Your SlideShare is downloading. ×
Managing Performance of Clustered Applications - 6350
Upcoming SlideShare
Loading in...5

Thanks for flagging this SlideShare!

Oops! An error has occurred.


Saving this for later?

Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime - even offline.

Text the download link to your phone

Standard text messaging rates apply

Managing Performance of Clustered Applications - 6350


Published on

1 Like
  • Be the first to comment

No Downloads
Total Views
On Slideshare
From Embeds
Number of Embeds
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

No notes for slide


  • 1. Managing Performance of Clustered Applications Yori Lavi Founder, Chief Product Officer [email_address]
  • 2. Managing Performance of Clustered Applications ABSTRACT
    • As more enterprises turn to J2EE to deliver their mission-critical applications over the Web, there is a greater reliance on clusters to provide mission-critical services to ensure maximum service availability and scalability. The challenges in clustering a J2EE application can be attributed to the very nature of distributed computing. Clusters promise to bring affordable mainframe-level computing power and reliability, using cost-effective servers. But clustering also introduces increased management complexity and amplifies the need for new transaction-level management approaches that can address the management complexity from a service delivery perspective.
  • 3. Agenda
    • Intro
    • Managing workloads – the performance perspective
    • Market trends
    • New clustering techniques create new challenges
    • Gap analysis: where existing tools fail
    • Moving forward – techniques to bridge the gap
    • Q&A
  • 4. Managing workloads - the performance perspective
    • Modern workloads – not your uncle’s mainframe workload
      • Diversity
      • Variance
      • Rapid change
    • Visibility – measure the right metrics
    • Active measures – control the right resources according to the right metrics
    • Anticipate deployment challenges based on combination of historical data and business projections
  • 5. Market trends
    • Application integration
      • SOA, ESB, ad hoc custom integration
      • Method:
        • Wrapping existing application with interfaces
        • Developing J2EE infrastructure tier between applications/services.
        • New flow management tools
      • Benefits:
        • Faster implementation of new business process. Better business flexibility
        • Focus on transaction level services (vs. application level services).
      • New challenges:
        • Increased need for visibility. (finer granularity: User, Transaction, resources per Trx/user/tier/machine…)
        • Increased need for (near) real-time control
        • More is riding on exact capacity planning
        • Higher need for understanding Cost of Transaction
        • Less control tools
  • 6. Market trends
    • Hardware consolidation
      • Clusters, load balancing, virtualization
      • Method:
        • Multiple, similarly configured tiers, operating together to server heterogeneous collection of users and applications
      • Benefits:
        • Higher availability
        • Lower TCO
        • Scalable architecture
      • New challenges:
        • Complexity is exponentially higher
        • Platform availability is higher, but service availability could be lower
        • Capacity planning tools are less effective
        • Lower visibility of transaction path and resource consumption makes problem isolation very difficult
        • Global configuration changes are not as effective as in a silo environment
  • 7. New techniques create new challenges
    • The business need is focused on service to user (Transaction). J2EE management tools focus on single ‘vertical’ tiers
    • ITIL require close tracking of service levels to the business, IT configuration management and IT procedure to manage both
    • In Silo-ed environment, over-provisioning and global parameter changes are effective
    • In clustered environment there are many more unpredictable variables
    • Number of components in a Silo are smaller than in integrated/cluster environment; therefore identification of changes and problems is more difficult in the new environment
  • 8. Gap analysis: where existing tools fail
    • Existing management tools disparately monitor either a user’s perspective or a single vertical tier in the stack/cluster or the cluster’s routing
    • Impact of a single tier problem was limited to a single silo
    • Context in a silo is more homogeneous
    • In Multi tiered, cluster environment:
      • The context (who, what, where, when) is varied, masked and hard to follow
      • Any problem has a wide impact (multiple applications and users are integrated)
      • ID problem, business triage, problem isolation - needs to be near-real-time
      • Time to problem remediation is critical
    • Existing management tools lack the context to provide the visibility and have no active remediation measures required in the new environments.
  • 9. Moving forward – techniques to bridge the gap
    • Requires new holistic approach that looks both at business context (WWWW – who, what, when, where) and physical context (Tiers, machines, resources, time line)
    • Implement holistic approach through the application life cycle:
      • Pre deployment testing should look for SLA compliance per tier and per transaction
      • Capacity planning before deployment should have the real granular data: resource consumption per business service per tier
      • Cost of transaction should be known to the business well ahead of deployment
      • When deploying – need to monitor multiple levels. Both global (per transaction) SLA and per component (each leg a transaction perform on a each tier)
      • When deploying – need to monitor multi dimensional SLA (elapse time, CPU, IO, Memory, topology, calling ratio between tiers…)
      • When looking at a tier – look for resource consumption per transaction context, on a timeline
      • Once an SLA is breached, the NOC should be able to perform a quick business triage
      • Localizing the problem is quicker when the holistic approach is practiced
      • Utilizing active measures to remediate immediate problems
      • Employ techniques to control resource consumptions based on business policies
  • 10. Moving forward – techniques to bridge the gap
    • Techniques
      • Single repository infrastructure
      • Collection and aggregation platforms
      • Point collection techniques
        • Instrumentation
        • Correlation
        • Tracking
      • Consideration
        • Low overhead, production grade
        • Quick Time To Value
        • Flexibility of monitoring and control tool.
    • Build, buy or both?
  • 11. Business tracking of transactions fills the gap in SOA management solutions
  • 12. Business tracking of transactions fills the gap in SOA management solutions (cont.)
  • 13.
    • [email_address]
    Q&A OpTier is an innovator in Transaction Workload Management solutions that assure IT business service levels and optimize resources for the enterprise