About clock synchronization in the cloud
Dear all, I was wondering whether the features of the system clock of a virtual instance (compute or storage) may be of interest for the user. I wrote down 500 words on the topic (see Google doc below), and I'd be pleased to know your opinions, taking advantage of the intense "brain storm" during next OGF. https://docs.google.com/document/pub?id=1imafqyrVscdTbdfjngJ4Xfbg1DIoebdyzyQ... Bye Augusto
Sounds like a good discussion topic! It might be a little out of scope for OCCI, but who knows where the discussion might lead to or what it might highlight. Andy From: occi-wg-bounces@ogf.org [mailto:occi-wg-bounces@ogf.org] On Behalf Of Augusto Ciuffoletti Sent: Monday, September 19, 2011 9:33 AM To: occi-wg@ogf.org Subject: [occi-wg] About clock synchronization in the cloud Dear all, I was wondering whether the features of the system clock of a virtual instance (compute or storage) may be of interest for the user. I wrote down 500 words on the topic (see Google doc below), and I'd be pleased to know your opinions, taking advantage of the intense "brain storm" during next OGF. https://docs.google.com/document/pub?id=1imafqyrVscdTbdfjngJ4Xfbg1DIoebdyzyQ... Bye Augusto ------------------------------------------------------------- Intel Ireland Limited (Branch) Collinstown Industrial Park, Leixlip, County Kildare, Ireland Registered Number: E902934 This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
Clocks are another type of virtualized hardware resource... On 9/19/2011 2:39 PM, Edmonds, AndrewX wrote:
Sounds like a good discussion topic! It might be a little out of scope for OCCI, but who knows where the discussion might lead to or what it might highlight.
Andy
*From:*occi-wg-bounces@ogf.org [mailto:occi-wg-bounces@ogf.org] *On Behalf Of *Augusto Ciuffoletti *Sent:* Monday, September 19, 2011 9:33 AM *To:* occi-wg@ogf.org *Subject:* [occi-wg] About clock synchronization in the cloud
Dear all,
I was wondering whether the features of the system clock of a virtual instance (compute or storage) may be of interest for the user. I wrote down 500 words on the topic (see Google doc below), and I'd be pleased to know your opinions, taking advantage of the intense "brain storm" during next OGF.
https://docs.google.com/document/pub?id=1imafqyrVscdTbdfjngJ4Xfbg1DIoebdyzyQ...
Bye
Augusto
------------------------------------------------------------- Intel Ireland Limited (Branch) Collinstown Industrial Park, Leixlip, County Kildare, Ireland Registered Number: E902934
This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
_______________________________________________ occi-wg mailing list occi-wg@ogf.org http://www.ogf.org/mailman/listinfo/occi-wg
I agree that the present trend of OCCI is not on resource features: OCCI gives a way to define and "bind" resource types, and gives the provider the freedom to decide such features. But the present model is influenced by the original, resource specific model. A discussion about a feature not yet figured out before the generalization process may be of interest, after all. I'm not sure if a clock can be considered as a resource in the OCCI way: as is, I see the clock more as a feature (attribute?) of (certain) resources. Like a processor speed for a compute resource. However, a master clock might be considered as a resource by itself, and other resources could be bound with OCCI links to indicate synchronization. 2011/9/20 Gary Mazz <garymazzaferro@gmail.com>
Clocks are another type of virtualized hardware resource...
