VMUG Virtual Event – Technical Chat with Cormac Hogan about new features in Virtual SAN

vmug_virtual

This is a transcript of the technical chat with Cormac Hogan about the new features and functionality in VSAN. It was captured from the Feb 17th VMUG Online Virtual Event.

VMUG Virtual Event 2015

Scheduled Chat –Cormac Hogan discussing Virtual SAN
All times shown in CST

100 Most Recent Messages | All Messages

[ 02/17/15 11:30 AM] Cormac Hogan (VMware): Good morning/afternoon/evening – we’ll get started in just a moment
[ 02/17/15 11:30 AM] Cormac Hogan (VMware): Lots of new stuff to talk about over the next 30 minutes
[ 02/17/15 11:31 AM] Cormac Hogan (VMware): VSAN, VVOLs (yes, who would’ve believed it) and some new CORE vSphere storage features too
[ 02/17/15 11:32 AM] Cormac Hogan (VMware): So I think I will be begin with some new VSAN stuff
[ 02/17/15 11:32 AM] Cormac Hogan (VMware): I guess from a scalability perspective, the next release of VSAN will go to 64 nodes, and support 62TB VMDKs
[ 02/17/15 11:33 AM] Cormac Hogan (VMware): We’re also going to fianlly make use of the Virsto acquisition we made a number of years ago
[ 02/17/15 11:33 AM] Cormac Hogan (VMware): there is a new on-disk format for VSAN using VirstoFS – so we’ll have better snapshot and clone performance than before
[ 02/17/15 11:34 AM] Cormac Hogan (VMware): Of course there is also the all-flash version of VSAN coming too
[ 02/17/15 11:34 AM] Cormac Hogan (VMware): iirc, we were getting 90K IOPS hero numbers from a single node
[ 02/17/15 11:34 AM] Cormac Hogan (VMware): from an availability perspective too, we’re getting fault domains
[ 02/17/15 11:35 AM] Nikki Kueker – VMUG HQ: @Everyone, please feel free to ask Cormac your questions!
[ 02/17/15 11:35 AM] Catalin.Bostan: why new fs?
[ 02/17/15 11:35 AM] Steve.Smith: @Cormac – so the 90K IOPs is for all-flash VSAN?
[ 02/17/15 11:35 AM] Cormac Hogan (VMware): through fault domains, we have the ability to place virtual machine replica copies on different racks..
[ 02/17/15 11:35 AM] Cormac Hogan (VMware): new fs question first …
[ 02/17/15 11:35 AM] Wouter.Kursten: that’s hot, now only to convincce our customers to trust it. That still seems to be the biggest thing I have run into proposing vsan: how to convince the customer it’s trustworthy
[ 02/17/15 11:36 AM] Cormac Hogan (VMware): its really for performance – the virstoFS snap and clones is superior to the v1 on-disk format in version 5.5
[ 02/17/15 11:36 AM] Catalin.Bostan: i want to ask what kind of checksum new fs use
[ 02/17/15 11:36 AM] Cormac Hogan (VMware): Yes @Steve – the 90K was All-Flash
[ 02/17/15 11:36 AM] Cormac Hogan (VMware): but we also doubled up on the IOPS figures for hybrid between 5.5 and 6.0
[ 02/17/15 11:37 AM] Cormac Hogan (VMware): I hear you @Wouter
[ 02/17/15 11:37 AM] Cormac Hogan (VMware): The product is only just 1 year old – I can understand customers reluctance around that
[ 02/17/15 11:37 AM] Catalin.Bostan: i use zfs on storages-full checksum fron wire to disk–paranoic but..effective,so what method virstoFS use?
[ 02/17/15 11:37 AM] Cormac Hogan (VMware): There is none on the filesystem
[ 02/17/15 11:38 AM] Cormac Hogan (VMware): But there is a hardware based checksum available via the VSAN Ready Nodes
[ 02/17/15 11:38 AM] Catalin.Bostan: at what level on the hardware is checksumming performed?
[ 02/17/15 11:38 AM] Matt.Osburn: Does VMware actually intend for vSAN to replace commodity storage long-term? Or are they looking to keep it niche…
[ 02/17/15 11:39 AM] Stan.Tracy: HI
