Macworld Forums: Mac mini with Lion Server ideal for SOHO and SMB - Macworld Forums

Post your comments for Mac mini with Lion Server ideal for SOHO and SMB here

"Peace of mind"? Have you done a study on whether ECC memory actually performs any better than non-ECC memory? Is this just another one of those IT requirements that aren't necessary anymore? Same with redundant power supplies. I ran a system with eight Xserves (original versions) and I replaced one power supply. I also had 300 Macs at any given time and I don't remember ever replacing a power supply in any of them. The mini draws a lot less power than the original Xserves, putting a whole lot less stress on it's small power supply. As far as monitoring, I liked to look at the green lights but Mac hardware runs very well when you have it properly configured and leave it alone (I rarely turn the power off on any of my Macs, it's like a light bulb--power cycling (hot-cold) is what blows them out).

The mini is not a high-powered, computational type server but it satisfies the needs of a small to medium size shop. I would, as always have more than one available but that's just my old-time IT gut speaking.

rob53, on 29 December 2011 - 07:55 AM, said:

Features like ECC memory and redundant power supplies in servers aren't desired because the failures they address are considered likely. They're desired because in the specific application those failures and their attendant consequences, should you be unlucky enough to be in that tail of the bell curve, are outright unacceptable.

rob53, on 29 December 2011 - 07:55 AM, said:

ECC memory is like any other kind of redundancy. Most of the time, it's not needed, but when something goes wrong, it's good to have. in any event, the price difference between unbuffered ECC and Non-ECC DDR3 1333MHz on crucial.com is not huge for a server.

rob53, on 29 December 2011 - 07:55 AM, said:

I agree completely, which is why I love the price point. They're small and cheap. Get 2-3, and life is good.

Would the mini server make a good BackUp for a total of about 40 iMACs/MacBooks all have Time Machine. I had no luck with Apple's Time Capsules had two already both failed just after warranties expired. I like the ease of use of Time Machine but problems start occurring when backUp drives start getting full. Suggestions?

@bastion and @jcwelch, I know what you're saying but my comments were made based on 15 years managing and supporting a large in-house Mac installation. This started in 1990 and ended in 2005 when I changed positions. The older equipment had more problems but the newer equipment had evolved into a state where I was confident it would work 24x7x365. The redundancy mindset is good, as long as it isn't taken to extremes. We're talking about small Mac servers providing services for at most a couple dozen desktops. That's where the mini-server is positioned.

The reason I asked about ECC memory testing is to see if anyone has ever run tests on standard memory to determine any failure rate. Historically, ECC memory was created because common memory had problems. Is that a concern anymore? Please don't give me the standard IT answer. I've been there and nobody ever complained about changed data.

The same goes for power supplies. Power supplies from 10-20 years ago were built the same way they've been built since the dawn of time. Newer power supplies have been re-desigend, primarily to fit into smaller spaces but also to provide cleaner, more consistent power. I would like to hear from service vendors on the frequency of power supply replacements (only Macs, not garbage PCs).

If I were setting up a small office system server, I would never set them up with only one server. I'd always have a secondary server so I could keep system access available during patching and in the event of a failure (nothing is perfect). My point is, Apple manufactures good quality hardware that runs a lot better than many other brands. Let it do its job.

GTPSr, on 29 December 2011 - 09:12 AM, said:

You can plug a USB hard drive into an Airport Extreme base station and provide network backups.

If the Airport Extreme dies, you just replace it and the backups on the hard drive aren't affected.

