Presentation is loading. Please wait.

Presentation is loading. Please wait.

Handling Apps, Databases and VDI Workloads Trent Steele Southeast Region Sr. Systems Engineer (336) 482-5725

Similar presentations

Presentation on theme: "Handling Apps, Databases and VDI Workloads Trent Steele Southeast Region Sr. Systems Engineer (336) 482-5725"— Presentation transcript:

1 Handling Apps, Databases and VDI Workloads Trent Steele Southeast Region Sr. Systems Engineer (336) 482-5725 @thevbox

2 The Business IT…What are we here for?What drives your business?

3 What drives our businesses?

4 By 2016 more than 80% of server workloads will be virtualized* More than 55% today are virtualized, but getting to 80% or more virtualization presents challenges** −Visibility −Manageability −Cost Most private clouds will be based on 100% virtualized environments Virtualization Trends in Enterprise IT 4 * Gartner, Virtualization Key Initiative Overview, 22 July 2011, ** Gartner, Virtualization Key Initiative Overview, 11 April 2014,

5 Applications

6 Application Workloads:

7 Business Owner I need a new app for easier web-based billing It’s all about the App!!!  We need to: −Get capacity now −Get s/w stacks deployed −Simulate production  Once in prod, we need −Plan capacity for app −Place on Tier 1 capacity −Provision the app, web, and database servers −Set up the load balancer −Set up the firewall −Set up data protection −Set up mgmt −Manage the app −…−… Developers Just getting the infrastructure to develop on is so slow! Operations How do we get the HW, manage the app, and deliver the SLA in production?

8 Databases

9 Understanding Database Workloads

10 Physical NAS/SAN How Traditional NAS/SAN breaks virtualization Physical Servers Physical Switches VLAN/vSwitch Host VLAN/vSwitch Host Read tunedWrite tuned50/50 Mix Datastore

11 Tintri VMstore Datastore With Tintri: Interchangeable Storage Physical Servers Physical Switches VLAN/vSwitch Host VLAN/vSwitch Host Datastore Real-time, automatic, read/write optimization

12 SQL VM on traditional storage 12 Read tuned Datastore How can you expect a consistent SLA when you have disparate VM workloads on a fixed type of datastore?

13 SQL VM on traditional storage with RDM’s 13 Read tuned Datastore So you overprovision a datastore or use RDM’s with a specific SLA? CRAZY!

14 VDI

15 VDI: Top Mistakes With This Workload 1.Not calculating user bandwidth requirementsNot calculating user bandwidth requirements 2.Not considering the user profileNot considering the user profile 3.Lack of an application virtualization strategyLack of an application virtualization strategy 4.Improper resource allocationImproper resource allocation 5.Protection from Anti-Virus (as well as protection from viruses)Protection from Anti-Virus 6.Managing the incoming stormManaging the incoming storm 7.Not optimizing the virtual desktop imageNot optimizing the virtual desktop image 8.Not spending your cache wiselyNot spending your cache wisely 9.Using VDI defaultsUsing VDI defaults 10.Improper storage designImproper storage design

16 Understanding VDI Workloads

17 Performance −Small random-write workloads −Burst Traffic −Boot storms −Antivirus scans Complexity −Reference architectures and best practices guides with 100’s of pages and multiple “knobs to tune” −No way to isolate VMs and identify performance problems Cost −Storage overprovisioning Understanding the VDI Workload

18 Storage must deal with the different states of VDI −Boot storm, steady state, AV scanning, recompose/patching Latency can impact user experience and throughput can limit scalability −Legacy and flash-only storage uses a big-hammer approach to performance Ideally VDI requires VM level QoS and performance allocation −Mix and match different users or even workloads on same storage Performance: VDI Storage Performance Needs MorningsBusiness HoursAV ScanRecompose Reads Writes IOPS Log off

19 Complexity: Simplifying VDI deployments Reducing the Storage Best Practices to be followed by 99%

20 The Tintri Product Family Tintri VMstore T620Tintri VMstore T540Tintri VMstore™ T650 2,000 VMs1,000 VMs500 VMs TINTRI GLOBAL CENTER™ TINTRI REPLICATEVM™ Tintri VMstore™ T650

