Measuring hosting performance: shared <--> vps <--> dedicate

Rails
1

Every app we build has to be deployed, and the price has to be right.

  • dedicated servers start at 99$/m (Dreamhost, …), and
    shared hosting go up to 47$/m (PlanetArgon) or 60$/m (TextDrive)
  • Railsplayground: 2$/m
  • TextDrive dedicated servers : 1000$/m

Is there an objective way/tool to measure, test and compare the
performance of the hosters?

Alain

2

No, not really…

I think somebody has tried, search the archive…

anyways…serverpronto.com does dedicated for 30$…been using them since
november 2005 without any problems…i signed up to see what the catch
could be…so far there hasnt been one…you need to know how to admin a
server though…

Mikkel B.

www.strongside.dk - Football Portal(DK)
nflfeed.helenius.org - Football News(DK)
ting.minline.dk - Buy Old Stuff!(DK)

3

Mikkel

i signed up to see what the catch …you need to know how to admin
a …

This thread is not about ‘catch’, reputation, or service level, it’s
about sheer performance, as a number.

To oversimplify:

"I tested my app., and it consumes
x cpu units/hour, on average
y db units/hour, on average
z bandwidth units/hour, on average
.
What hosting plans fit those needs?
"

From there, you start checking prices, service levels, reputation, ease
of installation, guarantees, etc, etc… and then you make your choice.

Alain

4

Is there an objective way/tool to measure, test and compare the
performance of the hosters?

IIRC there are some forums where people posts benchmark stats etc, but
your best bet is probably word of mouth from people running live
mission-critical apps. If you really want stability and performance,
dedicated is really the only option in my opinion. None of that
managed or cpanel crud either. Your best bet would be to either get
an unmanaged box and run it ALL on your own, or a managed from
textdrive since they are the only ones I’d trust. Dreamhost and
PlanetArgon are probably great as well, Ive just never tried them
myself. I have an unmanaged at layeredtech.com and its been my best
server to date. I never had anything close to comparable with shared
or vps. Seems like they’ve got a lightning fast network. Remember
though, you’d be 100% responsible for
troubleshooting/installing/maintaining everything. If your ok with
that then you cant beat it.

Joe

5

Rimu all the way! True VPS (running RHEL!) for like $20/mo, and
dedicated servers for like $160 or so. Plus they don’t mind newb
questions (like why the eff isn’t my subdomain working? Oh, I see,
incorrect permission on public/!).

  • Rabbit
6

On Feb 15, 2006, at 2:13 AM, Alain R. wrote:

Alain

Rails mailing list
[email protected]
http://lists.rubyonrails.org/mailman/listinfo/rails

Alain-

Maybe using the depo app from the agile rails book would be a good

benchmark app. Its widely available and well known by most people.
And its simple enough but has enough features to get a decent
benchmark. I can run it on a rimu vps and a layeredtech dedicated
server.

And by the way, you can get a nice dedicated server from layeredtech

for $69/month. Athlon 3000+, 80gig hd and 512Gig ram.

Cheers-
-Ezra Z.
WebMaster
Yakima Herald-Republic Newspaper
[email protected]
509-577-7732

7

Ezra

  > Maybe the depo app .. would be a good benchmark app. ...
  > I can run it on a rimu vps and a layeredtech dedicated server.

The ideal solution would be :

Once :

1/ Design a “typical” and standard web application.
That’s the hard part, but anything is better than what we have now.
(This client web. app will be installed on each host server to test)

2/ Design a test script that:
- exercises the app and simulates a chosen load/number of users.
- collects and stores the perf. results in a unique DB
This will run on 1 unique master machine, and test ALL the
monitored servers.

Once for each target server :

3/ Install the app on each target server/host.
=> Know/find 1 person with an account on each target host/server.
(On Textdrive, some machines are known to be less reliable than
others.).

4/ register the target url in the master server

and finally, in the master

Every hour, for each server :

5/ run the script and collect the results

==> The accumulated results would be displayed in a big matrix.

That would be a starting point, but it would be damn useful.

Alain