You are viewing a plain text version of this content. The canonical link for it is here.
Posted to modperl@perl.apache.org by Josh Chamas <jo...@chamas.com> on 2002/09/14 02:32:35 UTC
[OT] Survey for VMWare server hosting of mod_perl, linux, + apache
Dear mod_perl users, [ SURVEY BELOW, please respond privately ]
This is the *2nd survey* that I am sending to the list
about a possible business venture where we provide dedicated
hosting services specific to the needs of the mod_perl user community.
What we got from the last survey was the need for less expensive
and more reliable dedicated server options than were being offered.
However we are not interested in offering basic commodity dedicated
servers, companies like dellhost.com & rackspace.com do it well
enough already.
What we would like to do instead is offer a VMWare style
Virtual Machine slice of a high end ($12K+) x86 server configured for
high availability with RAID-10 SCSI hard drives & redundant power/fans.
This will let us not only minimize system failures for the customer,
but also take care of installations & upgrades without the usual
delays & downtime associated with customizing physical dedicated servers.
Don't worry, the survey is coming, but this is a new type of product
offering, so I'll elaborate some more... the basic computing unit
that we will sell will be fixed at first. Since its a slice of a
VMWare ESX server, we will be able to offer up to 16 power slices
bundled as one logical server. The basic pricing would
be as follows:
VM POWER UNIT PRICE:
Storage: 4G
Memory: 100M
CPU: 100Mhz P4 Xeon x86
Bandwidth: 100Kbs
IP Address: 1
Price: $150 / month
Additional bandwidth pricing: $50 / 100Kbs.
Pricing includes the following services:
RedHat Linux 7.x installed
Free Installation & Service Changes
Hardware RAID-10 SCSI disk subsystem
Remote Console Access
Full root level control of your server, its not a shared virtual server!
Primary &/or secondary DNS for up to 5 domains
Basic port monitoring
Redundant power supplies & fans in base hardware
Month-to-month contract
Basic support for configuring your installed apache & sendmail software
As an example, a 2xVM UNIT server at $300/mo. would have
8G SCSI/RAID storage, 200M memory, 200 Mhz Xeon CPU,
200Kbs bandwidth, and 2 IP addresses. For an additional
$150/mo., the bandwidth could be increased to 500Kbs.
==========================================
SURVEY HERE
==========================================
. Would you buy this service?
. In what time frame would you buy this service?
. How many VM POWER UNITS would you purchase for your server?
. How many of these logical servers would you buy?
. If you feel that the price per power unit is wrong,
what would you suggest instead of $150 USD?
. If you feel that the capacity per power unit is wrong,
what should be changed for the same price? Please note that
at least at first, we will need to sell all capacities as one unit.
Storage: 4G ->
Memory: 100M ->
CPU: 100Mhz P4 Xeon x86 ->
Bandwidth: 100Kbs ->
IP Address: 1 ->
. Would you need email fully hosted by us for an additional fee
on a separate system?
. Would you need a off system backup solution such as tape,
or remote disk for an additional fee?
. What Service Level Agreement (SLA) do you expect for the network?
None, 99.9%, 99.99%, 99.999%
. What Service Level Agreement (SLA) do you expect for your system(s)?
None, 99.9%, 99.99%, 99.999%
. Is there anything else you would like us to consider when developing
this service?
================================
SURVEY DONE
================================
Many thanks if you came this far. Free (--your favorite drink--)
should we ever cross paths for your interest :)
Regards,
Josh
________________________________________________________________
Josh Chamas, Founder phone:925-552-0128
Chamas Enterprises Inc. http://www.chamas.com
NodeWorks Link Checking http://www.nodeworks.com
Re: [OT] Survey for VMWare server hosting of mod_perl, linux, +
Posted by Steven Lembark <le...@wrkhors.com>.
> BTW, 99.999% uptime over a year means 5 minutes, 15 seconds TOTAL
> downtime - for the *entire year*. Apply that to a month, and you're
> talking a tad over 26 seconds. It should be obvious why five 9s is so
> expensive to attain.
One advantage to attempting it on VM's, however, is that
the uptime can be on the same "machine" running on multiple
pieces of hardware. Duplicate copies of the VM with failover
isn't hard to manage with VMWare and multiple units. Short
of a shotgun, bomb, or earthquake the system can be made to
fail onto the same machine; this is pretty hard to manage
with real machines.