On 9/19/2011 2:39 PM, Edmonds, AndrewX wrote:
Sounds like a good discussion topic! It might be a little out of scope for OCCI, but who knows where the discussion might lead to or what it might highlight.****
** **
Andy****
** **
*From:* occi-wg-bounces@ogf.org [mailto:occi-wg-bounces@ogf.org<occi-wg-bounces@ogf.org>] *On Behalf Of *Augusto Ciuffoletti *Sent:* Monday, September 19, 2011 9:33 AM *To:* occi-wg@ogf.org *Subject:* [occi-wg] About clock synchronization in the cloud****
** **
Dear all,****
** **
I was wondering whether the features of the system clock of a virtual instance (compute or storage) may be of interest for the user. I wrote down 500 words on the topic (see Google doc below), and I'd be pleased to know your opinions, taking advantage of the intense "brain storm" during next OGF.****
** **
https://docs.google.com/document/pub?id=1imafqyrVscdTbdfjngJ4Xfbg1DIoebdyzyQ... ****
** **
Bye****
** **
Augusto****
** **
------------------------------------------------------------- Intel Ireland Limited (Branch) Collinstown Industrial Park, Leixlip, County Kildare, Ireland Registered Number: E902934
This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
_______________________________________________ occi-wg mailing listocci-wg@ogf.orghttp://www.ogf.org/mailman/listinfo/occi-wg
_______________________________________________ occi-wg mailing list occi-wg@ogf.org http://www.ogf.org/mailman/listinfo/occi-wg
-- Augusto Ciuffoletti Via Milano 118 38122 - Trento
From my point of view, this is out of scope. Getting the hardware clock "right" is a job that has been solved on the operating system level, using Distributed Systems protocols like NTP. I agree that, for many services, exact time is crucial (try to deliver a PDC with a broken clock…), but this would be either something that a) may be part of an operating system template mechanism, where you know that such basic services are available or b) is ensured through quality of service parameters (as part of the contract or SLA between consumer and provider) Making it an attribute seems to be the better of the two options for OCCI, but it still doesn't make that much sense -- what does an attribute on the clock (which way ever this attribute is structured) tell me other than that the clock is there and in sync (with a drift of x and skew of y) at any time, and why would I want to discover this information via the OCCI interface? Just my two cents… -Alexander Am 19.09.2011 um 22:39 schrieb Edmonds, AndrewX:
Sounds like a good discussion topic! It might be a little out of scope for OCCI, but who knows where the discussion might lead to or what it might highlight.
Andy
From: occi-wg-bounces@ogf.org [mailto:occi-wg-bounces@ogf.org] On Behalf Of Augusto Ciuffoletti Sent: Monday, September 19, 2011 9:33 AM To: occi-wg@ogf.org Subject: [occi-wg] About clock synchronization in the cloud
Dear all,
I was wondering whether the features of the system clock of a virtual instance (compute or storage) may be of interest for the user. I wrote down 500 words on the topic (see Google doc below), and I'd be pleased to know your opinions, taking advantage of the intense "brain storm" during next OGF.
https://docs.google.com/document/pub?id=1imafqyrVscdTbdfjngJ4Xfbg1DIoebdyzyQ...
Bye
Augusto
------------------------------------------------------------- Intel Ireland Limited (Branch) Collinstown Industrial Park, Leixlip, County Kildare, Ireland Registered Number: E902934
This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies. _______________________________________________ occi-wg mailing list occi-wg@ogf.org http://www.ogf.org/mailman/listinfo/occi-wg
-- Alexander Papaspyrou alexander.papaspyrou@tu-dortmund.de
I think Alex hits this on the nail by noting "as part of the [...] SLA", which would be then related to the provisioning of the compute resources but not necessarily to the current OCCI spec where consideration for SLAs is not (yet?) present. Andy -----Original Message----- From: alexander.papaspyrou@tu-dortmund.de [mailto:alexander.papaspyrou@tu-dortmund.de] Sent: Tuesday, September 20, 2011 11:27 AM To: Edmonds, AndrewX Cc: augusto.ciuffoletti@gmail.com; occi-wg@ogf.org Subject: Re: [occi-wg] About clock synchronization in the cloud
From my point of view, this is out of scope. Getting the hardware clock "right" is a job that has been solved on the operating system level, using Distributed Systems protocols like NTP.
I agree that, for many services, exact time is crucial (try to deliver a PDC with a broken clock.), but this would be either something that a) may be part of an operating system template mechanism, where you know that such basic services are available or b) is ensured through quality of service parameters (as part of the contract or SLA between consumer and provider) Making it an attribute seems to be the better of the two options for OCCI, but it still doesn't make that much sense -- what does an attribute on the clock (which way ever this attribute is structured) tell me other than that the clock is there and in sync (with a drift of x and skew of y) at any time, and why would I want to discover this information via the OCCI interface? Just my two cents. -Alexander Am 19.09.2011 um 22:39 schrieb Edmonds, AndrewX:
Sounds like a good discussion topic! It might be a little out of scope for OCCI, but who knows where the discussion might lead to or what it might highlight.
Andy
From: occi-wg-bounces@ogf.org [mailto:occi-wg-bounces@ogf.org] On Behalf Of Augusto Ciuffoletti Sent: Monday, September 19, 2011 9:33 AM To: occi-wg@ogf.org Subject: [occi-wg] About clock synchronization in the cloud
Dear all,
I was wondering whether the features of the system clock of a virtual instance (compute or storage) may be of interest for the user. I wrote down 500 words on the topic (see Google doc below), and I'd be pleased to know your opinions, taking advantage of the intense "brain storm" during next OGF.