GTPSr, on 29 December 2011 - 09:12 AM, said:

The mini itself, probably not. But, if you pair it with a Promise Pegasus R6, that's 12TB connected over thunderbolt, talking to the network via GigE. Should work rather well.

rob53, on 29 December 2011 - 09:34 AM, said:

I'm giving you the standard answer with regard to any form of redundancy or backup. It's unneeded until it's not. in terms of "real world", historically, there were, and still are, issues with background radiation/cosmic rays affecting RAM states. This can be heightened during heavy solar activity, although engineering improvements have helped mitigate this problem. As far as other errors, from Wikipedia:

Work published between 2007 and 2009 showed widely varying error rates with over 7 orders of magnitude difference, ranging from 10−10−10−17 error/bit·h, roughly one bit error, per hour, per gigabyte of memory to one bit error, per century, per gigabyte of memory.[2][8][9] A very large-scale study based on Google's very large number of servers was presented at the SIGMETRICS/Performance’09 conference.[8] The actual error rate found was several orders of magnitude higher than previous small-scale or laboratory studies, with 25,000 to 70,000 errors per billion device hours per megabit (about 3–10×10−9 error/bit·h), and more than 8% of DIMM memory modules affected by errors per year.

It all depends on your comfort level. I don't find ECC to carry an especially onerous overhead in either cost or performance, and for server usage, I think both are acceptable penalties for the extra layer of safety ECC provides.

rob53, on 29 December 2011 - 09:34 AM, said:

replacement rates are a somewhat meaningless metric, because they don't take into account uptime/availability requirements. If you're talking five 9's or higher, then the fact you *may* never have a problem is meaningless. You still have to build that system to handle those needs even if something goes wrong. Focusing only on best-case leads you down almost the same path as only focusing on worst case. There's no one metric that provides the right answer for all needs.

rob53, on 29 December 2011 - 09:34 AM, said:

I worked in a small office, around 50 or so people. There's no way a mini or even two could have handled their needs, because it was a scientific computing company that ate CPU, memory, hard disk and network for lunch. For them, despite their small size, you would have needed gobs and gobs of minis to handle their needs, and, any advantage the mini gave you over anything else was gone. Apple does make good server hardware, but so do other top-tier companies, like IBM, Dell, and HP. Letting the hardware do it's job is not the same as blindly trusting it.

OriginalMacRat, on 29 December 2011 - 10:13 AM, said:

Caveat: While some people have gotten this to work it has never been a supported configuration.

rob53, on 29 December 2011 - 09:34 AM, said:

This is from a couple of years ago.

http://www.cs.toront...igmetrics09.pdf

I still think you haven't quite understood the point I was trying to make. You seem to be focused on risk. "The chance of failure X occuring is 0.00001 and I consider that risk acceptable." My stance isn't primarily about risk; it's about consequences. "Someone's going to be in the lower tail of the bell curve and I can't tolerate the fallout if it's me."

Think of it as the difference between contraception and abstinence.

Why yes, someone HAS done a study of DRAM error rates. Check out http://storagemojo.c...on-dimm-street/ which reports on a study by Google and Prof. B. Schroeder of the U of Toronto, which begins:
<blockquote>
A 2½ year study of DRAM on 10s of thousands Google servers found DIMM error rates are hundreds to thousands of times higher than thought — a mean of 3,751 correctable errors per DIMM per year. Another piece of hallowed Conventional Wisdom bites the dust.
</blockquote>

Yes, Virginia, there are DRAM errors. ECC is needed for critical operations.

You say at the end of this review that your review of Lion Server notes several troubling problems with Lion Server software.

http://www.macworld....ion_server.html

The software is what this is all about, right?

How can you give this a 4.5 star rating when the software that you buy it for is 3.5?

You seem to be burying the server software problems with a glowing review of the hardware.

Luckily, I have Snow Leopard Server on a Mini serving my needs. I know that is not being sold anymore... but....

Any other advice for people wanting to buy a server now? What are their options? That would be more helpful. We know the Mac Mini is
a good piece of hardware.

I see that this product is a real priority for Macworld to review. Released on July 21, reviewed on December 29.

Could've waited a couple more months and you would have gotten a new model to procrastinate on.

Oh, btw, a Pegasus R6 would be overkill for a Time Machine backup. If they need that much storage, an OWC Qx2 would be nearly half the price and have enough performance connected over Firewire 800. If they don't need that much storage, their other Firewire products would fit the bill nicely.