- reduce the cost,
- increase availability (HA),
- maintain security,
- plan for disaster recovery (DR).
- How is virtualization and federation helping?
- Applications are multiplexed to dynamic computational resources.
- Applications aggregation decreases the utilization variation.
- Understand well how much CPU, traffic, memory per application
- What is the profile of your server/applications, utilization seasonality, average, peaks, …
- Run experiments
- Virtualization is not adding any computational power
- Must satisfy basic needs for virtualization
- CPU
- RAM
- Disk
- Networking
- ...
- OS, special SW, databases require licenses.
- Licenses are based on number of processors.
- Some companies do not support the SW when ran in virtual networks.
- Each VM may need its license.
- Installation-based licensing versus instance-based licensing.
- This may higher up the total cost.
- Use consultant analyzing apps and support for SW upgrades.
Servers consolidation
HW requiements
License policy
Administration
- Higher administration in virtualized networks - an app per server is easier to monitor then multiple apps.
- SW to monitor the servers and balance the load.
- It requires tools and experience.
- Patching, updating tools and management.
- On premise, in house management, planning, … similar to current model.
- Virtual private cloud (VPC), private, isolated section, outsourcing to cloud providers - may have the advatages of public cloud.
- Cloud providers operate really large computational centers.
- Thousands of computers per administrator.
- Buy large quantities of cheap specialized HW (networking, UPS, computers, …).
- Cheap electricity, water power station, PUE < 1.3.
- Upfront versus over time expenditure.
- Easier to measure departments cost.
- What is better 1 PC computig a job 100 hours or 100 PCs computing it 1 hour?
- AWS simple monthly calculator.
Planning for cloud deployment
Not all apps good for cloud, the legacy applications may be too complex to move to cloud
- Planning is similar to virtual private cloud.
- Consolidation of servers.
- Apps server profile to choose cloud provider.
- Use consultants, too many choices today, difficult to compare.
- Still in charge of many administration aspects - educated staff.
- Service provider takes care of the OS, platform patches, security.
- Transparent scaling, load balancing
- Easy to grow, scale.
- Restriction on applications - they need to be written for PaaS.
- Easy for startup companies.
- Converting current aps to SaaS can be difficult.
- Provider, application dependency, lock-in
- Virtual, private cloud. Large investment.
- Currently big investments in on premise servers.
- Private + public = hybrid cloud.
- Backup to public cloud (encrypted).
- Move some services to public cloud.
- Take advantage of public cloud elasticity.
Cloud
Virtualizartion
&
Cloud
Jan Šedivý
Department of Cybernetics CTU, Prague
CIO job?
Virtualization
- Run several applications on the same box
- Virtualize – use 20% of available resources
- Utilize the whole box 80%
- server hardware,
- power & cooling,
- server storage (central memory required),
- networking (switches),
- space (rack space),
- administration savings.
How to estimate the savings
- 30 - 50% # of servers reduction
- Pay Back on Investment 1-2 years
Expected savings
- Managed by the service provider (SLA).
- Elasticity - almost end less computational power - easy to grow.
- HA - availability and geographical zones.
- r reliability, n redundant components.
- r(n) = 1-(1-r) * n
r (1) = 99%,
r (2) = 99.99%,
r (3) = 99.9999%
Private cloud
Public cloud - economy of scale
Company business model
Technology
Run several applications on the same box
Virtualize – use 20% of available resources
Utilize the whole box 80%