Building a Smart Cloud Strategy Marlin McFate Public Sector CTO, Riverbed Technologies
Behind every digital experience is a human one RIVERBED’S PURPOSE: ADVANCING THE HUMAN EXPERIENCE IN THE DIGITAL WORLD
Questions What can I move? What will the performance be? What do I have to do to be able to move?
! Difficulties Self-inflicted problems Unintentionally painted ourselves into a corner Can’t access cloud in the most efficient manner Many legacy applications
Battle Plan for Cloud Migration Identify App Ecosystem Predict App Response New Network Hotspots Continuous SLA—Before, During, After Before migration, the application eco-system must be identified at all levels Network map Application map User map Applications can be latency sensitive It is critical to identify the impact of added latency Tier dependency must be identified for a seamless migration schedule and strategy Changed workload and path can create unintended network hotspots Predict new hotspots before they cause problems Conduct double- failure studies Monitor applications Prior to the migration Post-migration Productivity report
Identify Ecosystem Can’t move something you don’t understand Rationalize applications
Identify Ecosystem Rationalize network path Understand current end-user performance
Predict App Response What is the performance? What will the performance be? What elements are latency sensitive?
Identify Network Hot Spots Where will capacity need to change? Where will my current architecture be inadequate?
Performance: Before, During, After Monitor end-user experience: For every enterprise application On any physical, mobile, or virtual device “ “Click to render” time Automatically baselines “normal” performance Correlates app performance to device health/performance Automatically compare/validate changes
This is Not Trivial IT Ops Network Ops App Ops DevOps LOB Switch PACKETS SNMP, CLI, WMI, SYN FLOWS EUE TELEMETRY Switch Router Devices UC and App Servers Cloud SteelHead
Assessment Factory Engine Support major application relocations Perform a fully automated proactive evaluation study for applications that are candidates for migration. The study for each application includes: Evaluating “at risk”—chatty links that experience degraded performance with increased latency Providing insights into application dependencies (application topology) Identifying comprehensive network impact Providing insights into underlying network topology Use the study results of each application in assessing the overall migration impact as multiple applications get migrated Input Application Details Assessment Factory Processing Engine Output—Assessment Factory Engine Document OP1—At Risk Evaluation (communications with high turns) OP2—Application Topology OP3—Network Impact Assessment OP4—Network Topology
Implementation Approach Do we know the app to study? Gather application footprint details NO Tap packet traffic at core or distribution level STEP 1: Application footprint details COVERAGE BIDIRECTIONAL TRAFFIC STEP 2: Data collection setup COVERAGE PICK AN APP TO STUDY STEP 3: Tools configuration STEP 4: Data validation and usage SNMP/CLI Entry/exit traffic + server level traffic needs to be covered YES PHYSICAL NO Identify virtual environments to consider for flow export AR11 STEP 5: Assessment factory processing engine OP2 STEP 6: Assessment factory engine output document * —Automated YES NO ANALYZE DATA WAN USING NETFLOW SERVER TRAFFIC USING PACKETS COVERAGE For CLI: obtain cli show command outputs For SNMP: identify the repository to obtain performance metrics Identify list of access layer devices from which packets need to be captured, based on application repository Identify the WAN devices for the Datacenters that would see application traffic Obtain a list of servers with high turns (Front end and Back end traffic) and associate them to their respective apps OP1 + OP2 FULL COVERAGE (Preferred) LIMITED COVERAGE USING ACCURATE FLOW EXPORT CONFIGURATIONS USING ACCURATE FLOW EXPORT CONFIGURATIONS Identify Network devices using trace routes and LLDP outputs Perform scripting to convert data as needed Tap packets Enable flow export and collect application data BACK AND FORTH NEEDED TO ACHIEVE PROPER COVERAGE PACKET FEEDS NETFLOW EXPORTS FROM FLOWGATEWAY CORRECTIONS NEEDED IF VALIDATION FAILS FOR APP TOPOLOGY NetIM AR11 NetProfiler CORRECTIONS NEEDED IF VALIDATION FAILS WITH SNMP/CLI ACCESS BIDIRECTIONAL TRAFFIC BIDIRECTIONAL TRAFFIC PER INTERFACE Output—Assessment Factory Engine Document OP1—At Risk Evaluation (communications with high turns) OP2—Application Topology OP3—Network Impact Assessment OP4—Network Topology Data Validation and Data Usage Data Validation and Data Usage Data Validation and Data Usage OP4 METADATA * OP1 METADATA * OP2 METADATA OP3 METADATA * PROCESSED OUTPUT * Assessment Factory Processing Engine
Battle Plan for Cloud Migration Identify App Ecosystem Predict App Response New Network Hotspots Continuous SLA—Before, During, After Before migration, the application eco-system must be identified at all levels Network map Application map User map Applications can be latency sensitive It is critical to identify the impact of added latency Tier dependency must be identified for a seamless migration schedule and strategy Changed workload and path can create unintended network hotspots Predict new hotspots before they cause problems Conduct double- failure studies Monitor applications Prior to the migration Post-migration Productivity report