This week virtualization.info is in Prague, Czech Republic, to follow and report about the Burton Group (a Gartner independent subsidiary) Catalyst Conference 2010 Europe.
The event features a specific track about virtualization and IaaS cloud computing and virtualization.info will cover some of the sessions that are part of this track.
This morning the first one is titled Building a Viable Cloud Adoption Strategy, performed by Drue Reeves, Vice President and Research Director at Gartner.
Reeves is on stage: we know that cloud computing has a lot of issues (security, vendor viability, ROI, liability, etc.).
He says that we passed the definitional phase in cloud computing and this is the year of planning, building and maintaining.
Customers are primarily divided in two groups: the ones that are ready to happily jump on the cloud computing bandwagon and the ones that want to know nothing about it.
To approach cloud computing in the proper way and fall somewhere in between these two extremes we need a proper adoption strategy in several steps:
- Build a cloud core team, which includes executive sponsors and key stakeholders.
Set clear business objectives and cloud adoption principles by managing conflicting issues and make key decisions about priorities.
Scope the effort.
- Assess the business processes and applications, with the help of business impact analysis (BIA) and risk assessment. (Reeves says that most companies seem to lack the knowledge to answer this point).
A key aspect of the application assessment is the definition the application Quality of Service (QoS) requirements. Application dependencies (hardware and software requirements) and interdependencies (for multi-tier applications or middleware) must be mapped too.
Cost analysis is another critical point (again, Reeves states that most companies don’t know their internal costs so there’s no basis for comparison): customers have to carefully watch for hidden and growth costs, and these aspects should be considered in vendor’s evaluations.
Last but not least, the impacts to organization should be assessed. This includes the changes to procedures and IT governance, the resource required to build and operate, the time to implement, and the staff and end-users training to maintain and operate.
- Select the vendor, a process made of three sub-steps.
First, customers should file requests for information (RFIs), including questions about vendor’s viability, profitability, security procedures, IT implementation, etc. Two-three vendors should be the target, some of them may not answer as their cloud offering is not negotiable.
- Mitigate residual risk and limit liability. As some risk cannot be avoided the best approach is to have an exit strategy, for instance by using cloud brokers and liability insurance.
- Steady stay. The cloud team and the organization must resist the temptation to avoid some steps above, and routinely verify everything as planned.
That’s all for this session.
The next session we’ll cover, this afternoon, is titled Hardware IaaS: An Immature, Yet Rapidly Growing Market, performed again by Drue Reeves.
Reeves on stage.
He describe his thesis: the market is changing rapidly, with many new players, pricing and QoS changes.
At the moment it primarily is a playground for SMBs and enterprises delivering non-critical applications. We need enterprise-grade clouds.
First of all, a clarification on definitions: unless the offering includes rapid compute instance provisioning and multi-tenancy per server, and leverages some sort of virtualization, it can’t be really called IaaS cloud computing but it’s more likely the dear old hosting/co-location.
The market leader is Amazon. The vendors that will have a significant market share are AT&T, Verizon, CSC, BT and IBM. Interesting newcomers are Rackspace, Terremark, Savvis, Softlayers, Flexiscale.
Smaller or niche players include: OpSource, GoGrid and ElasticHosts.
As you can see from the list, the IaaS market is made of players coming from different existing markets: the telecom providers, the integration providers, the hosting providers and the brand new, pure IaaS providers.
The market drivers are the desire to externalize IT, the increase in competition, service price and the rise of enterprise-grade clouds.
The technology trend is to use virtualization. No matter what hypervisor they chose, hardware virtualization is the foundation block of every IaaS offering today.
Vendors like IBM, Cisco, HP, Dell, NetApp and EMC are expected to deliver a number of server and storage products that are cloud-ready. Enterprise management vendors like CA will follow.
There’s an ongoing standardization for the IaaS market, led by technologies like OVF, the VMware vCloud API, XML-based SLAs, and the Red Hat Deltacloud API.
The open source effort is significantly increasing thanks to products like Eucalyptus, ParkPlace, Project Caroline, Numbus, Ubuntu.
Other market trends that are emerging.
Virtual private clouds, VPNSs and VLANs, offered as additional features of cloud platforms to guarantee greater data protection, offsite replication, etc.
There’s an increase in transparency as cloud providers are accepting SAS 70 audits and are willing to discuss their infrastructures details. SAS 70 and ISO 27001 certifications are becoming differentiation features to win customers over competition.
The pricing model is changing: reserved capacity is an emerging model while the price goes down because of increased competition. Some vendors won’t survive the price competition or the amount of features offered by biggest players.
Providers are likely to offer two different IaaS cloud platforms: a low-cost one for non-mission critical applications and another for critical applications that require demanding SLAs.
SLAs will become more negotiable while insurance companies will become a relevant part of the equation.
Reeves closes his session saying that the market will dramatically change by 2011.
The last session virtualization.info covers today is titled Hybrid Clouds: Two Worlds Collide, performed by Drue Reeves and Chris Wolf, Research Director at Gartner.
Reeves and Wolf are on stage.
The presentation is focused on a specific hybrid IaaS cloud computing architecture that doesn’t exist today but that Burton Group believes it could be the future.
In the described scenario customers demand for new virtual machines through the self-service provisioning portal of his on-premises cloud infrastructure which is connected to multiple external, public cloud platforms.
When the orchestration layer recognizes that the private cloud doesn’t have enough resources to serve the requested VMs, or if the VMs have specific requirements that cannot be satisfied locally, the VMs are built on-premises, but then are sent across the Internet to the first federated public cloud platform that can serve them.
The migration from the private cloud to the public cloud happens transparently and accordingly to corporate policies on SLAs and security. The remote VMs are management from the same local management console as they are served locally.
For this scenario to happen Reeves and Wolf believe a few components are needed:
- a unified architecture that include a virtual infrastructure, a cloud OS
, a service catalog and more components that are in the work today.
- an infrastructure authority (IA) that centrally store metadata and that maintains dependency maps for compute resources, memory, network, storage, security policies, etc.
The IA also stores and enforces the organizational policy to ensure that the migration can happen without breaking any corporate rule.
- license mobility, that allows the service provider to manage guest operating systems and application licensing on behalf of the customer.
Here Wolf recommends a license model based on vCPUs, users and seats to increase the mobility.
- metadata standards, where OVF is a good candidate but it has to turn into a runtime standard rather than just a packaging standard.
- machine readable and negotiable SLAs, that could fit the OVF standard.
- common API and a data model.
Wolf highlights how the number of “standard APIs” is rapidly increasing and we already have too many (EC2, vCloud, Deltacloud). A plug-in model could be the answer.
- infrastructure transparency, needed for the IA to make decisions about where to deploy VMs.
That’s all for our coverage of the first day of the Catalyst 2010 Europe. Stay tuned for tomorrow, when the focus will shift on desktop virtualization and client hypervisors.