[ 02/17/15 11:39 AM] Julian.Miranda: when speaking about VirstoFS, is this a complete replacement of VSAN’s VMFS-L filesystem?
[ 02/17/15 11:40 AM] Catalin.Bostan: well , i think intention is to crush netapp and EMC…and many other..
[ 02/17/15 11:40 AM] Cormac Hogan (VMware): @Matt – I’m not sure if there is a plan like that … but maybe we can return to a more technical discussion?
[ 02/17/15 11:40 AM] Cormac Hogan (VMware): The other items that we introduced are really around operations
[ 02/17/15 11:40 AM] Cormac Hogan (VMware): making it easier to identify and replace compoents, etc.
[ 02/17/15 11:41 AM] Jonathan.Sherman: Is there a clean method to implement an all-vSAN platform without leveraging non-vSAN storage during the implementation?
[ 02/17/15 11:41 AM] Zach Schaffner VMUG Leader: when setting up vsan is it best to keep identical disk layout in all hosts meaning 1 ssd and 1 spindle of the same size
[ 02/17/15 11:41 AM] Cormac Hogan (VMware): @Julian, yes, VirstoFS is the replacement for VMFS-L, although the new release can run with either v1 or v2
[ 02/17/15 11:42 AM] Julian.Miranda: thanks… upgrade considerations?
[ 02/17/15 11:42 AM] Cormac Hogan (VMware): To get the benefits of VirstoFS fast snapshots and cloning, the on-disk format should be v2
[ 02/17/15 11:42 AM] Matt.Osburn: Are there min-latency recommendations when physically placing vsan hosts in a cluster?
[ 02/17/15 11:42 AM] Jeremy.Nofs: What is the brief overview how an upgrade would go?
[ 02/17/15 11:42 AM] Cormac Hogan (VMware): rolling upgrade supported – evacuate all data from one node at a time,
[ 02/17/15 11:43 AM] Bill.Halverson: What would you recommence for a base test lab to implement VSAN? Are there any specific papers related to setting up a test environment?
[ 02/17/15 11:43 AM] Jeff.Crisconi: there are a lot of similar capabilities between what can be done in VMFS and vVOLs, snapshotting, thin provisioning etc. is there a recomendation on which provider to use?
[ 02/17/15 11:43 AM] Cormac Hogan (VMware): @Matt, no real latency considerations per se. Our recommendation is for everything to be on the same L2
[ 02/17/15 11:43 AM] Catalin.Bostan: ok, what you think is best quality of VirstoFS–killler feature?
[ 02/17/15 11:43 AM] Jonathan.Sherman: And is anyone aware of plans to support 2-node clusters? The minimum 3-node requirement is drawing attention to non-vSAN virtual storage solutions.
[ 02/17/15 11:44 AM] Cormac Hogan (VMware): @jeremy – host one enter mmode, evacuate host one, upgrade it, bring it back into cluster, exit mmode, move to node two, repeat and rinse
[ 02/17/15 11:44 AM] Cormac Hogan (VMware): @Bill – check out William Lam’s virtuallyghetto sites for lab setups, nested environemnts, etc
[ 02/17/15 11:45 AM] Cormac Hogan (VMware): @Jeff on VVols, nope, no recommendation. I would just see what each vendor has to offer capability wise
[ 02/17/15 11:45 AM] Cormac Hogan (VMware): @Jonathan – we cannot discuss futures unfortunately
[ 02/17/15 11:45 AM] Cormac Hogan (VMware): @Catilin – the improved snapshot and cloning features for sure – far superior to VMFS
[ 02/17/15 11:46 AM] Julian.Miranda: When you upgrade the host to ESXi 6, does the hypervisor automatically use v2 for vSAN?
[ 02/17/15 11:46 AM] Cormac Hogan (VMware): @Zach – yes, definitely go for uniform configurations
[ 02/17/15 11:46 AM] Cormac Hogan (VMware): @Julian – nope, it is an independant step
[ 02/17/15 11:47 AM] Roger.Dills: External storage? supported but without scary caveats?
