Well, I guessed that if you only heard it from me, you would rightfully put it to anecdotal evidence or hearsay. So I added links to the finding of Fred Langa (a person with much more credibility than I have) and a link to a technet post, assuming those were credible enough.-Shades
On the contrary - if a thing like this had been reported by you (or some other DC member), it would be something interesting to look into. When a post with that level of lack of quantifiable comes from some relatively well-known source, I get suspicious ("You won't BELIEVE how much homegroups suck" probably gives ad revenue), and I also hold "well-known folks" to higher standards than normal individuals.
Langa's article is extremely weak in quantifiable data, has imprecise language and terms, and doesn't even try to present a likely explanation. IIRC the TechNet post just referred back to the Langa article, so it's not a reliable data point in and of itself. And even the MVPs there post questionable stuff from time to time, anyway :-)
The machines that do not use NetBios are much faster retrieving content from network accessible folders than the other ones.-Shades
Can you quantify "retrieving content"? Not necessarily a full detailed breakdown with graphs and stuff, but a "first connection to remote machine is slow" versus "file listing is slow" versus "transfer speed of one 10gig file is 6MB/s vs 10MB/s on <other tech>" are
extremely different scenarios.
About the other part:
More often than not, 'Less is more'. Software and services/protocols that are not needed, you best get rid of to prevent Windows taking unnecessary actions. By 'getting rid of' I mean disable, not remove as there might be a future task for which you might need it again.-Shades
I agree with the "less is more" philosophy in general, but I don't really feel it applies to Homegroup. Caking layers upon layers on a protocol is bad, simplifying authentication isn't
necessarily bad.
From your "workgroup vs homegroup" list, the only thing that should have any impact on
throughput would be IPv4 vs IPv6. But if that's a 10% difference, something is very, very wrong on your LAN - the TCP headers are larger, but not that
Personally, I would not be surprised if the Homegroup requirement of IPv6 could be the cause for slowdowns in an IPv4-only network. The NIC in the computer could wait a millisecond or so, because of the incomplete/wrong IPv6 configuration before reverting back to the working IPv4 configuration, each time a TCP/IP packet needs to be ACKnowledged. This adds up when transferring (big) files. I would also have no problems imagining that the 'spanning over a subnet' introduces extra overhead in some network drivers.-Shades
Nope. You might have a delay at a broadcast "is there anybody OUT there?" level at IPv6, but it's not per-packet. Once you start communicating between hosts, you're on one protocol level. And if IPv6 is a
requirement for Homegroups, you wouldn't have any connectivity on IPv4 anyway. "Spanning over a subnet" would only be relevant if you're actually doing that.
As most people only have access to IPv4-only networks, which I don't see changing any time soon, the unnecessary Homegroup slowdown will remain a problem.-Shades
Most people won't have IPv6 internet connectivity, but we're talking LAN connectivity here. If Homegroup is IPv6-only, you wouldn't see a a slowdown on IPv4 networks.
Now,
if there really are slowdowns related to Homegroup, I'd like to know about it - and especially
why. The stuff I've heard so far seems about as reliable as homeopathy, though.
Recent Posts