https://docs.google.com/document/pub?id=1imafqyrVscdTbdfjngJ4Xfbg1DIoe bdyzyQSPmDgOCY
Bye
Augusto
------------------------------------------------------------- Intel Ireland Limited (Branch) Collinstown Industrial Park, Leixlip, County Kildare, Ireland Registered Number: E902934
This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies. _______________________________________________ occi-wg mailing list occi-wg@ogf.org http://www.ogf.org/mailman/listinfo/occi-wg
-- Alexander Papaspyrou alexander.papaspyrou@tu-dortmund.de ------------------------------------------------------------- Intel Ireland Limited (Branch) Collinstown Industrial Park, Leixlip, County Kildare, Ireland Registered Number: E902934 This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
Thank you for your replies, first... Here my point is that clock synchronization is irrelevant in many (possibly all) current use cases, but is essential for others that are just potential. The user that hits such a use case is unable to implement clock synchronization in the cloud (as far as I know), so he/she may be happy to pay an extra to have a clock within 1 second from UTC on certain VMs, and add a new use case. The simplest idea that comes to my mind (maybe wrong) is to provide clock specs through a mixin. The boundary of "what is subject to a SLA" is subtle. But a feature included in a SLA is somewhat static, and does not (necessarily) go through the user interface. Clock synchronization is a switch you can set on/off, and that requires specific actions on the side of the IaaS provider. I would be happy to know if anybody experienced the case where a lack of synchronization in the cloud was a problem. If only I'd know what PDC is ... Augusto 2011/9/20 Edmonds, AndrewX <andrewx.edmonds@intel.com>
I think Alex hits this on the nail by noting "as part of the [...] SLA", which would be then related to the provisioning of the compute resources but not necessarily to the current OCCI spec where consideration for SLAs is not (yet?) present.
Andy
-----Original Message----- From: alexander.papaspyrou@tu-dortmund.de [mailto:alexander.papaspyrou@tu-dortmund.de] Sent: Tuesday, September 20, 2011 11:27 AM To: Edmonds, AndrewX Cc: augusto.ciuffoletti@gmail.com; occi-wg@ogf.org Subject: Re: [occi-wg] About clock synchronization in the cloud
From my point of view, this is out of scope. Getting the hardware clock "right" is a job that has been solved on the operating system level, using Distributed Systems protocols like NTP.
I agree that, for many services, exact time is crucial (try to deliver a PDC with a broken clock.), but this would be either something that
a) may be part of an operating system template mechanism, where you know that such basic services are available or b) is ensured through quality of service parameters (as part of the contract or SLA between consumer and provider)
Making it an attribute seems to be the better of the two options for OCCI, but it still doesn't make that much sense -- what does an attribute on the clock (which way ever this attribute is structured) tell me other than that the clock is there and in sync (with a drift of x and skew of y) at any time, and why would I want to discover this information via the OCCI interface?
Just my two cents. -Alexander
Am 19.09.2011 um 22:39 schrieb Edmonds, AndrewX:
Sounds like a good discussion topic! It might be a little out of scope for OCCI, but who knows where the discussion might lead to or what it might highlight.
Andy
From: occi-wg-bounces@ogf.org [mailto:occi-wg-bounces@ogf.org] On Behalf Of Augusto Ciuffoletti Sent: Monday, September 19, 2011 9:33 AM To: occi-wg@ogf.org Subject: [occi-wg] About clock synchronization in the cloud
Dear all,
I was wondering whether the features of the system clock of a virtual instance (compute or storage) may be of interest for the user. I wrote down 500 words on the topic (see Google doc below), and I'd be pleased to know your opinions, taking advantage of the intense "brain storm" during next OGF.
https://docs.google.com/document/pub?id=1imafqyrVscdTbdfjngJ4Xfbg1DIoe bdyzyQSPmDgOCY
Bye
Augusto
------------------------------------------------------------- Intel Ireland Limited (Branch) Collinstown Industrial Park, Leixlip, County Kildare, Ireland Registered Number: E902934
This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies. _______________________________________________ occi-wg mailing list occi-wg@ogf.org http://www.ogf.org/mailman/listinfo/occi-wg
-- Alexander Papaspyrou alexander.papaspyrou@tu-dortmund.de
------------------------------------------------------------- Intel Ireland Limited (Branch) Collinstown Industrial Park, Leixlip, County Kildare, Ireland Registered Number: E902934
This e-mail and any attachments may contain confidential material for the sole use of the intended recipient(s). Any review or distribution by others is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
-- Augusto Ciuffoletti Via Milano 118 38122 - Trento
participants (4)
-
alexander.papaspyrou＠tu-dortmund.de -
Augusto Ciuffoletti -
Edmonds, AndrewX -
Gary Mazz