Virtual Ascetic

everything about virtualization, storage, and technology

  • Home
  • About

Your likes and shares help increase my enthusiasm to share knowledge. Please leave a comment if you like any post

Using Virtual Data Services for IO acceleration in vSphere

September 1, 2015 By asceticadmin

#STO5704

I have always been interested in engineering type of sessions at VMworld and this one perfectly fit the bill because there was some deep dive into the mechanics of VAIO. The breakout session started off with description of the data service architecture and how the data service engages host level solid state storage to optimize IO performance for read/write operations.

The Agenda included the following –

  • SDS in vSphere 6: VVOL and VAIO overview
  • Caching filter architecture with VAIO
  • Demo: Caching filter installation, configuration, and performance
  • Demo: vMotion with write-back caching
  • Caching filter and vSphere Cluster functionality

Key benefits of VAIO framework

  • Storage backend agnostic – so works with various types of storage
  • Automated cluster-wide installation – no installation required on each and every host. The upgrade and version enforcement is also applied cluster wide.
  • Comprehensive VMware solutions support – HA, vMotion, storage vMotion, etc
  • Filter configuration through SPBM (Storage Policy Based Management) on a per virtual disk level – if your VM disk fails then the failure does not impact other VMs. The filter failure will be limited to just one VM.
  • VM and virtual disk life cycle management notifications
  • Filters are user level libraries
  • Filters can persist and filter private data along virtual disk

VAIO Architecture

Figure – VAIO Architecture

 

All components can run the filter. The VAIO framework will have the IOs back into user space. Currently, vendors using VAIO need to provide a VC extension, daemon, and CIM provider that will work with the VAIO framework.

Vendor Daemon – collapse all IO requests to the single daemon and gets forwarded and sent back as well. CIM provider offers integration to VC.

Virtual Disk Lifecycle events

VAIO filter - Virtual Disk Lifecycle events

  • VAIO Conceptual design – IO Pipeline
  • VAIO framework sits inside the FDS layer. VAIO framework hands IO to filter.
  • IO filter can directly talk to IO pipeline
  • The IO filter can talk directly to flash or other hw and do IO acceleration
  • Allows you to pool SSD and create caching
  • Caching filter architecture is daemon based
  • Daemon is automatically started and restarted – beauty of it. Unix services have long used daemons for critical services
  • Daemon is provided sufficient resources
  • Daemon diretly starts IO from VM address space
  • VAIO provides Shared memory for communication and data sharing

Caching Filter Architecture

VAIO conceptual design - IO pipeline

Shared memory is used across instances and daemons to reduce latency

Zero memory copy in design – daemon addresses the start of IO from VM memory.

Write back cache is used for dirty data

There are more details to follow – if you are at VMworld and if there is a follow-up session do attend it – Session number #STO5704.

Filed Under: VMWorld Tagged With: anil sedha, caching filter, EMCElect, filter daemon, IO acceleration, IO filter, shared memory, vExpert, vitual data services, vmworld 2015, vsphere, write back cche

Let’s Get Social

  • Google+
  • Linkedin
  • Twitter


Recent Posts

  • VMware’s Photon OS 1.0 Now Available!
  • VMware’s Photon OS 1.0 Now Available!
  • VMware’s Photon OS 1.0 Now Available!
  • VMware’s Photon OS 1.0 Now Available!
  • VMware’s Photon OS 1.0 Now Available!

Archives

  • June 2016
  • May 2016
  • April 2016
  • February 2016
  • September 2015
  • August 2015
  • February 2015
  • January 2015
  • October 2014
  • September 2014
  • August 2014
  • May 2014
  • April 2014
  • March 2014
  • February 2014
  • January 2014
  • October 2013
  • September 2013
  • August 2013
  • June 2013
  • May 2013
  • April 2013
  • March 2013
  • February 2013
  • January 2013
  • November 2012
  • October 2012