• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Advise on home server build...

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
I posted to this topic but it looks like it never made it. Oh well. I have to agree you have to take things into perspective. You can get a tbyte nas for not much more then $500. I understand most of these are not RAID5 but you see my point. For a budget home file/media server I would say that that you should consider the older IDE drives. You can normally get these a bit cheaper then SATA and you can pick up the IDE RAID5 cards on ebay for under $100.

If you can find a relatively cheap RAID 5 mobo that has enough ports for what you're looking for and you understand that you won't get top flight performance you should be fine. Just remember that software raid cards use the CPU for its processing, so you might need to increase your CPU/memory. For the most part I've used cheap Athlon XP's with 512mb to 2GB of ram and I've had no major performance issues.

If you go the card route, The LSI Logic's MegaRAID and the Promise Fastrack cards are solid cards at the lower end.

FreeNAS And Openfiler are excellent linux choices. They are very easy to install (driver support can be an issue) even if you don't know much about Linux. I've used both of these in several installs. I prefer Openfiler but that's a personal choice.

As for WHS I'm anxious to give that a try, I'll definately put it up on one of my fileservers.

http://arstechnica.com/news.ars/post/20070108-8573.html is a good read on it.
 
Originally posted by: Madwand1
Originally posted by: erwos
The people dissing it really need to research it - it's based off SBS 2003, and it's got some really stunning features. Performance is also pretty good from the benchmarks I've seen.

Link?
Don't have one offhand. I can tell you from first-hand usage that it seemed pretty snappy. As for features, the new CTP has some slick stuff where you can punch in through the server to rdesktop to any machine on the network (that's capable of it) using just IE. There's also easy-to-use automated backup tools, and the whole WHS storage pool concept. With Linux, you've got to add storage a RAID at a time if you want redundancy, unless you've got the exact same size of drive to grow an existing RAID 5, and G-d help you if you want to try combining multiple arrays into a single storage unit (eg, with LVM). With WHS, you can just slap in any random drives you want to grow the pool. Definitely more flexible.

Whether you could hack in the stuff with a Linux server is almost irrelevant - I can almost guarantee that the time you save by just using WHS versus configuring all this stuff in Linux will make up for WHS' license cost. I've been down the Linux road. It's fun, but it's ultimately not the grand solution that it's being made out to be for this particular usage. (Did I mention I used to sys-admin Linux for work, and that I run Linux on my personal desktop?)

I also whole-heartedly disagree with whomever said to use IDE drives. Modern hard drives will actually max out ATA/133 channels, if you believe the benchmarks. And the cabling is a nightmare. Stick with SATA, IMHO.

Finally, the performance aspect here is being blown WAY out of proportion. Software RAID is pretty damned fast in Linux, and it's the WHS system is no slouch, either. You're better off using those hundreds of dollars you saved on more storage. You also have less chance of failure - no hardware RAID card to fail.
 
Originally posted by: erwos
Don't have one offhand. I can tell you from first-hand usage that it seemed pretty snappy.
...
Finally, the performance aspect here is being blown WAY out of proportion.

In other words, "trust me, it's fast". And "its performance doesn't really matter".

Thanks, but I'd personally want something a bit more objective -- I happen to care about performance. It's fine if you don't and can live with "it seemed pretty snappy"; I'm not making a value judgment on performance for you.
 
Originally posted by: Knavish
While this last statement was a bit trollish...

If you don't enjoy playing with your computer, buy a dell :evil:

QFT! 🙂

Just for the record, I enjoy the hardware side, OC'ing, etc, a lot more than I enjoy learning a new 'programming' language. Having programmed in Fortran, Basic, DOS and a couple other ancient codes I've just lost the bug - and have no desire to catch it again.

 
We're talking home server build I don't think maxing out the IDE's 133MB/s (minus overhead etc...) is that big of a concern unless you're streaming huge files constantly. But I agree that I would only recommend this route if you're on a tight budget since you can get the drives and raid cards pretty cheap.

