Best Practices for Validating the Performance of Data Center Infrastructure Henry He Ixia
Game Changers Big data - the world is getting hungrier and hungrier for data 2.5B pieces of content 500+ TB ingested per day 2.7M Likes per day 300M photos per day 100+PB data in single Hadoop cluster 25M+ Subscriber 4M Rating per day 3M Searches per day 30M Plays a day 1B hours of streamed traffic in June 2012 75% select movies based on recommendation Big data, Web 2.0 and social networking applications and mobile apps are all contributing to unprecedented growth in unstructured data Don t let your data center be the bottle neck
Data Center Scale How do you validate the elastics scalability requirements to manage demands of Big Data applications? Compute Storage Networking All three vector limits need to be stretched to truly assess the big data application performance impact
Enterprise Challenges Public Private Hybrid Is my data secure? Which cloud? How will my application workloads perform? What is the capacity limits? How will the infrastructure scale as business grows?
Provider Challenges IaaS Compute Storage Network Processor Memory Nodes Drives Images Persistence Internal External Inter connect Infrastructure hardening Performance, quality of service delivery, security effectiveness of physical & virtual infrastructure Optimization of resources Performance assessment of compute, network and storage infrastructure Characterize the capacity limits Improving application performance End-to-end application workload performance and impact on user experience
Virtualization in Cloud Data Centers Server, Storage and Network virtualization are critical aspect of data center deployment Key Test Challenges: What is the tradeoff between # of VMs and I/O performance? Throughput/latency of virtual components/hypervisor? How does VM mobility impact application performance? QoS for different applications on a converged link? Can the network infrastructure manage traffic growth? How can I guarantee the performance of tier 1 applications on my virtual infrastructure?
New Test Tools Need to seamlessly validate both virtual and physical data center infrastructure Testing in a virtual environment requires embedded test tools Virtual test ports act as VMs to: Generate network and application traffic Measure throughput/latency Validate application performance Identify impact of VM mobility Virtual test ports
Physical ports: emulate the converged data center network Use Case: IO Performance OBJECTIVE: Validate the converged traffic performance of virtualized servers and CNAs TEST SETUP Install virtual Ixia ports inside the server Attach physical Ixia ports to CNA Apply real-world application I/O profiles with local and remote storage KEY METRICS TCP/UDP throughput Latency Storage I/O performance (IOPS) Error rates SUT Virtual ports: emulate VMs, generate & measure application traffic
Use Case: CNA IO Performance I/O Throughput Latency Measurements
OBJECTIVE: Use Case: Live VM Migration Measure how applications/services are affected during live migration TEST SETUP: Virtual test ports emulate and generate traffic between an HTTP Client and HTTP server. Migrate HTTP server to another host machine KEY METRICS: Transaction Latencies Transaction Rates HTTP Client HTTP Server Live Migration HTTP Server
Use Case: Live VM Migration Is HTTP adversely affected in the time it takes for the VM to achieve full operation state on new physical server? Transaction Time Transaction Rates
Use Case: Virtual Appliance Tuning OBJECTIVE: Compute resources are often assigned in an ad-hoc fashion and assigning more cores to VM can decrease performance Validate the optimal amount of compute resources to VMs TEST SETUP: Vary compute resource (CPU, memory) assignment Generate realistic application workload on virtual appliance interfaces Use application emulation to validate the QoE metrics KEY METRICS: Application Throughput Transactional Latencies Connection capacities and setup rates 1 CPU 2 CPUs Optimal 4 CPUs
Use Case Storage Performance OBJECTIVE: Validate the virtualized and real storage performance impact Understand the tradeoffs between different storage architectures FC, NAS/iSCSI or CDMI Assess the impact of caching and data integrity TEST SETUP: Use virtual Ixia ports or physical ports to emulate storage clients or initiators Generate realistic storage workloads with user defined random data within the I/O Validate the storage performance, user authentication and data integrity When caching is in place validate the data reduction performance KEY METRICS: Application Throughput Transactional Latencies Data Integrity Mismatch Data Reduction Performance of WAN Optimization
Use Case: Network Performance OBJECTIVE: Measure capacity, forwarding performance and energy efficiency TEST SETUP: Physical or Virtual test ports generate and measure traffic across switching infrastructure KEY METRICS: Port density Traffic throughput, latency and loss Power consumption
Use Case: Converged Storage Performance OBJECTIVE: Verify no drop (no pause control) in absence of congestion Verify rate drop (pause control) for the lower priority flows under congestion TEST SETUP: Configure switching tiers with at least two PFC priorities for LAN and SAN traffic Send different ratios of application and SAN traffic Oversubscribe egress ports to observe PFC operation KEY METRICS: Measure SAN and LAN traffic separately Application and Storage throughput Application and Storage latencies
Public test: Lippis Data Center Fabric Testing Live VM Migration: East-West HTTP Server-Server Traffic Flows Server Lookup & Data Retrieval: East-West Server-Database/ Database-Server Traffic Flows Internet Browsing: North-South Client-Server/ South-North Server-Client Traffic Flows IP-based Storage Requests & Retrieval: East-West Server-Storage/ Storage-Server Traffic Flows Lippis Report Cloud Simulation Test RFC 2544 RFC 2889 QoS 100% Throughput Low Latency/Jitter No Loss
Public test: EANTC/LightReading Cisco CloudVerse End-to-end cloud validation Multi-tenancy Active-active fabric Tiered cloud services Virtual security gateway Service provisioning time
Testing the Elastics Data Centers Measure performance of cloud infrastructure VM scalability Throughput and IO capacity of NAS or SAN architectures Forwarding performance of the networking infrastructure (N-S vs E-W) Emulate realistic application workloads Compute and storage tuning for various applications on virtualized infrastructure Flexible IO generations for simulation of various application workload patterns Ensure data security and data integrity Realistic subscriber emulation with authentication to validate enterprise access policies Validate data corruption with byte-by-byte pattern matching Validate the converged cloud infrastructure Unified platform to validate compute, storage & network in conjunction with applications Seamless validate both virtual and physical End-to-end Converged Data Center Validation
Thank you For more Information