Information Technology Reference
In-Depth Information
Capabilities for SLA
specification
include support for standards-based specifica-
tion of SLAs and templates, such as the standard WS-Agreement (WS-Agreement
2007). Service delivery is described through the {service, SLA} pair, defining
exactly what the client is expecting from the provider. The complete lifecycle of
the service is mirrored by the life-cycle of the corresponding SLA specification. As
such, the SLA has a lifespan that is at least as long as the period of service usage by
the service consumer.
The main challenge of SLA
discovery and negotiation
resides in providing a
comprehensive environment for discovering the SLA under which a service may
be offered and negotiating the parameters of the SLA clauses in order to obtain a
contract which is best fit for its use, minimising over- and under-provision. Service
discovery based on SLA is well understood and is gaining acceptance in business.
SLA negotiation however is not widely accepted at present and its business justifi-
cation is being debated. In March 2009 WS-Agreement (WS-Agreement 2007) has
been in its last steps to become a full standard. It offers the only standard in this
area that has met some acceptance. Several implementations of this specification
have been developed since 2007, including some available as Open Source software
from the Grid Resource Allocation Agreement Protocol (GRAAP) Working Group
(http://forge.gridforum.org/sf/projects/graap-wg). However, the technical means to
perform negotiation are not yet there; the WS-AgreementNegotiation protocol, for
example, is still at early stages of maturity.
Extensive experimentation across many vertical market sectors has indicated
that the business justification for SLA negotiation is not widely accepted and when-
ever it is considered to be applicable, this is merely in relation to SLA discovery
(Dimitrakos 2009a). This finding is consistent with findings of several European
research projects such as TrustCoM, NextGRID, Akogrimo and BEinGRID (Parkin
et al. 2008). More importantly, the business reasoning for providing a capability to
re-negotiate SLAs remains unclear (as opposed to cancelling an SLA and replacing
it with a new one). Extensive experimental analysis in various business sectors by
the BEinGRID consortium (Dimitrakos 2009a) has also confirmed acceptance of
either simple short-term SLAs for use of IT resources or of complex legal contracts.
The latter are perceived as a means of treating higher value or higher risk offer-
ings by the parties involved, their definition typically involve qualified lawyers
and would not be automatically renegotiated. Furthermore national law in some
European regions obligates that renegotiation is treated as a negotiation of a new
contract. Nevertheless, it appears as if in some cases companies are willing to enter
a fixed long-term contract, and allow for short-term contracts (typically referencing
the over-arching long-term legal contract) that can be negotiated automatically,
within a limited scope.
The SLA
optimization
capability matches the information offered in SLAs
to the available resources. This improves the provider's scheduling strategy,
allowing the provider to improve the utilisation of its resources. It also allows
implementation of the business rules which govern the allocation of resources
based on KPIs such as the return value of the incoming SLA requests. Most
schedulers are designed to optimise the resource usage based on the incoming