I agree that linux in general isn't for everyone, I had a hella of a time getting my MythTV DVR up and running. But with customized distro's like OpenFiler and FreeNAS the setup and config is as easy as installing windows, easier even. You're not installing a full OS but a lean specialized OS for the function. Yep Microsoft's home server might be the best route for many people but that doesn't mean its the only or even the best route for everyone.

Besides the home server isn't even out yet, who knows what the reality will be like.
 
Originally posted by: GZeus

Just for the record, I enjoy the hardware side, OC'ing, etc, a lot more than I enjoy learning a new 'programming' language. Having programmed in Fortran, Basic, DOS and a couple other ancient codes I've just lost the bug - and have no desire to catch it again.

While it's no where near as bad as having to write a Fortran program, getting linux going via one of the prettier GUI based distributions is still not quite as easy as Windows. It's probably a bit like when I sit down in front of a Mac. I can run programs, surf the web, etc, but when it comes to system details it takes me forever to dig them out of their menus.

...the people who like tinkering with hardware and the people who like tinkering with software are definitely not the same group!
 
Originally posted by: Madwand1
In other words, "trust me, it's fast". And "its performance doesn't really matter".
What's with the hostility? Those two statements don't contradict each other. It is pretty fast (better than 20 mB/s, which exceeds most embedded NAS solutions), but for streaming media, raw performance doesn't matter very much. Full bit-rate 1080p is probably not going to exceed 50mbit/s. That's, what, a bit more than 6 megabytes a second?

There aren't any big fancy pages of benchmarks because WHS isn't in release yet. No one runs extensive benchmarks on a product that's still basically under development. But as of right now, it's fast enough for the OP's needs for the forseeable future. As someone who tested the beta, I do know that much.
 
Originally posted by: erwos
What's with the hostility?

No hostility, just disdain for seeing a claim being made for good performance based on benchmarks, and then no substantiation of that claim, no benchmarks, and a further attempt to downplay that claim into insignificance.

This is AT, it's a technical forum. Performance measurements are the lifeblood of reviews done in sites like these. Making performance claims without any substantiation is the problem here -- not my pointing out of that.

Originally posted by: erwos
It is pretty fast (better than 20 mB/s, which exceeds most embedded NAS solutions)

This would be pathetic for such a product, well under even single drive performance. WHS hardware requirements are much higher than the typical current consumer NAS box. It should really perform much better, and I hope that it will -- that way the current consumer NAS vendors will be pushed to produce better products.

Vista can already perform much much better, and based on what I understand, I'll claim that due to the OS and lack of RAID support, Vista will perform better than the released WHS (1.0) for transferring large files from the server.

Vista has SMB 2.0, and additional improvements in file and network handling. These would not be in WHS, which is based on the older 2003 code base.

WHS also apparently has 100% storage inefficiency when duplication is used. This adds to the cost and reduces from the data capacity. A decent RAID 5 implementation would both perform better and have greater storage efficiency. (And decent RAID 5 implementations can be found on-board, in *nix, and moderately-priced add-on controllers.)

As I said earlier, IMO WHS has its place and offers significant convenience to some people -- esp. when the hardware choices are also made for them. But there are limitation; it's not the greatest thing in file servers, and IMO technically capable people shouldn't be going ga-ga over it.
 
Originally posted by: Madwand1
No hostility, just disdain for seeing a claim being made for good performance based on benchmarks, and then no substantiation of that claim, no benchmarks, and a further attempt to downplay that claim into insignificance.

This is AT, it's a technical forum. Performance measurements are the lifeblood of reviews done in sites like these. Making performance claims without any substantiation is the problem here -- not my pointing out of that.

Your disdain is SO much better than hostility. Do you think I'm somehow mollified by that?

And, what review? We're talking about software that's still very much in beta. I said that I got good performance out of it, not that I wrote a full-blown review of it. You are a step removed from calling me a liar when you just disregard my own usage as irrelevant. If you want further confirmation of those numbers, feel free to wander over to the WHS boards and ask.

