Liveblogging: A Five-Step Framework for the Value of Cloud Computing

Posted in: Cloud, Technical Track

I took part in a webinar on cloud computing today, including some of the top names in cloud computing services. As Pythian has some MySQL clients using cloud computing, I was particularly interested…

I was interested by the many levels of what cloud computing means, including such categorizations as Facebook apps being a part of the cloud. I think many of us consider cloud computing to mean “virtual infrastructure as a service” and overlook some pretty robust cloud computing that’s already out there, such as “application components as a service” and “software [platform] as a service”.

Following are my notes:

“Our objective today is to cut through some of the noise associated with ‘cloud’ and get to a real world approach for getting some serious value from the cloud.”

Frank Gillett, VP and Principal Analyst, Forrester Research
Gillett’s part of the talk is entitled “Understanding and segmenting cloud services”.

The term “cloud” is associated with a variety of attributes.

If people say the cloud or “*aaS” (ie X as a service), they mean:

  1. Standardized IT-based capability (compute, storage, network) from a provider that has one or more of the following:
    • Is accessible via Internet protocols from *any* computer (HTTP, REST, SOAP, or a web browser panel interface)
    • Is always available, scales automatically to meet demand (resilient, HA, scales up and down automatically)
    • Pay-per-use or paid by advertisements (ads, subscription-based or pay-per unit — compute time, bandwidth, storage)
    • Web or programmatic-based control interfaces (ie, Facebook, Microsoft Virtual Earth 3D)
    • Offers full customer self-service (customers can provision, manage and terminate services without involving the service provider. Control is via a Web interface or programmatic calls to service APIs)

“If you only take away one thing from this presentation: There are two cloud infrastructure-as-a-service markets emerging.”

    1. Cloud services for end users
      • Web-based services — ie, Google Maps API on your site, not visible to the end user’s customer.
      • Software platform as a service — Google App Engine, EngineYard. No control of OS, end-user only provides code.
      • App components as a service
      • virtual infrastructure as a service — aka compute as a service, hardware as a service, elements of this with Amazon EC2
      • NOT CONSIDERED CLOUD — physical infrastructure as a service — Traditional Data center services market, such as collocation or managed hosting.
    2. Cloud services for service providers

Cloud-infrastructure-as-a-service — includes enterprises building internal clouds, and the “as-a-service” explained above in the “end users” section.

In general, cloud computing simplifies what folks have to deal with, gets rid of barriers — if you’re a developer and don’t want to deal with the OS, the “software platform as a service” cloud is for you.

After this point it’s mostly what services are offered…..

Jeff Barr AWS – Review of Amazon’s services

Amazon Web Services (AWS) principles:

  • easy to use
  • fast
  • elastic
  • highly available
  • secure
  • pay as you go

Many Uses for AWS:

  • elastic computing (such as seasonal effects such as retail, or periodic such as payroll processing, scalable websites)
  • media distribution
  • scalable web sites
  • backup and recovery
  • financial applications

Amazon’s services:

    • Amazon Elastic Compute Cloud (EC2)
      • Virtually unlimited computing power — Quickly scale up/down by obtaining and booting new server instances in minutes.
      • Full root access to a Linux virtual computer — $0.10 / hour 32-bit 1 core, $0.80 / hour 64-bit 8core.

Recent features:

    • Linux, OpenSolaris, Windows Server
    • Availability Zones for reliability (different parts of data centers)
    • Elastic IPs for more flexibility
    • Persistent storage using Elastic Block Store (EBS)
  • Amazon Simple Storage Service (S3)
    • distributed data store
    • REST/SOAP web services API
    • Simple (Buckets, Objects, Keys)
    • SLA 99.9% uptime
    • Busy — 70K TPS (scalability, durability, availability)
    • Pay-as-you-go — Storage $0.15 / GB / month (exact fractions)
    • Data Transfer — tiered, $0.10 / GB – $0.17 / GB
    • Requests: nominal charges
  • Amazon Simple DB
    • semi-structured data
    • no index maintenance
    • no performance tuning
    • Query/SOAP API
  • Amazon Simple Queue Service (not discussed)
  • Amazon Flexible Payments Services (not discussed)
  • Amazon Mechanical Turk (not discussed)

Adoption Model that rPath has built — Billy Marshall, Founder and Chief Strategy Officer

  • Level 1: Virtualization – Virtualize infrastructure and applications
    • NOT talking about virtual machines or zones.
    • This is about decoupling your application from your infrastructure
    • Be careful about platform “lock-in”
  • Level 2: Cloud Experimentation – define reference structure, can experiment in Amazon EC2 because no commitment.
    • identify which applications to test out
    • idea is to test some apps, with the idea of moving many over
    • baseline metrics and anecdotal returns to build momentum
  • Level 3: Cloud Foundations – Lay the foundation for the scalable application architecture
    • deploy a build and lifecycle management platform
    • release policies and procedures are documented
    • time to market for new apps will be drastically changed
    • if maintenance isn’t automated, there are problems here
  • Level 4: Cloud Exploitation – select the desired cloud environment, begin broad-based deployments, manual provisioning and load balancing.
    • here you solicit which apps can be deployed in the cloud
    • what is best if you don’t have to size for the peak demand?
    • not all apps are suited to run externally in the cloud
    • again, be careful of “cloud lock-in”
  • Level 5: Hypercloud (“cloud nirvana”) – making dynamic decisions about application workload, capacity arbitrage and self-service application provisioning.
    • be able to select a target environment at run-time of which cloud environment to use.
    • can be lots of organizational resistance
    • lots of technology needs to be understood to get to this stage

Quick poll where in the adoption model are you?
level 1 41.9%
level 2 22.5%
level 3 3.2% — where Pythian is!!
level 4 3.2%
level 5 0.0% (not really possible)

rPath has rBuilder and rPath Lifecycle Management Platform. 42% of all Amazon Machine Images (AMI) were built with rBuilder. rBuilder is offered free online.

rPath offers services of readiness assessment, implementation planning, actual implementation, and training.

Jeff Schneider, CEO, Momentum SI – Review some of their services:

One-day crash course on cloud computing
Help implement pilots in the cloud
How to change organizational-level policies and procedures to include a scalable delivery approach

Interested in working with Sheeri? Schedule a tech call.

1 Comment. Leave new

Google have a good group which has discussed some of these questions in detail.

For now, cloud = on-demand computing, whether for peak load, testing, batch processing, whatever CPU intensive task has a limited lifecycle.
As the provider market increases/matures, hopefully we will see the prices drop making level 4 and level 5 of that framework more realistic. The increase in complexity of IT architecture for 4 and 5 in large as well.

Have Fun



Leave a Reply

Your email address will not be published. Required fields are marked *