SATA mini SSD install question
-
Hi im setting a PFsence router up mainly to load balance 2 connections, ive only got a sprae 500GIG HDD and TBF dont really want to waste all the space on a router, would one of these be ok http://www.ebay.co.uk/itm/KingSpec-8GB-SATA-DOM-MLC-KDM-SA-51-08GMJ-SSD-42-MB-S-Solid-State-Disk-/190703794464?pt=UK_Computing_Other_Computing_Networking&hash=item2c66d4f920
eg would it be enough space and fast enough i know there not the fastest of drives in the world but at the price there quite cool will also be less heat/noice generated than a HDD.i want to do a full install of Pfsence and as i say just use it for load balancing 2 connections would it be fast enough for this
Thanks Ash
-
Use an USB flash drive, cheaper and easier to find.
-
I actually Run those for my freeNAS and PFsense boxes
Use an USB flash drive, cheaper and easier to find.
sure, but flash drives are slower, not internal, and not as reliable.
I prefer FAST solid state, not slow solid state but that's me. -
thanks both for the reply's i do have a 8GB USB 3.0 flash drive which has a read of about 35MB and a write of 5MB i thought that might of been too slow (the mobo im using also dose have USB 3.0 support) if its ok im happy to use that as i already own it But if it would be too unreliable would the sata option be better, i know it sounds silly but i dont want to pay £40 for a 32gig SSD because it would be a lot of wasted space although obviously the read/write would be vastly superior the last time i messed with PFsence was on a p2 256mb ram and a 6gig HDD and i think it was about 2009 this was a basic test so im not sure of the real requirements for HDD speeds,
Im wanting to balance a 20meg connection and a 12meg together if that makes any diff on the HDD speed requirement the 12 meg will later become a 60 but not for some time yet
if it comes to it i will just use the 500gig HDD that's in it but as said before it would be a waste of space and also more heat/noise with a HDD than some form of flash storage :)
Thanks again Ash
-
I use a Transcend DOM (Disk On Module) solid state storage emulating a hard drive. It has storage capacity of 1GB and has been the main storage of my home pfSense for about 4.5 years. I use the variant with IDE interface but there are variants with SATA interface. Cost at time of purchase was under US$30. Raw writes with a largish block sizes are reported to run at over 9MBps.
-
I'd advise that you get yourself a small 2.5 inch HDD new at a low low price. SSD is still in the "Works fine for some experts" phase if you ask me.
There are not many "My HDD sucks" entries in the forum but tons of "Whats up with my crashed SSD" threads. -
ahh right i use a SSD on my pc and have had no probs at all with it (had 2 actually upgraded a 128 to a 240) but i suppose in this circumstance its a different environment so could have probs, this is why best to ask on advice ay, i could get access to a laptop drive quite easy for free TBF just more had my eyes set on using SSD for less noise heat ETC lol,
Ash
-
Run it. Do a full install. Make sure you use 2.03 release of pfsense ;D
Should be fine.
-
I can, in good conscience, only recommend less than amazing SSDs for stuff that supports trim "out of the box" for most people.
-
I doubt it will make enough difference to shake a stick at in the noise/heat area, but knock yourself out.
Either it will work fine; or it will die young. Reports are mixed, opinions frequently seem heated. The price is not attractive, especially if you can get a laptop drive for free. You should, however, do what makes you happy. According to one side of the debate, it will work just as well and as long as (or longer than) a spinning disk, and according to the other it won't. The price and the existence of the debate are two strikes against them for now, for me, but I'm not you.
If you are not using squid, there's little reason not to simply use the embedded install, loading from a USB or CF into RAM and then running as fast, quiet and cool as you could possibly want. If you are using squid, I hold that there is excellent reason to use an entire 500GB (or more) disk to get a squid cache that's worth something. I cache from 0 - 4Gb, and it makes the second and succeeding times I have to install system updates fly…
-
I actually have this neat (debatable) little equation I stick too - 2GB free un-used ram dedicated for every 30GB or so hard drive cache. Cache has to be indexed in RAM. For most any install, 100 GB is way more than enough HDD even if you go crazy with on-board RAM.
-
…and I have 16GB of RAM, most of which isn't doing a lot with squid not running - so I'm in good shape by your formula once the non-cache portions of the disk and disk manufacturer nominal size inflation are accounted for.
-
If you run and actually acquire 500 Gigs of disk cache, your system will crash.
