HPE Storage Users Group

Does anyone know when we can expect the FC connections on the storage nodes to be capable of 16Gbps? I'm looking into purchasing another 3PAR with a couple of switches. I see that there are 16Gb HP switches, but the backend ports on the 3PAR are currently only capable of 8Gb.

My vendor is encouraging me to purchase the 16Gb switches anyway to have them in place for when the 3PAR is capable, but it just seems strange to me to have my servers connecting in at 16Gb only to hit the 8Gb controllers on the 3PAR.

Also, I'm sure my whiny customer isn't going to want me to try to dip into the well twice to upgrade the 3PAR at the time. If HP is going to 16Gb soon, maybe we'll just wait.

The array ports are always the last to move to the newer speeds, but there's also the question of why you'd need that amount of bandwidth?

The first devices to get the new speeds are normally switches, then it's the hbas and lastly it's the storage.

You certainly don't need 16gb speed on the server end, but the 16gb technology does have it's advantages, not just speed. You also don't need 16gb ports on your switches, it would be more cost effective to buy 8gb sfps for the hbas and the switches.

We'll probably be on 32gb switches by the time storage arrays have 16gb natively. The only reason you aren't being quoted an 8gb switch is that you can't buy them any longer I don't think.

Don't forget, the 3par can have 6x 8gb ports per controller, and there's 4 controllers, that's a lot of ports and a lot of bandwidth!

This is quite a good doc explaining the changes in 16gb, including the new 64/66 encoding.

Which actually means the throughput doesn't double from 8-16 but it's more reliable for distances.

http://www.brocade.com/downloads/docume ... 313-00.pdf

I second all that. 16Gbps is a hell of a lot of bandwidth.

I'm now having flashbacks to awhile back when brocade tried to rename 16Gbps FC to Gen6 FC or whatever, and Brocade and Cisco went at it like a bunch of angry teenagers.

At the end of the day, I'm not sure anybody in the SMB space that these arrays are targeted at is driving enough bandwidth to saturate a 16Gb link.

We were on 4Gb until we put in our 7200's last year, and there was no noticeable difference when we upgraded to 8.

I would say put it in with 8 and walk away. No need to wait for something you don't need.

Thank you all for your useful replies.

Considering many people just moved to converged 10gb where storage and network share a 10gb link I agree 16gb HBA is a hell of a lot of bandwidth and a few of them open throttle would begin to stress the back end at some point.

16 gb ports may seem "excessive" to some, however consider being able to run less cables per storage frame. Having just run 24 cables from my 7400 to my 8gb SAN, and consuming 24 switch ports... I relish the possibility of cutting that in half.

_________________
Richard Siemers
The views and opinions expressed are my own and do not necessarily reflect those of my employer.

I still wouldn't since reducing the number of ports reduces the number of internal queue objects and the number of buffer to buffer credits available within the switches. Of course if performance isn't important you could easily do that to reduce costs.

Queue depths and buffer credits etc all scale up with the port speed or else 16gb ports would not be able to run at 16gb, and performance aficionados would still be running massive bundles of 1GB/2GB FC links... I think what you meant is to watch out for limits like 128 initiators per port?

_________________
Richard Siemers
The views and opinions expressed are my own and do not necessarily reflect those of my employer.

I would say that the big issue here is understanding your workload and your environment.

There will be some that absolutely will benefit from 16gbps.

That being said, most, in particular those targeted by the 7200/7400, will probably not