05-28-2010
The difference comes down to random versus sequential access. A traditional mechanical drive gets very high transfer rates only when its heads don't move much, like when it's reading one large file beginning to end.
Imagine 100 512-byte files stored in different places on the hard drive, with the drive having to drop everything and reposition its heads to read each one; with seek times of 10ms, that's an entire second spent retrieving 50 kilobytes of data -- an order of magnitude or so slower than it's rated speed.
A solid state disk has no heads to reposition, hence almost no seek time, so blows the pants off mechanical drives for random access. But individual flash memory cells are a bit pokey so a mechanical drive still beats them for raw transfer once its heads are in place. Of course read speeds can be improved in either just by throwing more money at them, more things running parallel in either gets better speed.
Both of these situations assume no cache, of course. Given enough memory the system will just cache everything after its first load to make all seeking and transfer rates irrelevant. And tiny files are, well, tiny, unless you have millions of them.
One thing SSD's are terrible for is frequent writing. Writing to flash is slow and inefficient, and gradually wears them out. Hard drives wear out too of course, but there's a world of difference between "this drive will last a couple years continuous use, no matter what you do to it" and "you could kill this drive in a week if you treat it badly enough". DRAM-based SSD's are free from this of course.
Last edited by Corona688; 05-28-2010 at 04:47 PM..
7 More Discussions You Might Find Interesting
1. UNIX for Dummies Questions & Answers
Dear All,
I run a website for a non-profit. Does anyone know where I can get free or cheap software to run threaded discussions for our website?
Our website is obviously running off a unix platform.
Thanks (4 Replies)
Discussion started by: evertk
4 Replies
2. UNIX for Dummies Questions & Answers
Hi, I'm fairly new to UNIX and am looking at the potential of using UNIX servers as Application Servers (similar to what is done by Citrix Metaframe and Terminal Services in windows) does anyone know whether this is possible or am I firing in the dark?
Thanks. (1 Reply)
Discussion started by: tombobs
1 Replies
3. UNIX for Dummies Questions & Answers
Im trying to use a CISCO WiFi-BRIDGE and Fedora9 to connect to an I-net connection down the road. Then use the Fedora box to route the connection to 4-Xp workstations using the remaining 4 NIC's in the box.
My paln is to eventually connect my home network/I-net to my work network 1.5miles down... (0 Replies)
Discussion started by: Solomen
0 Replies
4. UNIX for Dummies Questions & Answers
Hi guys,
I have developed a small website using jspx pages.
Now i want to host the website to a server which is a linux machine.
The linux machine which i m referring to is already hosting some web pages(jspx) through a web server Which is running at port 8888.
My first question would be... (0 Replies)
Discussion started by: pinga123
0 Replies
5. UNIX for Dummies Questions & Answers
I've been wondering how I can reduce latency in downloading webpages. I've been pointed to network sockets (This guide in specific, beej's guide to network programming ) as some method to do so, but I can't figure out how to apply it in the way I need it.
I am iterating through webpages like... (0 Replies)
Discussion started by: seagaia
0 Replies
6. Shell Programming and Scripting
hi
i was trying to run the HTML script and was unable to run it as the apache server was not loaded on my linux server.....how do i check whether A[pache has been installed in my server or not.....???? (1 Reply)
Discussion started by: kullu
1 Replies
7. Shell Programming and Scripting
Hi,
Is it possible to have a script to interact with webpages. I want to create a script that logs a user into a specific site, and is able to get/post information.
Would anyone give me instructions on how it's should be done, and where I can find information on starting it out.
I know... (2 Replies)
Discussion started by: Pztar
2 Replies
LEARN ABOUT CENTOS
crosspost
CROSSPOST(8) System Manager's Manual CROSSPOST(8)
NAME
crosspost - create the links for cross posted articles
SYNOPSIS
crosspost [ -D dir ] [ -s ] [ file... ]
DESCRIPTION
Crosspost reads group and article number data from files or standard input if none are specified. (A single dash in the file list means to
read standard input.) It uses this information to create the hard, or symbolic, links for cross posted articles. Crosspost is designed to
be used by InterNetNews to create the links as the articles come in. Normally innd creates the links but by having crosspost create the
links innd spends less time waiting for disk IO. In this mode one would start innd(8) using the ``-L'' flag.
Crosspost expects input in the form:
group.name/123 group2.name/456 group3.name/789
with one line per article. Any dots in the input are translated into "/" to translate the news group into a pathname. The first field is
assumed to be the name of an existing copy of the article. Crosspost will attempt to link all the subsequent entries to the first using
hard links if possible or symbolic links if that fails.
By default, crosspost processes its input as an INN channel feed written as a ``WR'' entry in the newsfeeds(5) file, for example:
crosspost:*:Tc,Ap,WR:/usr/lib/news/bin/crosspost
To process the history file and re-create all the links for all articles use:
awk <history -F' ' '(NF > 2){print $3}' | crosspost
(where the -F is followed by a tab character.)
The ``-D'' flag can be used to specify where the article spool is stored. The default directory is /var/spool/news.
By default crosspost will fsync(2) each article after updating the links. The ``-s'' flag can be used to prevent this.
HISTORY
Written by Jerry Aguirre <jerry@ATC.Olivetti.Com>.
SEE ALSO
newsfeeds(5), innd(8).
CROSSPOST(8)