Very true, which is why I included "but then you have to collect metrics to compute and need a way to respond to requests to see the average." The implied addition was that the term average was defined at some referenced location and itself could be retrieved.
BTW, where in the RA will this discussion be captured and is anybody signed up to do it? I'd very much like to pass this around during discussions at MITRE.
Ken
On Mar 5, 2007, at 5:22 AM, Poulin, Michael wrote:
So, the general rule for an SLA is that all listed characteristics out to be measurable. In this case, a statement like "you will get 20% faster response than the average" is not a legitimate SLA statement because it cannot be measured - the 'average' value is not defined, i.e. the statement is not self-contained or complete. Actually, the latter note may be the second requirement to an SLA. Moreover, a statement like "an average response time should be less than 3 seconds" is also not self-contained because 'average' value changes over the time; so, the proper statement has to sound like "an average response time should be less than 3 seconds for any 30 minutes after first 1 hour of work".
I agree with Duane, nobody would care how long a response time is if "it is within an acceptable amount of time"... after "an acceptable amount of time" is clearly defined.
- Michael
From: Ken Laskey [mailto:]
Sent: 28 February 2007 02:59
To: Duane Nickull
Cc: Poulin, Michael;
Subject: Re: [soa-rm-ra] RE: updated service description
I more align with Duane's thinking here. The SLA has to specify what is being measured and what is the standard or limit against which it is being compared. An SLA can say you will get 20% faster response than the average, but then you have to collect metrics to compute and need a way to respond to requests to see the average.
Ken
On Feb 27, 2007, at 12:07 PM, Duane Nickull wrote:
But if you cannot see the response time of the other service requests for comparison, how can the consumer surmise they got the SLA they are assured of? Should they really care? As long as it is within an acceptable amount of time, it shouldn’t matter.
Duane
On 2/27/07 5:42 AM, "Poulin, Michael" <> wrote:
The "priority over al others" - like silver/gold user types - may be exposed as a 'response time', which can be easily monitored and measured. Moreover, why a SLA has to contain only run-time characteristics? It may contain a 'change control response time' (in days) as well, right?
The difference between SLA and Service Contract may be viewed as Policy section that cannot be a part of SLA.
- Michael
Fidelity Investments - Web Delivery
Phone: +44-173-783-6038
E-mail: <BLOCKED::">mailto:>
Web: http://www.fidelity.co.uk/ <BLOCKED::">http://www.fidelity.co.uk/>
Important: Fidelity Investments International, Fidelity Investment Services Limited, Fidelity Pensions Management and Financial Administration Services Limited (a Fidelity Group company) are all authorised and regulated in the UK by the Financial Services Authority and have their registered offices at Oakhill House, 130 Tonbridge Road, Hildenborough, Tonbridge, Kent TN11 9DZ. Tel 01732 361144. Fidelity only gives information on products and does not give investment advice to private clients based on individual circumstances. Any comments or statements made are not necessarily those of Fidelity. The information transmitted is intended only for the person or entity to which it is addressed and may contain confidential and/or privileged material. If you received this in error, please contact the sender and delete the material from any computer. All e-mails sent from or to Fidelity may be subject to our monitoring procedures. Direct link to Fidelity’s website - http://www.fidelity-international.com/world/index.html
From: Duane Nickull [mailto:]
Sent: 26 February 2007 18:51
To: Ken Laskey;
Cc:
Subject: Re: [soa-rm-ra] RE: updated service description
SLA can only encompass externally visible facets of a services operations, otherwise it is more or less moot. Why would someone even contemplate saying “your service request will be treated with priority over al others” if there is no way to verify or authenticate from an external standpoint? SLA’s should really just stick to things like timeouts, error reporting and other externally visible behaviors.
My $0.02.
D
On 2/25/07 2:21 PM, "Ken Laskey" <> wrote:
Michael,
I started to respond to this hours ago, then printed out your paper, and finally make it past numerous distractions (including a unexpected 3 inches of snow to shovel) to finally collect my thoughts.
