Menu

Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Messages - TL

#1
Setting up a new box - Looking for decent power use to performance, not an especially difficult deployment (~200 Mb service, home use, one WAN, no failover...)

Have a couple of available options, both on boards with Intel NICs.

Intel Core i7-2600 3.4ghz  "95W"
CPU Benchmark ratings:
MultiThread 5379 - single thread 1752 Cross-Platform Rating  11,300

Intel Core i5-3470 CPU @ 3.20GHz "77W"
CPU Benchmark ratings:
MultiThread 4660 single thread 1927 Cross-Platform Rating 10,737

They are not far apart, and the i5 actually has better single-thread rating as well as lower full-power rating (the two systems are fairly similar in power use (25-35W running) as I currently test, but I don't have the network up to connect them to yet so they are not working hard, and I have also not (yet) completely minimized differences between them (I won't be processor-swapping even though they share a socket type.)

I've used a similar i5 system in a much more difficult environment (100+ users) running pfSense in the past, which ran into limitations when trying to do traffic shaping after the pipe got upgraded (which was soluble by simply removing the downstream TS and leaving upstream TS in place - otherwise it was choking at 150 Mb downstream on a 400 Mb feed) but I doubt I'll have much need of traffic shaping for this job.

Not especially interested in buying any new hardware for the job at present - even if it used half the power, that's only going to save $25/year or so, so most new things will not pay off in a reasonable time at present, AFAICT.

If single-thread performance is more important, the i5 seems like a win - both lower maximum power use and better single-thread performance.

As time permits I'll be minimizing differences to see if I can get a better direct comparison, but if there's insight indicating that single-thread matters more most of the time, then I could save a bit of time by not running those comparisons and just concentrating on the getting the i5 system set up.

Thanks,
#2
I've been using pfSense for a number of years, saw the writing on the wall when they started in with the "gold" menu and have been waiting forever for them to fix some things, so OPNsense looks like the way to go, probably.

But, at present, pfSense is working, for limited values of working.

In transitioning, are there tools to import a pfsense backup and process what can be processed, and toss the rest, or is it ground-up configuration rebuilding time?

Does OPNsense solve my major pet peeve with pfSense as it sits, which is that traffic shaping (primarily the limiter, as I find that it does "fairness" the way I want, better - share available bandwidth fairly, rather than limit arbitrarily even if lots of BW is unused) AND run transparent squid sensibly (ie, squid cache hits are not limited, so we actually reap the benefit of squid in improving performance beyond what our outside line can do)?? I have tried many variants of many peoples claimed recipes for this on pfSense, and none have worked, and most have resulted in the system needing to be restored to a working configuration. Since ~pfSense 2.1.5 the two things simply can't run at the same time (before that they could, but cache hits were shaped.)

For almost a year now I just gave up on transparent squid as the limiter/bandwidth sharing is more important to overall function, but I'd like to get transparent squid back so I can re-implement squidguard filtering. Non-transparent squid with overbearing rules/individual setup is not a good solution IMHO, and the antiquated and never widely adopted auto web proxy discovery methods are largely pointless.

I found a few threads with "pfsense transition" in the title, but not much about the actual transition in the thread, when I searched for "pfSense" here. If I've missed a helpful, detailed thread, please point me to it. If not, let's make this be that thread.

Thanks.