At what speed is your File Server connected to the Network?

[Frazzle]

If you have a physical File Server at what speed is it connected to the network?

[synaesthesia]

10Gbit on ours mostly for futureproofing but don't really see much more need for 1Gbit in our usage case. Beforehand we had them on 2Gbit via link aggregation.

[crc-ict]

Teamed pair of 1GBit links for us

[PotNoodleTech]

2 x 10Gbit aggregated per VM Host and 1 x 10GB for Veeam Backup server.

[mjk]

10Gbit on ours mostly for futureproofing but don't really see much more need for 1Gbit in our usage case. Beforehand we had them on 2Gbit via link aggregation.

 

If I'm reading that right you are saying that you moved from an aggregated 2Gb/s that has a failover to a single 10Gb/s - thus reducing the reliability by 50%. Why not add in a 1Gb/s link back into the aggregation so that you can failover incase the module or cable goes bad ?

[PotNoodleTech]

I'd take single 10GB over double 1Gb any day of the week. I wouldn't want 10x less bandwidth (yes I am saying 10x less as link aggregation does not work in the way most people think it does...) for a "but what if" scenario which has never happened. Has anyone else actually had a port on your core switch (or a patch lead in a locked cab) fail and take a server out?

 

Not saying it couldn't happen, but not worth compromising your network bandwidth for a once every 10-20 year risk.

[Oaktech]

3 out of 4 ports on each host server teamed for general usage.

[mjk]

I'd take single 10GB over double 1Gb any day of the week. I wouldn't want 10x less bandwidth (yes I am saying 10x less as link aggregation does not work in the way most people think it does...) for a "but what if" scenario which has never happened. Has anyone else actually had a port on your core switch (or a patch lead in a locked cab) fail and take a server out?

 

Not saying it couldn't happen, but not worth compromising your network bandwidth for a once every 10-20 year risk.

 

Ports definitely go down on switches and cables fail - it's why we have failovers.

So why not have 10GB/s AND a 1gb/s failover if you can't afford the 2x 10GB/s ?

It seems ludicrous to me that you'd introduce a single point of failure in a critical system for no reason if the old fibre is just laying dark.

[mikkydoos]

Teamed pair of 1GBit links for us

 

Same here. No probs.

[synaesthesia]

If I'm reading that right you are saying that you moved from an aggregated 2Gb/s that has a failover to a single 10Gb/s - thus reducing the reliability by 50%. Why not add in a 1Gb/s link back into the aggregation so that you can failover incase the module or cable goes bad ?

 

Actually we do have in place already, just didn't think to mention it Just set up via vsphere so it was pretty much set and forget (added the 10Gb/s port to the team)

And indeed ports fail; we've not had any go on our cores but certainly a fair few on the lesser switches around the place and it's not something I'd take for granted.

Edited June 12, 2018 by synaesthesia

[AlanD]

Getting more than 1Gb/s out of a server with 10K or even 15K disks in a RAID can be a challenge...well unless you have some very serious kit...and/or have M2 type SSD storage. A single disk rarely delivers more than 200Mb/s -and less under "typical" load conditions when latency is always working against you.

 

Although we have a 10GB backbone, we only connect each of our servers as aggregated 2x1GB links -to separate switches in the stack (because the cost of a 10GB server card wouldn't give any increase in throughput).

 

 

So by all means use a 10GB connection - but at least the aggregated solution means we still have connectivity elsewhere in the school should a core switch (or server port) fail. In our case the aggregated solution cost nothing....

[Katy]

Has anyone else actually had a port on your core switch (or a patch lead in a locked cab) fail and take a server out?

Yes but it took two buildings offline rather than a server. Infants school was cat5 linked to nursery which was fibre back to main building. Fibre patch lead in the nursery cab broke one day whilst locked in a cab - coincidentally around the same time they were doing drain works to a new housing development and they broke the duct containing the fibre - so we assumed they'd broke the fibre and was about to go mental at them when turned out it was the patch lead. Only happened once in 15 years though.

 

My hosts have a 10Gb link plus two 1Gb teamed as backup/standby/failover (whatever the vmware terminology is), file server is a VM so it sits on that.

 

Back when I had a physical file server it had 2 x 1Gb (and we had no 10Gb anywhere)

[MartinT]

2 x 1Gbps teamed. It's plenty fast enough.

[pritchardavid]

Currently 2x10GBe Aggregated per Physical Server, have two servers in a box using sas drives. (Capable of having 70 drives)

 

Will be moving this to a Clustered two VM setup, to whatever needs to stay on site, will be moving everything possible to OneDrive.

 

[markwilfan]

@AlanD my raid 10 with spinny disks easily saturates a 10Gb link. A single spinny disk can saturate 1gb link. You only need circa 110MB/s to do 1Gb/s