Go to a terminal and type df -h
I bet its not bigger than 200 GB.Mine is only using 16GB and its been running a while…
-
/dev/ad4s1a 41G 380M 37G 1% /
devfs 1.0K 1.0K 0B 100% /dev
/dev/ad4s1d 9.7G 201M 8.7G 2% /var
/dev/ad4s1e 989M 3.5M 907M 0% /tmp
/dev/ad4s1f 369G 226G 114G 66% /squid
devfs 1.0K 1.0K 0B 100% /var/dhcpd/devWhatever you bet, you just lost it by 26GB. ;)
Now, the system that's only been up since July 1 is only at 86 GB so far. The primary users for that system won't be back until September, and left before it was brought up. It may well get wiped and started over before then - I kludged it up n a rush when something else died before it was replaced. The system above has been up since February. -
How much RAM is it using?
I added up your usage. Must be gobbling a chunk of ram.
-
Per dashboard, 89%, with a RAM cache of 6144M (ram cache size is played with to maximize RAM use while not going into SWAP use. If swap is used, Ram Cache is shrunk.)
The RRD memory graph report is more detailed, but seems to want to come with the whole picture.
Active holding near 77% for the week. Inactive 6.7, Free 1.79, Cached 2.30 and wired 11.50% (Average, not a lot of variation with min and max) Speaking of which, I just love the indistinguishable colors, and the tantalizing note that "change of style or color may not take effect until the next refresh" which implies that the colors could be changed, but the GUI provides no way to actually change them, that I can find.Or perhaps this one from system activity would suit you better?
12.513Gb just for Squid.14596 proxy 64 20 12995M 12412M kqread 1 63:30 0.00% squid
-
Seems good then. I am only running 1.5 GB RAM cache (4GB total ram) on my home pfsense and 30 GB Disk cache, which I have yet to fill. Memory usage bounces around between 30% and 70% depending. Its like it will bubble up to 70+% usage and then suddenly drop to 30% usage. No reboot or anything. I'm going to let it go a year or so and maybe I'll revise up to use my system. System free resources to a point is good, beyond that its waste.
-
I have 25 or so users on this system - it's not a home unit. The other one gets 85+ during the school year. If I can get even 5% cache hits it saves a lot of time with 10 Mb & 25 Mb download speeds (and typical daily traffic of 1-
1832 GB). When something like a big system update comes along, hit percentage can rise quite a bit. When vast sums of bandwidth are being piddled away on watching TV or Movies, not so much… (32GB day, 16GB to netflix. Uggggh.)I also get the memory usage dropping without a crash or anything I've noticed in the logs (everything seems to keep working, so I don't comb the logs too hard, but I have looked to see if there's something that correlates.) That happens more often on the other system, for some reason. I pushed the RAM Cache size up until I started using swap, after having it run at lower levels for quite some time, then backed it off. I figured I had the RAM, it might as well be working. Ram object max size is 7200K, not the default 32K. On your system I'd set it to 1500K. I really don't get the fascination with small files, when my reason for using proxy servers is to download big things once (though I'm perfectly happy to cache small things and don't have a lower limit set.) I don't believe or don't get the comment near the disk cache object size "If you wish to increase speed more than you want to save bandwidth, this should be set to a low value." Disk cache is about 100 times faster than my internet, so it makes no sense to me. Pulling large objects out of disk cache is going to beat the WAN network 10 times out of 10, and the bigger they are the more it wins.
Likewise, I use Heap LFUDA for both RAM and Disk Cache.
It looks like at the moment I have the disk cache size set to "only" 250000MB, despite having 369GB set aside for it in the partition, so it likely got trimmed down to the low watermark to be 226GB...
-
I get only about 5% cache hit. Mainly because most content is dynamic. I see this as a SERIOUS shortcoming of squid as its currently implemented.
The other systems are not running cache. They serve mainly as VPNs.
Home system cache hits this month:
Total/Average: 13 6 82.0 G 454.5 M 5.05% -
5% is perfectly respectable. If there are few users on the system and I do a bunch of large system updates to a bunch of computers, I can get it higher (I hit 48% on August 1, I think I ran around applying updates all day), but on a daily basis, 5% is decent, and I'm not sure how squid could do much better given the nature of the "modern internet" unless a bunch of people were all on the same network accessing the same things - which happens a little, but not all that much. But, if what squid can do is not worth it for your operation, not using squid does make deciding to go embedded/diskless that much easier.