Die Verwendung von Storage-Arrays wird allgemein als natürliches Design einer Virtualisierung angesehen. Die Verwendung dieser Storage-Arrays fürhrt allerdings oft zu Problemen mit der Speicherleistung. Flash-Speicher wird als Retter für die Speicherengpässe angesehen. Stellt sich nur noch die Frage, wo denn Flash am sinnvollsten eingesetz werden sollte?
Frank Denneman, Tech Evangelist und Author, sprach am VMware Storage Event vom 30. April 2015 über den Einsatz von Flash-Speicher. Ebenfalls stellte er in seiner Präsentation PernixData FVP™, mit der die Storage-Performance um ein vielfaches erhöht werden kann.
The hypervisor contains context about I/O. However when leaving the hypervisor, the context is lost. Typically the compute and storage stack is comprised of many components of different vendors. Consequentially unable to provide an integrated QOS throughout the stack.
And its going on for a few years and it’s a great advancement in the storage technology and which we haven’t seen for decades. One fine morning you wake up and you storage substrate is ready to go 10.000 times faster then your older storage substrate.
Well what do you do with it and it turns there’s many folks thinking about and and it turns out that every thought was to make something that’s already existed, which was a storage array, faster. And by itself it’s a great thing, that is needed.
But the REAL question is, is that the next big idea? Is that what Flash Tops at? Is it the Next Big Thing?
Or can we do something with flash, enable something that was not previously not possible.
To get the best performance is all about location, you want to get the native performance of the device
Flash behind the server HBA uses light speed to communicate, yet the industry determined we need faster speeds. We need to be able to remove as much components as possible. Going from networks, to PCI bus, to straight up memory devices.
The stack is comprised of many queues
Solving the Traditional Storage Design Problem
Application agnostic
Flash technology agnostic
Storage infrastructure agnostic
FVP does that in a manor that require no change to the virtual machine, That means No drivers or changes to applications/VMs or static partitioning of server flash.
It requires no change to your storage system and it requires, no new virtual appliance, no new datastores to manage. It’s the same old virtual machines that run on the same old datastores on the same old Storage Array except they are magically faster.
And that magically faster applies to both read intensive virtual machines as write intensive virtual machines.
Solving the Traditional Storage Design Problem
For more information: http://frankdenneman.nl/2014/07/21/stop-wasting-storage-controller-cpu-cycles/
Solving the Traditional Storage Design Problem
When to use memory and when to use flash?
Timing: current chipsets support up to 1.5TB-3TB per host, 6TB (Intel E7-v2) chipsets already announced