--
Steven Lembark 2930 W. Palmer
Workhorse Computing Chicago, IL 60647
+1 800 762 1582
Re: [OT] Survey for VMWare server hosting of mod_perl, linux, +
Posted by Chris <ch...@prather.org>.
> We have a SUN E10k (8 CPUs & 20GB RAM) and it has already been down 3
> days this year - Only one day was attributable to the system -
> which SUN was happy to reimburse me for :P
>
> Five 9s is truly a fantasy; if for nothing but the "human" factor - a
> system is likely NOT to be up longer than a 8 or 9 months
>straight. Besides, how much downtime is allowable to the
> application system itself? Sometime people think that ust because the
> h/w is UP ALL NIGHT doesn't mean your application as well.
>
> my $cents = 2;
I would never hold a ISP accountable for application uptime. They shouldn't
care about anything but the service they provided you. In this specific
case that means the VMWare slice of the server should be up with some
guarnteed reliability.
5 nines is a dream. But (and especially) with a scheme of shared hardware
you need to have some kind of guarnteed h/w uptime, or personally I really
wouldn't trust it. It's not so much the guranteed 5 nines or whatever, it's
the idea that they've done alot of research and planning about how to deal
with a situation where you have 5 VMWare hosts running on a box, and one of
them starts pegging the processor, or something funky happens and fries
some critical bit of hardware. The more nines in the gurantee, the more
confident they should be in assessing the situation, and money-back says
they're gonna put money on it.
> On Saturday, September 14, 2002, at 02:38 PM, mike808 wrote:
>> BTW, 99.999% uptime over a year means 5 minutes, 15 seconds TOTAL
>> downtime -
>> for the *entire year*. Apply that to a month, and you're talking a
>> tad over
>> 26 seconds. It should be obvious why five 9s is so expensive to
>> attain.
-Chris
Re: [OT] Survey for VMWare server hosting of mod_perl, linux, +
Posted by iudicium ferat <sn...@bellsouth.net>.
We have a SUN E10k (8 CPUs & 20GB RAM) and it has already been down
3 days this year - Only one day was attributable to the system -
which SUN was happy to reimburse me for :P
Five 9s is truly a fantasy; if for nothing but the "human" factor -
a system is likely NOT to be up longer than a 8 or 9 months
straight. Besides, how much downtime is allowable to the
application system itself? Sometime people think that ust because
the h/w is UP ALL NIGHT doesn't mean your application as well.
my $cents = 2;
-Sx-
PS - I did have a SUN Blade that had been up over a year -
but I had forgotten it was there and no one else was using it... :-D
On Saturday, September 14, 2002, at 02:38 PM, mike808 wrote:
> BTW, 99.999% uptime over a year means 5 minutes, 15 seconds TOTAL
> downtime -
> for the *entire year*. Apply that to a month, and you're talking a
> tad over
> 26 seconds. It should be obvious why five 9s is so expensive to attain.
Re: [OT] Survey for VMWare server hosting of mod_perl, linux, +
Posted by mike808 <mi...@users.sourceforge.net>.
> On Sat, 14 Sep 2002, Chris wrote:
> > Rackspace.com offers 5 nines uptime, and starts refunding your money.
On Saturday 14 September 2002 10:19 am, Dave Rolsky wrote:
> How long do you think it'd take from the time I noticed a problem to the
> time a ticket was created? 5 minutes? 10? That'd be enough time, in and
> of itself, to respond to quite a number of problems.
>
> I'm sure they offer excellent service, and I'm sure Joshua's company would
> as well. But five 9's is still in the realm of fantasy, unless you have a
> nearly unlimited budget. Four 9's is pretty hopeful too, I think.
You could measure the uptime only between 9am and 5pm local time, like
Microsoft does to get your five 9's. They're on the "Nights and Weekends
Free" program, so read the fine print carefully on those "Five 9's" claims.
BTW, 99.999% uptime over a year means 5 minutes, 15 seconds TOTAL downtime -
for the *entire year*. Apply that to a month, and you're talking a tad over
26 seconds. It should be obvious why five 9s is so expensive to attain.
Mike808/
--
() Join the ASCII ribbon campaign against HTML email and Microsoft-specific
/\ attachments. If I wanted to read HTML, I would have visited your website!
Support open standards.
Re: [OT] Survey for VMWare server hosting of mod_perl, linux, +
Posted by Dave Rolsky <au...@urth.org>.
