Organic Name Reactions for the students and aspirants of Chemistry12th.pptx
Data storage for the cloud ce11
1. Jim McDonald Chief Strategy Officer, Xiotech Data Storage for the CloudBalancing Performance, Capacity and Cost
2. Most data follows a similar pattern Write, read, ignore “Small” amounts of data follow a much more aggressive pattern Write, write,read, write,read, read, write, read,read, read,write, write, write, write,read, write The result is that for effective cloud storage you need to have both high capacity and highperformance And it needs to be cheaper than could be done in-house Storage for the Cloud
3. Traditional cloud storage was viewed as a data dumping ground All about $/GB, performance “irrelevant” Cloud storage is now becoming a lot more active: File synchronisation Big data Calendars … And performance is needed for cloud computing If it can’t do it faster and cheaper it won’t win Why are the Requirements Changing?
5. Over the past 10 years, on a performance/GB basis, drives have gone down 50x What Does This Mean?
6. A standard enterprise drive can handle roughly: An Exchange server with 125 users @ 2.4GB/user A 300GB database with 12 transactions per second A virtual desktop implementation with 3 users @ 100GB/user The capacity Vs. performance numbers are very imbalanced against performance For internal IT this is wasted expenditure For cloud service providers this is unobtainable revenue Performance is the Limiting Factor
7. Snapshots Thin provisioning Deduplication (not compression) Traditional tiering Housekeeping Things which Lower Performance
8. Traditional tiering is a great example of sacrificing performance for capacity Because tiering focuses on capacity Current tiering algorithms are especially bad for heavily-loaded environments Think sat nav before traffic information Enterprise storage: 10,000 IOPS Traditional Tiering Archive storage: 2,000 IOPS Performance Current tiering solutions are fine if you just want capacity (But if you just want capacity don’t bother with tiering)
9. Add disks Fine, as long as you don’t use the capacity Increase the performance per disk Cache (sometimes) Better understanding of drives Intelligent data placement algorithms Tier for performance Use all of the performance of all of the drives How to Fix This:Increase Performance
10. Cloud test suite provides clean copies of every current version of Microsoft Windows Desktop and server, all language patches, all valid combinations of patches (several thousand instances) Automatically uploads program and runs tests, stores results in centralised area, then destroys the VMs Performance is the enabler: The time taken to create and destroy the VMs is at the cost of the service provider and not the customer So if it can be reduced the service provider makes more money The time taken to run the tests is a fixed rate So if it can be reduced the service provider makes more money Example:Cloud Test Suite
11. Stop thinking like it’s 2000 “If I could get all the old data off my enterprise drives I’d be rich” Capacity is not expensive any more, performance is Measure anything other than archive storage in terms of performance not capacity Be wary of capacity-increasing (and performance-decreasing) “added value” features Do you really need more capacity and less performance? Closing Thoughts