Raid 5 write performance in proliant range

ginjaninja · 9 Aug 2019 at 11:24

thanks, i wasnt aware of that as an option
any ideas how likely it is to work with a cheap and cheerful esata caddy? such as ...

https://www.startech.com/HDD/Docking/usb-3-1-gen-2-esata-dock~SDOCKU313E

Armageus · 9 Aug 2019 at 11:28

I haven't personally tested it (Caddy or even the HBA), but see no real reason why it shouldn't.

Someone else may know more

Deleted member 138126 · 10 Aug 2019 at 18:17

Oof Veeam Reverse Incremental is *brutal* on the IO of whatever you're storing the backups on.

ginjaninja · 11 Aug 2019 at 17:07

rotor said:
Oof Veeam Reverse Incremental is *brutal* on the IO of whatever you're storing the backups on.

out of interest, for learning purposes, what is your requirement for reverse incremental?

ginjaninja · 11 Aug 2019 at 21:09

by switching my reverse incrementals to forward incrementals, and having a separate full backup job directed to external media (batch scheduled via windows task scheduler every 28 days), i am savings loads of time / load on the backup server.
creation of the single point on the external media is <6hrs vs 20 hours (to process the reverse incremental and copy the file). Now i dont need to worry about raid/controller spec, the limiting factor is the sata disk which 'is what it is'.

but have now gone full circle and am testing a backup copy job with an x day interval (backup copy jobs seem to support an x day interval) . Its is perhaps a moot point that backup copy jobs insist on maintaining a minimum of 2 restore points (i only want 1 point ie a full backup)....moot because the backup copy job is pointed directly at a repository on an external path that is wiped - as removable drives are formatted before use, so i always get just a full backup.

ginjaninja · 14 Aug 2019 at 14:13

Armageus said:
I haven't personally tested it (Caddy or even the HBA), but see no real reason why it shouldn't.

Someone else may know more

i am going to give it a go and report back.

I am new to SFF/SAS, will this cable go into this controller please?

https://www.reichelt.com/gb/en/mini...fkKVQS4NG3siitKrxLJRZgIUxB8aAi7HEALw_wcB&&r=1

https://h20195.www2.hpe.com/v2/GetDocument.aspx?docname=a00008183enw

Armageus · 14 Aug 2019 at 15:13

Yep - looks fine (Controller is Mini SAS HD, SFF-8644 cable is Mini SAS HD according to wiki)

ginjaninja · 14 Aug 2019 at 21:52

Armageus said:
Yep - looks fine (Controller is Mini SAS HD, SFF-8644 cable is Mini SAS HD according to wiki)

thank you,
Do you have any experience with making the most our REFS on Windows 2016/2019?
I am keen to minimise storage space without massively affecting backup times.
I am assuming that REFS is the goto tech / design for Veeam now?
but i read there are some challenges and the software is still 'maturing'

https://forums.veeam.com/veeam-backup-replication-f2/refs-3-0-and-dedup-t56658.html

khrall · 15 Aug 2019 at 00:57

ginjaninja said:
thank you,
Do you have any experience with making the most our REFS on Windows 2016/2019?
I am keen to minimise storage space without massively affecting backup times.
I am assuming that REFS is the goto tech / design for Veeam now?
but i read there are some challenges and the software is still 'maturing'

https://forums.veeam.com/veeam-backup-replication-f2/refs-3-0-and-dedup-t56658.html

We're successfully using REFS on 2016 for our IaaS and VCC customers. We've not experienced any issues.

bledd · 15 Aug 2019 at 17:30

I'm fresh off a Veeam course, ReFS is best for Veeam Repositories

ginjaninja · 15 Aug 2019 at 18:24

bledd said:
I'm fresh off a Veeam course, ReFS is best for Veeam Repositories

so refs it is.

so for 30 point forever forward incremental daily backup chain of ~2Tb of 30 windows servers.
on a raid 10 array with bbwc windows 2019 16 core 16gb dedicate 100% veean backup server .
what would peoples thoughts on the following be.