On Sat, 14 Sep 2002, Chris wrote:
> Rackspace.com offers 5 nines uptime, and starts refunding your money. It is
> a service differentiator that is used, and if you can hotswap drives and
> have really fast people running around you could perhaps maintain it.
>
> I've never actually been with a company that had the money to pay for that
> kind of hosting so I can't say how well it works.
Reading Rackspace's SLA, I see this key passage:
Network downtime exists when a particular customer is unable to transmit
and receive data and Rackspace records such failure in the Rackspace
trouble ticket system. Network downtime is measured from the time the
trouble ticket is opened by a customer to the time the server is once
again able to transmit and receive data.
How long do you think it'd take from the time I noticed a problem to the
time a ticket was created? 5 minutes? 10? That'd be enough time, in and
of itself, to respond to quite a number of problems.
I'm sure they offer excellent service, and I'm sure Joshua's company would
as well. But five 9's is still in the realm of fantasy, unless you have a
nearly unlimited budget. Four 9's is pretty hopeful too, I think.
-dave
/*==================
www.urth.org
we await the New Sun
==================*/
Re: [OT] Survey for VMWare server hosting of mod_perl, linux, +
Posted by Chris <ch...@prather.org>.
> 99.99% is a pipe dream. That's about 50 minutes _per year_. One
> reasonably bad incident could easily knock out a network for more than
> an hour. 99.999% is just funny. You'd _guarantee_ less than 5 minutes
> downtime per year? What do I get when you fail?
>
> Promising 99.999% only tells me you can't do math.
>
Rackspace.com offers 5 nines uptime, and starts refunding your money. It is
a service differentiator that is used, and if you can hotswap drives and
have really fast people running around you could perhaps maintain it.
I've never actually been with a company that had the money to pay for that
kind of hosting so I can't say how well it works.
-Chris
Re: [OT] Survey for VMWare server hosting of mod_perl, linux, +
apache
Posted by Dave Rolsky <au...@urth.org>.
On Fri, 13 Sep 2002, Josh Chamas wrote:
> ==========================================
> SURVEY HERE
> ==========================================
>
> . Would you buy this service?
Yes, if it were cheaper.
> . In what time frame would you buy this service?
Very soon (the next 1-2 months).
> . How many VM POWER UNITS would you purchase for your server?
1. Except ...
> . How many of these logical servers would you buy?
Huh/ What's the difference?
> . If you feel that the price per power unit is wrong,
> what would you suggest instead of $150 USD?
Given that I can go to rackshack.net and get a 1 Ghz Duron, 512 MB, 20GB
drive, and 400GB/month transfer for $99/month, this seems a bit expensive.
The big problem for me is the memory. 100MB for a mod_perl app (plus
database, etc.) is nothing. The disk space is a bit low as well.
> . If you feel that the capacity per power unit is wrong,
> what should be changed for the same price? Please note that
> at least at first, we will need to sell all capacities as one unit.
> Storage: 4G ->
> Memory: 100M ->
> CPU: 100Mhz P4 Xeon x86 ->
> Bandwidth: 100Kbs ->
> IP Address: 1 ->
More storage (maybe 6-10MB) and much more memory (500MB).
> . Would you need email fully hosted by us for an additional fee
> on a separate system?
No.
> . Would you need a off system backup solution such as tape,
> or remote disk for an additional fee?
That'd definitely be nice, but it depends on the fee. Doing a nightly
rsync or unison backup to another machine I admin isn't that big a deal.
> . What Service Level Agreement (SLA) do you expect for the network?
> None, 99.9%, 99.99%, 99.999%
99.99% is a pipe dream. That's about 50 minutes _per year_. One
reasonably bad incident could easily knock out a network for more than an
hour. 99.999% is just funny. You'd _guarantee_ less than 5 minutes
downtime per year? What do I get when you fail?
Promising 99.999% only tells me you can't do math.
> . What Service Level Agreement (SLA) do you expect for your system(s)?
> None, 99.9%, 99.99%, 99.999%
See above.
Seriously, I'd be interested in such a service but it just doesn't seem
that competitive. But maybe I'm not the target market. I'd be willing to
trade some uptime (i.e. no redundant fans/power/etc.) for a bit more
machine power.
-dave
/*==================
www.urth.org
we await the New Sun
==================*/