On your first point, I think of the service description as being the executive summary and table of contents. It gives you the critical overview and tells you where to find more details. So while Interaction Policies & Contracts in the description may explicitly lay out such things, these are more likely to be fully defined elsewhere and referenced from the description. Of course, one benefit of this is that the policy, for example, becomes reusable.
As for the personal nature of policy agreements, if it is a one-time agreement, I see it residing in the execution context (whose concrete manifestation is still to be discussed). If the agreement is intended to be reused, it should be catalogued somewhere. The idea of a general purpose somewhere also needs to be tackled, perhaps (definitely?) as part of the RA.
I then read your paper and if you went through the TC emails, you could find discussions touching on but never resolving many of these issues. So here are some general questions and some specifics ones relating to your paper.
- What is a service component? If we have a composite service, i.e. one that shows a single interface but is actually a combination of other services, do you consider the constituent services to be service components? In the RM, we distinguish between the capability and the service that provides access to the capability. Is an implementation of the capability also a service component? Personally, I think I'd say yes to both where changes to either can affect the client experience.
- Are "service functions" the same as RM actions?
- One of my favorites: what belongs in a SLA? Some views have it so complete that it basically becomes the service description, i.e. it describes functionality, constraints, interface details, and the birthdays of anyone who ever used it. Personally, I take a more minimalist approach and think of SLA as being strictly a set of well defined metrics against which compliance can be measured. I agree with with your line that "a SOA service ... is invoked ... when the client is fully aware of the service behavior and conditions." (We could quibble over "fully".) To accomplish this, I rely on an adequate service description, of which a metrics-laden SLA is a part. However, another part is an indication of how one can obtain metrics collected against the service. It is then up to the client to compare the collected metrics with the SLA and determine fitness for use. For example, if I have a service that says I will relieve 80% of world hunger and the metrics show me only relieving 70%, does my service provide value?
Enough for now.
Ken
On Feb 25, 2007, at 10:00 AM, wrote:
Hi Ken,
I have two notes to the update (actually, to the doc as it is with the update):
1) There is an entity in the Service Description diagram named "Interaction Policies & Contracts". I think that Contract belongs to another area than service description because it is 'personalized' based on consumer/provider agreement. Yes, logically, a Contract is a special type of description but it would be more comprehensive if we put it into Interacting with Services Model section.
2) You have mentioned <A discussion of how version designation/value should be impacted by change of version of any of its components (or any descriptions?) still needs to be discussed.> To me, a version of the service is one of very important characteristics of the service description (and, respectively, of the service contract). I would propose placing version as an entity in the Service Description diagram either under Identity or at the level of Service Description entity because version affects Service Reachability, Service Functionality, Interaction Policies, and Service Interface entities. I have published a few thoughts about SOA service versioning in SOA-WebServices Journal (http://java.sys-con.com/read/250503.htm ).
- Michael
------------------------------------------------------------------------------------------
Ken Laskey
MITRE Corporation, M/S H305 phone: 703-983-7934
7515 Colshire Drive fax: 703-983-1379
McLean VA 22102-7508
--
**********************************************************
Sr. Technical Evangelist - Adobe Systems, Inc. *
Chair - OASIS SOA Reference Model Technical Committee *
Blog: http://technoracle.blogspot.com *
Music: http://www.mix2r.com/audio/by/artist/duane_nickull*
**********************************************************
--
**********************************************************
Sr. Technical Evangelist - Adobe Systems, Inc. *
Chair - OASIS SOA Reference Model Technical Committee *
Blog: http://technoracle.blogspot.com *
Music: http://www.mix2r.com/audio/by/artist/duane_nickull*
**********************************************************
------------------------------------------------------------------------------------------
Ken Laskey
MITRE Corporation, M/S H305 phone: 703-983-7934
7515 Colshire Drive fax: 703-983-1379
McLean VA 22102-7508
------------------------------------------------------------------------------------------
Ken Laskey
MITRE Corporation, M/S H305 phone: 703-983-7934
7515 Colshire Drive fax: 703-983-1379
McLean VA 22102-7508
