Re: AWC EC2 queries vs self hosting and other options for a email services

Hi, On 03/01/18 09:09, Paul van den Bergen via luv-main wrote:
OK. As of reinvent 2017, AWS introduced bare metal server pricing. As far as I understand it, if the hardware breaks, you get a new machine, but it's up to you to manage DR.
dedicated instances - it's a long term contract for a VM (not a bare metal machine)
on demand - you pay per hour.
Typically you want a mailserver running 24/7, so about 750 hours every month.
spot price - you pay "bid" for a low priced VM. If someone outbids you, you lose the VM.
Do you mean, you have a box, working... then you get outbid and lose a working box? Kind Regards AndrewM

spot price means if you get out bid partway through your hour, you lose that VM and anything running on it (I think there is an option to dump to a s3 bucket or some such - but regardless) the price is usually so cheap that it's perfect for throwaway computational tasks - hadoop, grid (keep track of the state of your returns independantly), etc, and so on. Also, IIUC it only hits you when demand is larger than supply. On 3 January 2018 at 15:17, Andrew McGlashan via luv-main < luv-main@luv.asn.au> wrote:
Hi,
On 03/01/18 09:09, Paul van den Bergen via luv-main wrote:
OK. As of reinvent 2017, AWS introduced bare metal server pricing. As far as I understand it, if the hardware breaks, you get a new machine, but it's up to you to manage DR.
dedicated instances - it's a long term contract for a VM (not a bare metal machine)
on demand - you pay per hour.
Typically you want a mailserver running 24/7, so about 750 hours every month.
spot price - you pay "bid" for a low priced VM. If someone outbids you, you lose the VM.
Do you mean, you have a box, working... then you get outbid and lose a working box?
Kind Regards AndrewM
_______________________________________________ luv-main mailing list luv-main@luv.asn.au https://lists.luv.asn.au/cgi-bin/mailman/listinfo/luv-main
-- Dr Paul van den Bergen

hey folks, because I'm actively reviewing this and my brain is full I made a mistake in terminology before... Reserved Instance - a long term VM contract Dedicated Host - bare metal server apologies for the confusion On 3 January 2018 at 15:54, Paul van den Bergen <paul.vandenbergen@gmail.com
wrote:
spot price means if you get out bid partway through your hour, you lose that VM and anything running on it (I think there is an option to dump to a s3 bucket or some such - but regardless)
the price is usually so cheap that it's perfect for throwaway computational tasks - hadoop, grid (keep track of the state of your returns independantly), etc, and so on. Also, IIUC it only hits you when demand is larger than supply.
On 3 January 2018 at 15:17, Andrew McGlashan via luv-main < luv-main@luv.asn.au> wrote:
Hi,
On 03/01/18 09:09, Paul van den Bergen via luv-main wrote:
OK. As of reinvent 2017, AWS introduced bare metal server pricing. As far as I understand it, if the hardware breaks, you get a new machine, but it's up to you to manage DR.
dedicated instances - it's a long term contract for a VM (not a bare metal machine)
on demand - you pay per hour.
Typically you want a mailserver running 24/7, so about 750 hours every month.
spot price - you pay "bid" for a low priced VM. If someone outbids you, you lose the VM.
Do you mean, you have a box, working... then you get outbid and lose a working box?
Kind Regards AndrewM
_______________________________________________ luv-main mailing list luv-main@luv.asn.au https://lists.luv.asn.au/cgi-bin/mailman/listinfo/luv-main
-- Dr Paul van den Bergen
-- Dr Paul van den Bergen

On Wednesday, 3 January 2018 3:54:34 PM AEDT Paul van den Bergen via luv-main wrote:
spot price means if you get out bid partway through your hour, you lose that VM and anything running on it (I think there is an option to dump to a s3 bucket or some such - but regardless)
Some time ago we had a LUV lecture about using Amazon as part of a HPC system. The idea was to take advantage of spot prices and similar deals to get a lower cost for the work done than owning systems, espectially for the case of systems that aren't busy 365 days a year. Maybe we should track down the speaker and ask if they want to reprise that talk? Ongoing changes in costs of ownership and costs of EC2 as well as changes to the types of jobs run will make some changes to the most effective ways of running such things. -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/
participants (3)
-
Andrew McGlashan
-
Paul van den Bergen
-
Russell Coker