Re: [Netslices] Reliability & Availability in NetSling

 From a tenant's perspective, which arguably is the most important 
perspective the key characteristics of interest is  surely "The Mean 
Time Between Failures (MTBF) to maintain a defined QOS requirement." 
together with the duration of those periods.

With that information you can the viability of your application over the 
slice, and the impact on your business.

- Stewart

On 11/08/2017 08:47, qiangli (D) wrote:
>
> Hi All,
>
> The phrase “working as expected” inspired me. For a network slice, 
> what people expect is not only the stable working probability/time, 
> but also the guaranteed SLA.
>
> Try to consider from tenant’s perspective, I would be very concerned 
> about how much the promised SLA could really be achieved. I am not 
> sure which one (availability/reliability) this concept should belong 
> to according to ITU’s definitions.
>
> Best regards,
>
> Cristina QIANG
>
> *From:*Netslices [mailto:netslices-bounces@ietf.org] *On Behalf Of 
> *Stewart Bryant
> *Sent:* Thursday, August 10, 2017 5:58 PM
> *To:* netslices@ietf.org
> *Subject:* Re: [Netslices] Reliability & Availability in NetSling
>
>
> In general at IETF we associate reliable with the probability of 
> packet delivery, and a reliable transport protocol is one that retries 
> (at a cost of delay) until the packet is delivered or other factors 
> intervene. Availability tends to refer to the ability of the network 
> to receive packets for attempted delivery.
>
> I looked for a formal definition in IPPM which is the WG that is 
> concerned with measuring such things but could not find a definition.
>
> From the ITU we can find:
>
> *Term* : availability (performance)
>
> *Definition* : The ability of an item to be in a state to perform a 
> required function under given conditions at a given instant of time or 
> over a given time interval, assuming that the required external 
> resources are provided.
> NOTE 1 – This ability depends on the combined aspects of the 
> reliability performance, the maintainability performance and the 
> maintenance support performance.
> NOTE 2 – Required external resources, other than maintenance resources 
> do not affect the availability performance of the item.
>
> *Term* : reliability (performance)
>
> *Definition* : The ability of an item to perform a required function 
> under given conditions for a given time interval.
> NOTE 1 – It is generally assumed that the item is in a state to 
> perform this required function at the beginning of the time interval.
> NOTE 2 – Generally, reliability performance is quantified using 
> appropriate measures. In some applications, these measures include an 
> expression of reliability performance as a probability, which is also 
> called reliability.
>
>
> or
>
> *Term* : reliability characteristic
>
> *Definition* : The Mean Time Between Failures (MTBF) to maintain a 
> defined QOS requirement.
>
>
> - Stewart
>
> On 10/08/2017 05:42, Pedro Martinez-Julia wrote:
>
>     Dear all,
>
>     I personally think it is better to use the industrial definitions but,
>
>     being purist, we can find important differences between them. While
>
>     availability is the probability for a system to work as expected
>     in some
>
>     period of time (99.999% of time), reliability is a broader term that
>
>     refers to the different situations in which a system will be able to
>
>     overcome without breaking. In some cases, the latter can
>     incorporate the
>
>     former, but not in all of them.
>
>     For network slicing we can keep the definition commonly used by
>     industry
>
>     with the necessary details to make clear the aspects that
>     differentiate
>
>     them. I would keep "working as expected for some period of time"
>     related
>
>     to availability and "resistant to disparate situations" to
>     reliability.
>
>     Regards,
>
>     Pedro
>
>     On Thu, Aug 10, 2017 at 02:44:09AM +0000, GENG Liang wrote:
>
>         Hi Cristina,
>
>         Interestingly we were discussing this confusion with few
>         industrial partners recently. In telecommunication language we
>         normally use "Reliability" to refer the probability a network
>         is stably run (i.e. 99.999% of time). This is also regarded as
>         network "Availability". However, "Reliability" in industrial
>         verticals is more comprehensive - including not only network
>         availability parameter but also mechanics, electricity etc.
>
>         Personally I think, network slicing is still looking at
>         network regime where I believe Reliability means the
>         percentage of time a connection is available. But we you sell
>         this concept to industrial verticals, they may think differently.
>
>         ________________________________
>
>         Liang GENG
>
>         China Mobile Research Institute
>
>         From: qiangli (D)<mailto:qiangli3@huawei.com>
>         <mailto:qiangli3@huawei.com>
>
>         Date: 2017-08-10 10:04
>
>         To: netslices@ietf.org
>         <mailto:netslices@ietf.org><mailto:netslices@ietf.org>
>         <mailto:netslices@ietf.org>
>
>         Subject: [Netslices] Reliability & Availability in NetSling
>
>         Hi All,
>
>         I was confused when I was reading some NetSlicing related
>         materials. It seems that “Reliability” supported by Netslicing
>         refers to the probability that a network slice could work
>         stably, or other similar metrics. But, shouldn’t this be the
>         defination of “Availability”? Then what does reliability mean
>         in NetSlicing?
>
>         Best regards,
>
>         Cristina QIANG
>
>         _______________________________________________
>
>         Netslices mailing list
>
>         Netslices@ietf.org <mailto:Netslices@ietf.org>
>
>         https://www.ietf.org/mailman/listinfo/netslices
>