davidktw
Arch-Supremacy Member
- Joined
- Apr 15, 2010
- Messages
- 13,547
- Reaction score
- 1,301
Okay, so lets differentiate the discussion into 2 different tracks. First, satisfying TS's budgetary requirement. At between $300-$600, indeed there isn't much OOTB solution that is performant. I agree on this point. For only way to achieve a relatively performant NAS within the limit of $600 is to DIY a modest box choosing only the necessary components. Spending only roughly $200 on a pair of WB1TB Blue, leaving us roughly $400 for a PSU(roughly $100), Box(roughly $60), RAM(2GB DDR3 = $18), MB(ASROCK B75M-DGS = $98) and CPU(G550 = $30), DVD-RW($23). Still have $70 bucks for a keyboard, mouse, CAT 5E network cable and so this and that like an extra fan.
It based the above price based on the latest price list of Best Bargain for Sim Lim SQ.
As for the OS. The latest LTS version of Ubuntu 32bits Server Edition (console mode), using Linux S/W Raid, and EXT4 FileSystem.
I am pretty sure it will be able to give a sustain at least 60MB/s using iSCSI, other maybe other file sharing protocols too. Of course, a lot of tedious work and some good Linux skill sets to get stuffs moving
Therefore to answer TS question, at $600, it is possible, but do TS have the means to piece up all these stuffs together.
Now moving on to a more debatable discussion on the EXTRA requirement that you have placed in below...
The keyword here is having the NAS giving performance similar to a local SATA drive. The answer to such requirement is no, TS budget can't do it. But to be frank, even most Enterprise budget can't do it either. The only few interfaces that I can think of that achieve such network latency is Fibre Optics at least 4Gbps bandwidth or Infiniband DDR and above. That's the requirement from a RAW performance talking about network latency only. I have not even go into storage capability. Here I will using SATA 3G in comparison.
Like I say not even most Enterprise solution come close to such capabilities. Those using 1Gbps Ethernet solutions which are already more than 5 figures themselves won't hit such performance, let alone any consumer budgets.
After having the necessary network interfaces for remote storage, then we move on to harddisk. Do you think if I place a SATA harddisk spinning at 7200rpm remotely or locally under the SATA interface going to make any difference ? The answer is not likely. In that case, the next topic is on harddisk capability. Earlier on you reply to TS the following:
Then my response to the above statement is NOT RELEVANT. Even after you throw in 5 figures for network configuration, that match up with the exciting enterprising solution, data are not going any faster than the weakest link in the storage solution - harddisk. Here I tackle the crux of the statement "Using the NAS as a secondary store ..." and "try to move 1T at a go".
What I'm stressing is it doesn't matter it is local or remote over the best network connectivity. How fast external transfer between 2 harddisks is not going to change the fact that the bottleneck will be the speed of transfer of the harddisk. I hope it's clear. The statement above is fundamentally flawed because it's the harddisk that is slow, not whether it is NAS or not.
Answering to your "requirement" of having a hard disk farm and expandability and so forth, you didn't exactly match up with TS's requirement. You added NEW requirements of having expandability. Nothing is free of course, expandability comes with its own cost too. As such I won't tackle this aspect because clearly it's out of scope.
Using my budgetary quote on $600 system and I can assure you can get to at least 60MB/s of transfer speed, will you consider this faster than USB2 which flood at around 25MB/s ? One thing for sure is those NAS at such price range are marked up for branding, OOTB experiences and some bells and whistles regarding software packages and of course a customized box solution. After adding all these extra costs, one should at least expect something around $1K if wanted the same raw performance versus the $600 DIY NAS box I proposed above.
Even so, the entry price for Synology DS212j has no issue achieving 80MB/s for iSCSI but one will need to fork out at least $700 in Singapore for the box only. Add additional 2 suitable harddisks will set one back like around $1K. Looking at the price, I don't see how it requires near 3K or above as you have quoted earlier. Mind if I say so, even really Enterprising solution wouldn't be 3K because of the "Enterprising" nature that I will discuss subsequently.
No I absolutely do not concur that there is speed issues with 1TB file transfer. I have already explained very clearly above that this argument is flawed. It has nothing to do with whether it's a NAS or not, it is a fundamentally harddisk bottleneck. Placing a slow harddisk on fastest Fibre Optics won't make it fly for speed throughput.
Mind your terminology. I quoted CAT 5E, not CAT 5. Be very specific about such things and not use them loosely. CAT 5E are capable of Gigabit throughput. NICs and Switches will go into Gigabit network when they detected a CAT5E and above network cable. I have been using CAT 5E cables for years in gigabit network, I don't see how CAT 6 makes a lot of differences in my life. I did the necessary performance test and find negligible performance gain to convert into CAT 6 for raw performance purpose.
If you use the word CAT 5, your fiber cable installer will not assume it is CAT 5E.
I have already mentioned, using CAT 6 or CAT 5E cable for consumer is what I called a placebo effect. The owner feels happy having a better cable, by all means. The owner wants to prepare for nuclear fallout and still the CAT 6 miraculously works, by all means. The fact is for normal consumer home, you will find almost no differences in performance. Don't believe ? Connect both ends of the cable to 2 linux boxes, and test out using iperf.
