Virtualization changes storage in three key ways:
1) I/O is concentrated on fewer ports and randomized across physical disks. 2) Storage utilization is low due to abstraction overhead. 3) Options include shared storage on SAN/NAS, raw device mapping, or NFS with pros and cons for each. Effective virtualized storage requires balancing these changes.
Designing IA for AI - Information Architecture Conference 2024
How Virtualization Changes Storage Requirements
1. How Does Virtualization Change Storage?May 18, 2009 Stephen Foskett, Director of Consulting, Nirvanix
2. Abstract The virtualization of servers destroys everything that storage folks thought they knew about I/O and throws in a new layer of abstraction to boot Creating storage for virtualization is not the same as storage for most other apps, and storage virtual servers on a SAN or NAS is not the same as using internal disk This session will walk through what virtualization changes about storage, the various storage options, pros and cons, and what the future looks like with FCoE, UCS, 10 GbE, and VMware vStorage APIs 2
9. Virtualization As An I/O Engine Server virtualization is the single greatest I/O driver in the modern data center CPU power and memory capacity are easy to ramp up, I/O is not Unbalanced systems will not perform well 5
10. I/O Is Concentrated Then… Each server had its own storage and LAN ports I/O utilization was low Now… 6 All I/O is concentrated on just a few ports LAN SAN LAN SAN
11. I/O is Randomized Then… Now… Sequential I/O is mixed together randomly Disk is virtualized and re-combined 7 I/O was mainly sequential Requests were grouped physically on disk Storage could read ahead and cache data
12. I/O is Accelerated Then… Now… Combined I/O Packets arrive quickly Quicker protocols: 10 GbE, 8 Gb FC 8 Channels were under-utilized with little contention for resources Speeds were low: 1 GbE, IDE/ATA In the same amount of time… 1 GbE handles 1 packet from 1 host... 4 Gb FC handles 4 packets from 4 hosts... 8 Gb FC handles 8 packets from 5 hosts... 10 GbE handles 10 packets from all 6 hosts...
13. Converged Data Center I/O Now… All I/O is concentrated on just a few ports Soon… 9 I/O is converged on 10GbE and extended into server hardware LAN SAN LAN SAN
15. Wasted Space Each level of abstraction adds overhead Overall utilization is low! 11 Raw array capacity Usable array capacity LUNs presented to host Configured datastore Server 1 virtual disk Server 2 virtual disk Server 3 virtual disk Server 1 used capacity Server 3 used capacity Server 3 used capacity
16. Thin Provisioning Thin provisioning allocates storage as-needed Example: 500 GB request for new project, but only 2 GB of initial data is written – array only allocates 2 GB and expands as data is written What’s not to love? Oops – we provisioned a petabyte and ran out of storage Chunk sizes and formatting conflicts Can it thin unprovision? Can it replicate to and from thin provisioned volumes? VMware adding thin provisioning to vSphere 4 (standard at all license levels!) Some storage arrays do thin (3PAR, HDS, NetApp) 12
17. Server Virtualization Demands SAN and NAS Server virtualization has transformed the data center and storage requirements 86% have implemented some server virtualization (ESG 2008) VMware is the #1 driver of SAN adoption today! 60% of virtual server storage is on SAN or NAS (ESG 2008) Server virtualization has enabled and demanded centralization and sharing of storage on arrays like never before!
18. VMware Storage Options:Shared Storage Shared storage - the common/ workstation approach Stores VMDK image in VMFS datastores DAS or FC/iSCSI SAN Hyper-V VHD is similar Why? Traditional, familiar, common (~90%) Prime features (Storage VMotion, etc) Multipathing, load balancing, failover* But… Overhead of two storage stacks (5-8%) Harder to leverage storage features Often shares storage LUN and queue Difficult storage management VM Host Guest OS VMFS VMDK DAS or SAN Storage
19. VMware Storage Options:Shared Storage on NFS Shared storage on NFS – skip VMFS and use NAS NTFS is the datastore Wow! Simple – no SAN Multiple queues Flexible (on-the-fly changes) Simple snap and replicate* Enables full Vmotion Use fixed LACP for trunking But… Less familiar (3.0+) CPU load questions Default limited to 8 NFS datastores Will multi-VMDK snaps be consistent? VM Host Guest OS NFS Storage VMDK
20. VMware Storage Options:Raw Device Mapping (RDM) Raw device mapping (RDM) - guest VM’s access storage directly over iSCSI or FC VM’s can even boot from raw devices Hyper-V pass-through LUN is similar Great! Per-server queues for performance Easier measurement The only method for clustering But… Tricky VMotion and DRS No storage VMotion More management overhead Limited to 256 LUNs per data center VM Host Guest OS I/O Mapping File SAN Storage
21. Which VMware Storage Method Performs Best? Mixed Random I/O CPU Cost Per I/O VMFS, RDM (p), or RDM (v) Source: “Performance Characterization of VMFS and RDM Using a SAN”, VMware Inc., 2008
22. Which Storage Protocol Performs Best? Throughput by I/O Size CPU Cost Per I/O Fibre Channel, NFS, iSCSI (sw), iSCSI (TOE) Source: “Comparison of Storage Protocol Performance”, VMware Inc., 2008 And iSCSI is even better in vSphere 4!
24. Which Storage Protocol is For You? FC, iSCSI, NFS all work well Most production VM data is on FC Either/or? - 50% use a combination (ESG 2008) Leverage what you have and are familiar with For IP storage Use TOE cards/iSCSI HBAs Use a separate network or VLAN Is your switch backplane fast? No VM Cluster support with iSCSI* For FC storage 4 Gb FC is awesome for VM’s Get NPIV (if you can) FCoE is the future Converged storage and networks adapters (CNAs) Cisco UCS
25. Storage in VMware vSphere 4 Thin provisioning is standard for all levels Dynamic expansion of VMFS volumes Any-to-any Storage Vmotion High performance I/O Paravirtualized SCSI Enhanced iSCSI stack Jumbo frames Data Protection APIs (A) Pluggable Storage multipathing (E+) 21
26. The Organizational Challenge How will server, storage, and networking teams deal with integration? Each discipline has its own best practices Each has its own prejudices They can be forced together, but will it work? 22
27. Who Is Nirvanix 23 The Premier “Cloud Storage” Service Provider for the Enterprise Backed by Intel Capital, Mission Ventures, Valhalla Partners, Windward Ventures and European Founders Fund 2007 “Storage Products of the Year”2008 “Top Startups to Watch” 2008 “Product of the Year” Over 500 customers including leading Fortune 10, Media & Entertainment and Web 2.0 companies
28. 24 Thank You Nirvanix We manage your storage, so you can manage your business www.nirvanix.com twitter.com/nirvanix Stephen Foskett sfoskett@nirvanix.com Enterprise Storage Strategies Blog: bit.ly/ESSBlog Personal Blog: blog.fosketts.net Enterprise IT Content: gestaltit.com
Hinweis der Redaktion
Up to 256 FC or iSCSI LUNsESX multipathingLoad balancingFailoverFailover between FC and iSCSI*Beware of block sizes greater than 256 KB!If you want virtual disks greater than 256 GB, you must use a VMFS block size larger than 1 MBAlign your virtual disk starting offset to your array (by booting the VM and using diskpart, Windows PE, or UNIX fdisk)*
Link Aggregate Control Protocol (LACP) for trunking/EtherChannel - Use “fixed” path policy, not LRUUp to 8 (or 32) NFS mount pointsTurn off access time updatesThin provisioning? Turn on AutoSize and watch out