in optimized configuration from SSD CPGs. Where in between you end up will depend on those major contributing factors.

[Lagcat]

Thank you for your reply

the 3par is not really a huge issue as it is being phased out for Netapp

i will give an example of a single job which is giving me 160mb

disk type: SAS with SSD pool
backup type: i backup by datastore to confirm all new vms built are instantly included in the backup job
Netapp: i have 1 datastore per volume/lun
VM disk: thin provisioned

job is a incremental with full synthetic on friday, doing transform into rollbacks

Backup Job 1:
backs up datastore 1 & 2
datastore 1 size: 2tb, provisioned space850gb, free 160tb (very small store)
datstore 2 size: 4tb, provisioned 2.48tb, free 2.85tb (still not a huge store)
number of vms in job: 31

increment takes 27min - processs rate 125mb, bottleneck source 99% - processed 1.4tb, read 129gb. transferred 58.8 - if i can provide more stats i will

at the time of the job there is no load on the netapp - there is also no other veeam jobs running on the proxy at this time - i have no overlaps on my jobs currently - i am mainly just trying to get the best performance before things get busy

so really i should be expecting a much higher transfer rate i guess?

[ssimakov]

Incremental backup speeds tend to be lower than active full and that's true even for hybrid storage configurations. This processing rate of 125 MB/s, is it with storage integration enabled or disabled?

Judging by the description it looks like you use NetApp as block storage for VMFS and not NFS, could you please confirm? With NetApp block LUNs are typically slower than NFS (and the difference will grow bigger in 9.5 with storage integration disabled). What is the configuration of aggregate that backs these LUNs?