I _didn't_ downplay the performance of WHS. I downplayed the performance needs of the OP. Are you disagreeing with my assertion that media streaming is something that any halfway decent NAS can do, even a relatively slow one? (Caveat: you'll obviously want more performance with more clients, but the OP didn't say he was streaming 100megabit 1080p video to 30 clients at a time, either.)

Originally posted by: Madwand1
This would be pathetic for such a product, well under even single drive performance. WHS hardware requirements are much higher than the typical current consumer NAS box. It should really perform much better, and I hope that it will -- that way the current consumer NAS vendors will be pushed to produce better products.
"Really should" and "really does" are two different things. Most consumer NAS products don't do much better than that. Bitching about performance which is perfectly in line with the competition is a little silly, IMHO.

And I only said that it exceeded 20mB/s - not that it was the limit. My testing was with a mish-mash of relatively old drives. It did well enough with those, but I wanted to put a floor on my claims.

Originally posted by: Madwand1
Vista can already perform much much better, and based on what I understand, I'll claim that due to the OS and lack of RAID support, Vista will perform better than the released WHS (1.0) for transferring large files from the server.

Vista has SMB 2.0, and additional improvements in file and network handling. These would not be in WHS, which is based on the older 2003 code base.

For a guy who was just complaining about my lack of benchmarks, you seem to be real short on them yourself. Link?

Originally posted by: Madwand1
WHS also apparently has 100% storage inefficiency when duplication is used. This adds to the cost and reduces from the data capacity. A decent RAID 5 implementation would both perform better and have greater storage efficiency. (And decent RAID 5 implementations can be found on-board, in *nix, and moderately-priced add-on controllers.)

You're wrong, at least in the target case. WHS detects duplicate files automatically and only makes a single copy of them. That is to say, it has _file-level_ redundancy detection, rather than a striping+parity scheme like RAID 5.

Something else to remember is that there's nothing stopping you from using RAID 5 with WHS. However, RAID 5 has its own set of limitations that you are casually ignoring. There's no way to expand it except using the exact same size of drive that's already in the array. Performance degrades immensely if you lose a drive. And you are generally screwed in the case of RAID cards if the manufacturer decides to stop making the card, and yours dies. You lose some space in the worst case, but there are advantages that go with that trade-off.

Originally posted by: Madwand1
As I said earlier, IMO WHS has its place and offers significant convenience to some people -- esp. when the hardware choices are also made for them. But there are limitation; it's not the greatest thing in file servers, and IMO technically capable people shouldn't be going ga-ga over it.

How do you know how people should react to it before you've even tried it? Look, at least use it before you knock it. I was impressed. Almost everyone who's tried it has been impressed. It really is shaping up to be a good product. I know what's out there in the Linux space - for both work and play, I've managed large, multi-TB Linux NAS setups. This is, at the very least, a pretty good competitor if you just want Samba sharing and some extra features for a mostly Windows network.
 
Originally posted by: erwos
And I only said that it exceeded 20mB/s - not that it was the limit. My testing was with a mish-mash of relatively old drives. It did well enough with those, but I wanted to put a floor on my claims.

Well, that's the number you posted. If you want to say that it can go higher than 30 MB/s for example, you should have posted that number.

Originally posted by: erwos
For a guy who was just complaining about my lack of benchmarks, you seem to be real short on them yourself. Link?

You brought up the subject and mentioned that you had read benchmarks showing good performance, both of which you've failed to substantiate.

I have no problems substantiating my claims; I've posted several measurements in the past. Here's an example:

D:\tools>dir \\amd-vista\f\test\test0\10.gb
Volume in drive \\amd-vista\f is hrd5-6-6464
Volume Serial Number is 24A1-BA99

Directory of \\amd-vista\f\test\test0

08/24/2006 06:01 PM 10,000,000,000 10.gb
1 File(s) 10,000,000,000 bytes
0 Dir(s) 75,692,900,352 bytes free

D:\tools>time 0 nul
The current time is: 12:39:56.67
Enter the new time:
D:\tools>xcopy /y \\amd-vista\f\test\test0\10.gb n:\test\test9
\\amd-vista\f\test\test0\10.gb
1 File(s) copied

D:\tools>time 0 nul
The current time is: 12:41:41.91

That's 10 GB actual file transfer in ~105s which is around 95 MB/s.