1. refs disk cluster size? 64kb?
2. any refs settings beyond cluster size?
3. veeam compression level
4. veeam storage setting ? local disk largest blocks?
5. corruption guard
6. maintenance tasks. create full restore point regularly?
7. dedupe settings/considerations.

i am major noob in terms of understanding the significance/value of these.
my goal is to minimise disk capacity use without significantly hindering backup times, whilst ensuring the integrity of backup points.

thanks

khrall · 15 Aug 2019 at 22:50

ginjaninja said:
so refs it is.

so for 30 point forever forward incremental daily backup chain of ~2Tb of 30 windows servers.
on a raid 10 array with bbwc windows 2019 16 core 16gb dedicate 100% veean backup server .
what would peoples thoughts on the following be.

1. refs disk cluster size? 64kb?
2. any refs settings beyond cluster size?
3. veeam compression level
4. veeam storage setting ? local disk largest blocks?
5. corruption guard
6. maintenance tasks. create full restore point regularly?
7. dedupe settings/considerations.

i am major noob in terms of understanding the significance/value of these.
my goal is to minimise disk capacity use without significantly hindering backup times, whilst ensuring the integrity of backup points.

thanks

1.) 64k
2.) Not that I'm aware of
3.) Optimal
4.) Local disk (not largest blocks) Largest blocks will affect dedupe ratio, only designed for really large backup files
5.) Yes
6.) Make sure it's synthetic full, as this will leverage fast clone feature of ReFS & also save space.

ginjaninja · 16 Aug 2019 at 21:48

thank you, server and cable have been ordered, will update once the results are in.

blueboy2001 · 22 Aug 2019 at 07:28

Always make sure you have plenty RAM for ReFS. I think Veeam reccomend 1GB per TB of storage up to a certain point - that's over and above any other memory usage on the box

ginjaninja · 23 Aug 2019 at 20:17

blueboy2001 said:
Always make sure you have plenty RAM for ReFS. I think Veeam reccomend 1GB per TB of storage up to a certain point - that's over and above any other memory usage on the box

more luck than judgement i have enough. 1GB per TB would get quite pricey for larger backups...my disk will only be 8TB and actual backup chain size closer to 2TB. 16GB RAM should be enough for 2019+VEEAM+REFS.

Customs made me pay £15 for the privilege of importing a 2M SAS to ESATA cable from the states..hope it works

ginjaninja · 19 Nov 2019 at 04:38

Armageus said:
External SAS HBA (e.g. HP SC08GE) and SAS SFF8088 to eSATA breakout cable?

thanks for advice sas hba plus cable supports both our esata caddies.

ginjaninja · 19 Nov 2019 at 04:42

my veeam performance on refs is a bit lack lustre ....
any suggestions on %read/write cache on p408 with raid 10?

ginjaninja · 27 Nov 2019 at 11:43

Spoke too soon, REFS on RAID 10 is working well. Synthetic Full backups on forward incremental are using no extra space, and taking a 5 minutes to calculate (as opposed to reverse incremental taking closer to 10 hours) and Veeam is hitting 1GB/s . to the array.

Vince · 27 Nov 2019 at 11:49

Didn't read the whole thread but my backup is far bigger each night and I use a HP StoreOnce 4500 (dedicated backup and de-dupe) which is a DL380 chassis with HP custom firmware. I can process some 20+ tb per night and still fit within a few hour window. The 4500 imo is a solid dedicated appliance

From Veeam last night a full backup on a 2.2TB Exchange server took a total of 24 mins.

Semple · 27 Nov 2019 at 11:52

Vince said:
Didn't read the whole thread but my backup is far bigger each night and I use a HP 4500 (dedicated backup and de-dupe) which is a DL380 chassis with HP custom firmware. I can process some 20+ tb per night and still fit within a few hour window. The 4500 imo is a solid dedicated appliance From Veeam last night a full backup on a 2.2TB Exchange server took a total of 24 mins.

StoreOnce 4500?

N.B I work on the StoreOnce R&D team.

Raid 5 write performance in proliant range

Deleted member 138126

Deleted member 138126