I know this sounds ridiculous, but what is ridiculous is not what I say, it's what most people assume. I am daring to say that more than half of the people that uses CAT 6 cables don't even test the network cable performance in the actual usage scenario. They just assume since it's only a couple of bucks more than CAT 5E cable and since they are going to ALWAYS go for future proofing, hence lets get it and over with it. After all, it's not easy to take out the cable under the tiles or inside the wall, why save now and later a few years down the road need to change.
This is the typical "kaisu" mentally that most people have and thats why CAT 6 cable selling at hot pancakes. Recently I see CAT 7 cables capable for 10Gps already in the market. I'm sure you will find a few wealthy folks getting CAT 7 cable soon. They are going to prepare for year 2500 when 10Gbps network becomes common.
I have already put out the citeria. Consumer are putting their cables at HOME. This premises is so darn lack of electrical interferences. How many cables are these fellas going to bundle together ? I tell you what is in the data centre where cables are hanging above the server racks. They are bundle in tens or hundreds together. They lay across the rooms of data centre in tens of metres, not just 10m from the living room to the master bedroom. They are situated in a room with hundreds of server churning away. Each server with possibility 6 pairs of high speed fans, hundreds of harddisk, and hundreds of mainboards all screaming at the top of their "lungs" with Electro Magnetic waves.
Each time these waves hit the network cable, it's going to attempt bit flips and tons of other EM interferences. From the network perspective, this is BAD BAD BAD. TCP/IP transmission will get checksum error, retransmit is going to happen. Unnecessary network load for retry. Higher latency to affect heartbeats between clustered systems. A few millions investment is not going to function at their peak because some network manager wanted to save 10K laying CAT 5E cables when he can use CAT 6 or even CAT 7. Whose head is going to roll. Get the picture ? If I'm the network manager, I will lay CAT 10 cables.
At home, SUPER overkill. Weak EM interferences, not running critical business worth millions, just a couple of slow network. Waste of MONEY. I rather use the money and buy more harddisks.
If you really want some reference to support my claim, read here at http://www.cat-5-cable-company.com/faq-cat6-v-cat5e.html. However this is not the source of my claim. The source of my claim is actual bandwidth and latency testing on real cables. BTW the word "certified" in what I learnt is FUD. It's for people that want to hang on to something. Once the cable is testing out and working, certified or not doesn't matter. What matters is TRUTH and real testing. There are just so much certification that means nothing. It's the science behind it that matters.
I do have one question for you, what gives you the impression that your gigabit router is not the bottleneck and your cable is ? Just because router claims to be gigabit doesn't mean they will route fast enough. Just because a switch is gigabit doesn't mean the switching throughput is good enough. Don't just look at claims, look at real benchmarks that are tailored to your needs. Don't look at raw benchmarks because those benchmarks are meant to hit the extreme. Look at the hourglass, do you think it matters how big the funnels are at both ends when the centre is so narrow ? If you can't identify where the bottleneck is, then the benchmark is useless. It just tell you how good the equipment is without telling you how good your solution is.
If you take into consideration how an Operating System is designed, you will find the load upon the storage subsystem doesn't need to be constantly peak performance for the whole solution to be at the peak. That's because the peak of the solution has bottleneck somewhere else and not the storage component. Your examples on multimedia usage most of the time is processors, gpu and software bottleneck, not so much of the storage system UNLESS you are talking about hundreds of users using the same storage component. Most consumer don't reach more than 3 concurrent connections, let alone hundreds.
I have been working in this industry for more than 8 years, I can tell you what all these Enterprising "bullxxxx" are. It's all about keeping your head on your neck. It's all about convincing your management that for a multi million project, everything must work as planned. Lets get the budget down first and talk about the rest later. You think the budget come from after accounting all the required hardware, software and so forth ? Everything is buffered heavily. Why ? Whose head is on the chopping board ?
Lets talk about truth. When a harddisk break down, the server crash, and so forth. Replacement are not coming in from your nearest DIY hardware store nor SIM LIM SQUARE. They come from the vendors. There will be procurement delay. When the load spike, things must continue to run. When some servers go offline or maintenance or upgrades, the front facing system MUST continue to server the stipulated load and enforce the SLA. That's why systems, peripherals and so forth are normally over-provisioned. For some agencies I work with, they over spec some resources by over 60%.
Do you think all these costs they you gather from Enterprising solution justify for consumers ? The answer is NO. That's why the ball price in the Enterprise arena is not your consumer budget. Not because they pay more and get better hardware. It's about assurances, it's about migrating business risks and so forth. I don't suppose between a couple at home, the NAS failed for 1 week will stop the household from functioning right ?
That's why such EXTRA comfort that comes with Enterprising solution are so hefty. I hope it seems all clear to you. Don't ask the wrong group of people for the wrong purpose. What they quote are more than what you need.
Last edited:
. But if one is wiring up the house now then can get Cat6 cable cos since price difference not that much and being a SGians, we are kiasu so hoot the better one 
