Download ramp up speed
-
The modem isn't showing any uncorrectables and all of the signal and dB ratings are in spec.
The testing I've been doing is either on the pfsense VM or on a switch attached physical machine, but no extra hops.
NICs are vmxnet3 and wan/lan are physically different NIC. Broadcom brand.
Same result before and after openvm-tools
-
Out of the box settings should be fine. My TWC 300/20 attached, goes to full capacity (actually more like 15-20% above full capacity down) in a fraction of a second.
Any traffic shaping or limiters configured?
-
No traffic shaping or limiters. Running latest 2.3_1 release.
here's my graph. Notice the slight curve at the bottom… It seems to have a tiny bit of lag time between go and GO. Yours just shoots straight up.
modem stuff looks to all be in check.
-
So, this may not be a pfSense issue
I get the same trends of ramp up in my downloads with both my Macbook Pro and Windows 10 Desktop connected directly to the modem with fresh leases.
The question is, now what…
On a semi-related note, I have a spare PCI-Express BCM5709 dual port board. Is that worth swapping in to my ESXi box to replace my BCM5717 onboard? They use different drivers (tg3 vs bnx2i) so maybe there's some extra performance to be gained.
-
That looks like it ramped up to full speed in 1 update interval, within a second. That's nowhere near representative of as initially described. The bit of time until it ramps up seems like it could be something like loading a speed test page before starting the speed test. Seems like it might be reasonable if that traffic graph is typical.
-
On a semi-related note, I have a spare PCI-Express BCM5709 dual port board. Is that worth swapping in to my ESXi box to replace my BCM5717 onboard? They use different drivers (tg3 vs bnx2i) so maybe there's some extra performance to be gained.
Either of those will be more than adequate for 200 Mb, there won't be an appreciable difference between them.
-
It just feels like browsing is slow, regardless of the connection (direct, via pfSense).
I installed a chrome plugin to measure how long a page load takes and they're not long, I just expected them to a bit more … instant, having gone from 50down to 200down.
But at some point, render time and ad blocking negate any substantial benefits past a certain speed. I also imagine it's like a file copy, many small files is always slower than one large file.
My VM without ipv6 enabled seems to load pages with about the same overall speed.
-
Nearly all web pages aren't going to load any faster at 200 Mb than 50 Mb. You're not bandwidth-constrained web browsing at 50 Mb. A typical web page will download in a small fraction of a second at either speed. Unlikely you ever hit 50 Mb even for a split second loading any web page. A bit of browsing around to popular sites, the most ad-intensive sites only got up to 8-9 Mb for a fraction of a second. Just not enough data to transfer for that to matter.
-
Regardless of downlink speed, isn't there still a lot of overhead to loading a web page? Things like DNS resolution, browser side? A lot of how fast a page loads is tied to the server side (how fast it can give you information).
That may give a couple of things to look at, especially since you've said things feel/measure the same(similar) with a client behind pfSense or directly connected.cmb beat me to it.
??? -
Yeah responsiveness is all about latency once you're talking about that high of bandwidth. Web pages load noticeably snappier in the office where we have 10 Gb fiber to the colocation datacenter next door, to a 1 Gb drop, plus a fiber connection to Time Warner, with full BGP feeds from both so traffic takes the shortest path. Bandwidth when web browsing is never any higher than it is at home, rarely exceeding 10 Mb, but there is a significant difference in latency between a cable modem and fiber to a colo facility. About 20-30 ms to Google from home, 6 ms from office. Similar for pretty much anything, there are a lot of websites that are reachable in the office in less latency than it takes to get to the first hop on cable.
It's not a huge difference, but there definitely is one. You can't get that low of latency over copper or coax, regardless of how much bandwidth you have.
-
I use PFSense's DNS resolver to give me 0ms cached DNS response times. I also set the cache to be huge and auto-refresh entries prior to expiration. Nothing like a good cache hit rate.