I just bought a second Drobo and a Drobo Share.
Today I bought some new Cat6 cables and connected my Drobo Share (with the two Drobos) and my two Vista 64 PCs to my 5 port 1000 Mbit switch (Netgear GS105).
Both PCs get the expected higher bandwith with the new cables but the Drobo share seems to be limited to a speed of around 10000-11000 KB/s.
All Drobo Hardware has the latest fimware. 1.35 drobos /1.12 share
Is that normal ?
I expected more speed and I think the Drobo share has a 1000 Mbit LAN connection.
Yes, I agree with this, consistent with my results. Very slow in my opinion, especially as it’s equipped with Gigabit ethernet! I too have a gigabit network and feel very disappointed with DroboShare performance. To the extent that I refuse to use the DroboShare I purchased. The Drobo connected direct by Firewire can serve data/files much faster than DroboShare. Why should this be the case? I wonder if there’s any chance the DroboShare firmware could get an update to improve this situation?
I only got DroboShare so I could run DroboApps on it and retire some of the full systems I have doing daemon work.
For my use (not trying to run video off the Drobo), the slower sharing performance is tolerable, though I’d love for it to be faster.
per @jennifer - DroboShare is 10MB/sec = 80 Mbps, or 80% of a 100Mbps Ethernet. Pretty good. Rule of thumb: protocol overhad takes 20-25% of Ethernet’s performance (USB and Firewire, too).
Maybe Droboshare has more head room and a faster link, i.e. Gigabit Ethernet, might enable it. But certainly not a huge amount. Before I gave my Drobo/Droboshare to a nephew, I regularly saw reads in the 8-12 MB/sec range. Large files, .iso’s etc, gave the best performace.
Same experience and very disappointed, Drobo is not the cheap option, especially as the DroboShare is even more cash to shell out, upgraded my home network to gigabit in preparation of a speedy, sexy, NAS… but no, barely 10-12MBps. My switch confirms it’s a 1000Mbps connection (so was expecting a good ~40MBps - seeing as the limiting factor would be the USB connection between the Drobo and the DroboShare). Saying that - I’m assuming the Drobo-to-DroboShare connection is USB2? It’s not something stupid like USB1 is it? How can it be that slow!!! Feel a little cheated for the money.
Has someone been using the Jumbo Frame droboapp? does it make a different in the transfer speed?
I was thinking of upgrading my network to gigabit ethernet but from this thread I have the impression if would not make any difference for the droboshare transfert speed.
Your numbers are pretty close for 75% of max, which is a good measure of practical real-world performance due to overhead and other issues. Just your caps are a little funky there Dave…
Given a 75% USB 2.0 transfer, you get 45 MB/sec (360 Mbps)
But I think the bottleneck is really the DroboShare itself. Either its USB bus isn’t up to par, its CPU isn’t fast enough, or its network interface has a bottleneck.
We’d need to see a block diagram to really know though.
Reminds me a lot of early motherboards with Gigabit that had the GigE interface connected to the 32-bit PCI bus which created a bottleneck that limited the Gigabit interface to less than Gigabit speed.
Yeah I can never remember which way to have the caps but you got the jist so that’s ok.
I tend to agree, I think it’s the drooshare too, the other possibilities were just for completeness. I’ve seen samba configs that are really bad that have caused problems but by the looks of theirs it’s so standard that I doubt that there could be an issue there.
I’m not sure what the throughput figures are for a drobo connected directly with USB without the droboshare but I’m sure that they are not pushing the USB envelope so maybe it’s an amalgamation of the USB bus and the CPU not being able to drive the Gbit card.
I raised this with drobo about 6 months ago and all I got was
Subject: Re: Case 23366 - DroboShare Problem
“As I mentioned before, 6.5MB/s is about right for a DroboShare. This is partially due to the fact the Drobo connects to the DroboShare by USB. The only way that I know to increase performance would be to use the DroboApp Jumbo Frames.”
When I asked for escalation I got
“Yes I do have access to those who know the unit at that level, but unfortunately we are not allowed to disclose how things work. Like I said 6.5MB/s is about right for a Drobo on a DroboShare.”
Proprietary tech bites us about as often as it loves us…
A block diagram would be nice, but the only tech I see that has block diagrams these days are motherboards, and even that’s becoming a rarity.
Works for low-bandwidth needs - office-type files, MP3s, etc. But if you’re doing large transfers, especially video or large data files, users may find it lacking.
I suspect the 10 MB/sec max is a total, so a DroboShare connected two two Drobos would still deliver a max of 10 MB/sec out its network port.
If it ends up delivering 20 MB/sec out, then the two USB ports are on separate buses and the bottleneck is in the USB interfacing.
I’d give this a try but I don’t have two Drobos.
I realized I’m going to have to set up small machine to handle some of my other network needs as I retire my huge RAID server and automation server, so I might end up replacing DroboShare with a netbook/nettop type unit with Gigabit. It’ll mainly depend on the USB load on that machine.
I am disappointed too. I thought I upgraded my wireless router to Giga (Apple wireless Extreme) expecting to make lots of difference but nope it doesnt. I installed the Jumbo Frames same result too. Transferring 300 gigs of file took me nearly 12 hours and 800mb of files around 5-6minutes.
I get 17-22 Mbytes per second on my locally attached USB Drobo (admittedly, that is a best case, transferring video files, multi-gig’s each.)
You will get lower rates for smaller files. 10 to 12 Mbytes is suspiciouly close to 100 mbits/second.
If your hub is manageable, then you should check to make sure that the link is really 1000 mbit/s and was not auto-negotiated downward. If it is at all manageable, try to get it to NOT negotiate, and insist on 1000 mb/s full-duplex. If you’re a linux nerd, you can check on the ds side too.
It is really strange, but after twenty years, the auto-negotiation is still rather flaky on ethernet, and even expensive hardware gets it wrong sometimes.