21 Tintri Validated for VDI 21 1,000 user example One datastore (no LUNs, RAID, queue-depth etc.) One pool (one replica with linked clones) 8 Minutes to configure storage No storage tuning Ultrabook level of user experience Sub $60/VM Recompose during the day Server VMs and Desktop VMs on same storage Source: Tintri – VMware Joint Testing report.html

22 Storage makes up a substantial portion of upfront cost −Performance driving storage requirements −Over-provisioning traditional storage for performance or use of expensive flash-only arrays When planning VDI, consider −Upfront costs −Operational costs** Cost: Cost of Desktop Virtualization ** $1 spent on Hardware requires $3.8 more to support it of it (Source: VMware/IDC)

23 What Happens When It Doesn’t Work? UpsetAdministratorAngryManagement Solutions Provider / Manufacturer

24 Yesterday’s Concept of Storage – COMPLICATED!

25 Innovation Truly Changes Things

26 What If… Storage Was Smart? Apps, VDI, Databases Adaptation to the Now Understanding Different Workloads Never have Noisy Neighbor again Protection from Harmful Issues VM Auto- Alignment Visualize Bottlenecks Per VM Visibility Individually Understand Workloads 10:1 Rack Savings Lower Cost and Rack Footprint Built for Virtual Machines ZERO LUNS No Need for LUNS or Volumes ZERO Volumes vDisk QoS No Need for vVols vDisk Isolation Future Growth Can Adjust to Future Workloads Application Changes No Capacity Based Performance Crashes Predictive Scaling Performance Visualization

27 SAP Microsoft Exchange Physical data center Virtualized data center With traditional storageWith smart storage Storage array S A P M S F T MSFT VMs SAP VMs Storage array See, learn and adapt Virtualization Disruption in Storage

28 VM/vdisk Assessment Action Adaption Plan Execution Verification and Learning Smart-Storage: Ability to adapt to the workload

29 Tintri Application-aware Smart Storage Sees Learns Adapts 1 1 2 2 3 3 Automatically takes care of complex and mundane tasks Manages and protects VMs with a few mouse clicks Seamlessly supports multiple hypervisors Adapts to the needs of applications without human intervention Guaranteed I/O “lane” for each individual application Scales to thousands of VMs easily VM-level visibility for insight and control Quick performance troubleshooting and trend analysis Eliminates storage overprovisioning

30 Customer Testimony Eric Hester VP of Technology Green Cloud Technologies @GreenCloudTech

31 CAAS Green Cloud Tintri Partnership Presented by: Eric Hester Co-Founder / VP Engineering

32 Green Cloud Overview Provider of Cloud Services based in Greenville, SC ▪ Public Cloud ▪ Private Cloud ▪ Disaster Recovery 2 Data Centers 1000s of VMs ~ 1PB of Storage 4 Netapp Clusters 12 Tintri VMStores

33 Our Challenge ▪ All our workloads are unknown and unpredictable ▪ We don’t manage the applications ▪ Customers change their workloads ▪ Resellers can self provision large workloads quickly ▪ Disaster Recovery failovers ▪ Traditional SAN (in our case Netapp) environment takes enormous man power to tune and manage this evolving workload mix ▪ Things “work great right up until they don’t”. Very hard to predict, only solution is constant overbuilding.

34 The Solution ▪ We implemented storage tiers and focused on vendors built for virtualization. ▪ Tintri was tested for our premium tier that was meant for Database, Terminal Servers, VDI and other high IO workloads ▪ During testing we regularly told Tintri the thing that disturbed us the most was we didn’t have to think about their box. It just worked and worked well. ▪ This continued into production. Engineering and Operations got their lives back. ▪ We have since selected Tintri as our VDI storage platform as well because it performs so well.

35 Why we are Tintri Partners ▪ Tintri is VM Aware ▪ VM specific replication and snapshot configuration ▪ Per VM IO queuing ▪ Flash First vs Tiering Cache ▪ Write advantage with immediate read advantage for unpredicted workloads ▪ Reduced Complexity ▪ Self contained ▪ No spindle count management ▪ No storage area network to manage ▪ Excellent Support ▪ It Just Works…

Download ppt "Handling Apps, Databases and VDI Workloads Trent Steele Southeast Region Sr. Systems Engineer (336) 482-5725"

Similar presentations

Ads by Google