1 |
A. Khattri wrote: |
2 |
> On Wed, 19 Apr 2006, kashani wrote: |
3 |
> |
4 |
>> I'm not sold on the Google approach. |
5 |
>> |
6 |
>> Assuming someone was to build nine data servers we're talking roughly |
7 |
>> $3k per server (dual CPU, 4GB ram, raid 5 sata) or $30k with shipping |
8 |
>> and tax. |
9 |
> |
10 |
> Actually Google use the cheapest hardware they can find, buy in bulk, and |
11 |
> they assume stuff will fail so they plan accordingly. I very much doubt |
12 |
> they spend $3K per server... |
13 |
> |
14 |
|
15 |
Anyone trying to build the same with a purchase of less than 100 servers |
16 |
is not going to spend much less. |
17 |
|
18 |
2 x 2GB RAM = $1000 |
19 |
2 x CPU = $500 or so |
20 |
Raid Card = $300 |
21 |
Drives = $100 each |
22 |
1u chassis/MB/etc = $500 |
23 |
|
24 |
IIRC they drop the drives and shove everything into RAM. Which is fine |
25 |
when you have a limited data set and enough machines to shove it into |
26 |
RAM. Originally Google did have a limited data set. It was only after |
27 |
the infrastructure reached a critical size that they began Google mail |
28 |
and other large storage things. And had fours years to work out the |
29 |
operational kinks. |
30 |
|
31 |
I think it unlikely that someone with a single storage set has enough |
32 |
money or time to pay for a few Phds to write a custom filesystem, a few |
33 |
hundred servers (10TB/4GB), and the datacenter monkey necessary to |
34 |
replace gear constantly... oddly Google has both of these in spades. And |
35 |
if you read the whitepaper the smallest Google data cluster is nineteen |
36 |
servers, aka $40-60k for schleps like us, aka the cost of a SAN that |
37 |
that uses less power and burns less switch ports. |
38 |
|
39 |
This infatuation with the Google stuff that very few people (ie none of |
40 |
us) have the in house infrastructure to handle or the available cash is |
41 |
useless. Unless someone has actually built their own mini Google and |
42 |
wants to tell us all about it with nice numbers like total cost, source |
43 |
code, transactions per second, cost per GB of user data, throughput, and |
44 |
other data points. |
45 |
|
46 |
kashani |
47 |
-- |
48 |
gentoo-server@g.o mailing list |