1 |
>> Anyway, the point of all this is to prevent an HD failure from |
2 |
>> stopping the system. An SSD is much safer, right? |
3 |
> |
4 |
> SSDs are still relatively new technology, so predicting failure rates is |
5 |
> less reliable. What's wrong with using RAID-1? It's proven technology and |
6 |
> totally resistant to a single HD failure. |
7 |
|
8 |
Well, I've read great things about the reliability of SSDs. Here's a |
9 |
comment from Samsung: |
10 |
|
11 |
http://www.engadget.com/2008/02/23/samsung-puts-the-kibosh-on-ssd-reliability-worries/ |
12 |
|
13 |
"a pattern could be perpetually repeated in which a 64GB SSD is |
14 |
completely filled with data, erased, filled again, then erased again |
15 |
every hour of every day for years, and the user still wouldn't reach |
16 |
the theoretical write limit" |
17 |
|
18 |
So in theory, the things are very reliable, but we wonder how they do |
19 |
in the real world. |
20 |
|
21 |
I'm considering a Super Talent Ultradrive which uses the relatively |
22 |
new Indilinx controller. There are 60 reviews of these drives on |
23 |
newegg.com. Of these 60, there are only 2 reports of operational |
24 |
problems, most of the remainder are glowing tales of speed and |
25 |
silence. This is a "cheap" drive using technology that is new even |
26 |
for an SSD, and still the newegg.com reports aren't dominated by |
27 |
reports of "DOA!" or "Failed within 1 week!" like all of the |
28 |
newegg.com HD reports I've seen. Of course, this is far from |
29 |
empirical evidence of SSD reliability, but it's very encouraging. |
30 |
|
31 |
I shy away from RAID1 for a few reasons. I posted these a little while ago: |
32 |
|
33 |
1. RAID is another layer to learn, install, and maintain. |
34 |
|
35 |
2. RAID isn't foolproof: |
36 |
|
37 |
http://hardware.slashdot.org/article.pl?sid=08/10/21/2126252&from=rss |
38 |
http://blogs.zdnet.com/storage/?p=162 |
39 |
|
40 |
3. RAID is relatively expensive on a hosted server. Let's assume that |
41 |
without RAID, the hard drive in use fails every 3 years and causes 24 |
42 |
hours of downtime with good backups. That's a loss of .09% uptime |
43 |
over 3 years. If the server makes $100,000/year (and the same amount |
44 |
every day), that's a loss of $273 over 3 years. However, my host |
45 |
wants $105/month for a second 15k hard drive and RAID controller card. |
46 |
The cost of that over 3 years is $3,780. |
47 |
|
48 |
- Grant |