This one is Vista to Vista, add-on RAID 5 to on-board RAID 5. This is not the best Vista number I can demonstrate, but uses the motherboard I mentioned above with its on-board NIC and on-board RAID 5. (This was a pull from a machine running a Highpoint RAID 5 card to the machine running on-board RAID 5.)

Originally posted by: erwos
You're wrong, at least in the target case. WHS detects duplicate files automatically and only makes a single copy of them. That is to say, it has _file-level_ redundancy detection, rather than a striping+parity scheme like RAID 5.

Let's get this straight. You must turn on WHS duplication to get its redundancy from its array-management features. If you do so, you get file level redundancy which costs you 2x the space for 1x the data. I.e. if you store 100 GB of unique data, you take 200 GB drive space (perhaps a bit more for overhead, but this should be roughly right). This with a minimum of 2 drives; adding drives doesn't improve the storage efficiency.

Do you dispute this?

Contrast with RAID 5 -- where you use at most 1/3 drives' space for redundancy information. 3 drive RAID 5 => 1 drive loss. 8 drive RAID 5 => 1 drive loss. The efficiency gets better as you add drives.

Originally posted by: erwos
Something else to remember is that there's nothing stopping you from using RAID 5 with WHS.

WHS does not encourage you to use a RAID implementation. I've read notes which warn you about potential problems if you do so. Yes, you can probably get it to work, but it's not going to be formally supported by Microsoft, and it's probably not going to be formally supported by the RAID vendor due to the OEM nature of this OS. The whole non-RAID management claim of WHS is one of its selling point. Either their non-RAID implementation performs well and gives adequate redundancy and storage efficiency or it doesn't -- claiming that it's better than RAID and then saying that you can still use RAID is just a cop-out.
 
Originally posted by: Madwand1
I have no problems substantiating my claims; I've posted several measurements in the past. Here's an example:

*snip*

That's 10 GB actual file transfer in ~105s which is around 95 MB/s.

This one is Vista to Vista, add-on RAID 5 to on-board RAID 5. This is not the best Vista number I can demonstrate, but uses the motherboard I mentioned above with its on-board NIC and on-board RAID 5. (This was a pull from a machine running a Highpoint RAID 5 card to the machine running on-board RAID 5.)

I was actually looking more for the Vista / SBS 2003 comparison. I was also hoping for something a little more scientific than that (eg, iozone). IMHO, your results seem a little unbelievable - Samba is just not that efficient as a protocol - which is why iozone would be nice. That's what's driving me crazy here - you seem to think your anecdotal evidence is acceptable, yet mine is not.

And, more to the point, you don't seem to recognize that the OP implied that an expensive SATA RAID 5 card is off the table due to cost. Again: if he's trying to not spend more than $100 on a mobo, he's not buying a friggin' $500 RAID card.

Originally posted by: Madwand1
Let's get this straight. You must turn on WHS duplication to get its redundancy from its array-management features. If you do so, you get file level redundancy which costs you 2x the space for 1x the data. I.e. if you store 100 GB of unique data, you take 200 GB drive space (perhaps a bit more for overhead, but this should be roughly right). This with a minimum of 2 drives; adding drives doesn't improve the storage efficiency.

Do you dispute this?
Yes, because you're not reading what I wrote. If I have five copies of a file, WHS is only going to store it physically twice. RAID 5 is going to store it five times. If you're using WHS for automated backup, this is not an insubstantial savings. You're not gaining more efficiency with more drives, but you do get it in other cases.

I don't think it's overall as efficient as RAID 5, I agree, but there are some things to be said for it. Certainly, there are corner cases where it would be better, and some where it would be worse.

Originally posted by: Madwand1
WHS does not encourage you to use a RAID implementation. I've read notes which warn you about potential problems if you do so. *snip* Either their non-RAID implementation performs well and gives adequate redundancy and storage efficiency or it doesn't -- claiming that it's better than RAID and then saying that you can still use RAID is just a cop-out.

First of all, I've never seen anything official about not using RAID on WHS. You're going to have to substantiate that, preferably with a link.

Second, I never said it was entirely better. I said it was better in some ways, and worse in others. Reality is not binary. There is no right solution for everyone. I used my own judgement, and told the OP what I thought. You disagree. Fair enough.
 
Originally posted by: erwos
I was actually looking more for the Vista / SBS 2003 comparison. I was also hoping for something a little more scientific than that (eg, iozone). IMHO, your results seem a little unbelievable - Samba is just not that efficient as a protocol - which is why iozone would be nice.

Hint: Vista has SMB 2.

Why would iozone be more accurate than actual file transfers? This is nonsense. Iozone should be calibrated/compared with actual file transfers, to see if it's giving meaningful results and to see how well it models that application performance.

Iozone gives more data (notably the oh-so-apparently-impressive but unreadable 3D graphs) and can model some other applications better perhaps, but it's easily misused and polluted with cached performance. I don't see it (a) being more meaningful than actual file transfers (b) being generally reliable.

Thinking that iozone is somehow more scientific than actual file transfers is itself unsound and unscientific. There are formal reviews which include actual file transfers, if you happen to think being used in formal reviews makes it more "scientific".

I have done comparisons with 2003, and found Vista to perform better; much better in some cases. As to posting those measurements -- they're not really WHS, and use RAID, so are not ideal for comparison.

I've demonstrated good file transfer performance with Vista, and could do so with other benchmarks, but I don't see the point -- file transfer performance is meaningful and reliable in itself.

Originally posted by: erwos
That's what's driving me crazy here - you seem to think your anecdotal evidence is acceptable, yet mine is not.

I didn't accept "seemed snappy" as substantive. Your 20 MB/s figure is fine by me, but it demonstrates rather poor performance, not good performance. If it was really impressive, I'd ask about the details of the test, and perhaps even try to duplicate it myself, but it's bad, so I don't need to do so.

Originally posted by: erwos
you're not reading what I wrote. If I have five copies of a file, WHS is only going to store it physically twice. RAID 5 is going to store it five times. If you're using WHS for automated backup, this is not an insubstantial savings. You're not gaining more efficiency with more drives, but you do get it in other cases.

I understand the point of avoiding duplication of duplicate files. I don't do much of that. I think the other point of storage inefficiency of unique data is more important. For backups, any decent software should be able to do differential backups, avoiding further duplication in the first place.
 
Originally posted by: erwos
Originally posted by: Madwand1
WHS does not encourage you to use a RAID implementation. I've read notes which warn you about potential problems if you do so. *snip* Either their non-RAID implementation performs well and gives adequate redundancy and storage efficiency or it doesn't -- claiming that it's better than RAID and then saying that you can still use RAID is just a cop-out.

First of all, I've never seen anything official about not using RAID on WHS. You're going to have to substantiate that, preferably with a link.

I can't find the original note that I read, but there are notes advising you to disable RAID clearly in the installation / release docs.

I also found this note:

http://forums.microsoft.com/WindowsHome...ShowPost.aspx?PostID=1274906&SiteID=50

Raid is not a supported scenario for Windows Home Server. There are issues with certain RAID cards that others have seen. Some RAID cards show all the hard drives attached to the controller to Windows Home Server and others only show the pool.

I would suggest removing the RAID controller and installing Windows Home Server to a single drive and adding additional drives later through the drive extender.


Windows Home Server Team
 
What makes this board better than the standard boards? I don't really see anything that sets it apart other than ECC memory... which is a good thing.

(I was referring to the board blain had initially recommended)
 
Back
Top