[ 02/17/15 11:47 AM] Cormac Hogan (VMware): New installs will be v2, but upgrades require an additional on-disk format upgrade, which can be done with rolling upgrade mechanism
[ 02/17/15 11:47 AM] Zach Schaffner VMUG Leader: can you use two different sized disks on the hosts. Looking at having a 256gb SSD and then I have a 146gb and 500gb in each host.
[ 02/17/15 11:47 AM] Cormac Hogan (VMware): @Roger – they will be, but please check HCL as there are only some that have been certified
[ 02/17/15 11:48 AM] Chestin.Hay: HI
[ 02/17/15 11:48 AM] Cormac Hogan (VMware): Its possible, but you could have some inbalance. We do highly recommend uniform if at all possible
[ 02/17/15 11:48 AM] Cormac Hogan (VMware): So shall we talk about some CORE stuff?
[ 02/17/15 11:49 AM] Cormac Hogan (VMware): Anyone excited about NFS v4.1?
[ 02/17/15 11:49 AM] Catalin.Bostan: yeah
[ 02/17/15 11:49 AM] Cormac Hogan (VMware): Thats coming in 6.0 too.
[ 02/17/15 11:49 AM] Jonathan.Sherman: yes!
[ 02/17/15 11:49 AM] Catalin.Bostan: me šŸ™‚
[ 02/17/15 11:49 AM] Roger.Dills: The HCL Grail, and stick to it
[ 02/17/15 11:49 AM] Bill.Batchelor: yes
[ 02/17/15 11:49 AM] Cormac Hogan (VMware): That’s cool – its been a long time coming šŸ™‚
[ 02/17/15 11:49 AM] Julian.Miranda: šŸ™‚
[ 02/17/15 11:49 AM] Cormac Hogan (VMware): We also did some cool stuff around Storage DRS
[ 02/17/15 11:49 AM] Bill.Halverson: With the inovations of V6 would it be vaibale for us to consider VMware / VSAN across Campuses with 300MB connections with less then 100MS latency? Thoughts, considerations?
[ 02/17/15 11:50 AM] Cormac Hogan (VMware): @Bill – no strech clustering this time, but we’re looking at it seriously
[ 02/17/15 11:50 AM] Catalin.Bostan: is full compliant with rfc5661 or not šŸ™‚
[ 02/17/15 11:50 AM] Jorgen.Sandtorv: I use 2×100 in one host and 1×200 in the 2 other test hosts I use @Zach – work fine…
[ 02/17/15 11:50 AM] Cormac Hogan (VMware): On Storage DRS, if your storage array bendor implements VASA 2.0, SDRS will be able to understand datastore that are deduped, and thin provisioned, and running auto-tiering
[ 02/17/15 11:51 AM] Cormac Hogan (VMware): Oh, and by the way – VMFS-3 is now deprecated in 6.0, ok?
[ 02/17/15 11:51 AM] Cormac Hogan (VMware): But no-one is running VMFS-3, right? šŸ˜‰
[ 02/17/15 11:51 AM] Catalin.Bostan: heheh
[ 02/17/15 11:52 AM] Cormac Hogan (VMware): another cool thing – 6.0 has a file system checker for VMFS-5
[ 02/17/15 11:52 AM] Cormac Hogan (VMware): its called VOMA
[ 02/17/15 11:52 AM] Jorge.Torres: I’ve heard NFS multipathing; similar network configuration as iSCSI multipathing?
[ 02/17/15 11:52 AM] Cormac Hogan (VMware): yes – NFS multipathing for sure, which is not pNFS though
[ 02/17/15 11:52 AM] Wouter.Kursten: @cormac I am running a 4.1 environment untill the end of the month with vmfs 2 but please don’t ask LOL
[ 02/17/15 11:52 AM] Cormac Hogan (VMware): LOL
[ 02/17/15 11:52 AM] Cormac Hogan (VMware): I won’t
[ 02/17/15 11:52 AM] Catalin.Bostan: too bad šŸ˜¦
[ 02/17/15 11:53 AM] eric.hedin: what’s new in nfs 4.1?
[ 02/17/15 11:53 AM] Cormac Hogan (VMware): multipathing and security
[ 02/17/15 11:53 AM] Catalin.Bostan: and what about 4.2 šŸ™‚ is a draft from august i think
[ 02/17/15 11:54 AM] Cormac Hogan (VMware): you don’t have to rely on AUTH-DES anymore (you know, where you use no-root-squash)
[ 02/17/15 11:54 AM] Jorgen.Sandtorv: I found an esx 3.5 in our remote locations the other day.. uptime 1600 days..
[ 02/17/15 11:54 AM] Cormac Hogan (VMware): now you can use AD for accessing NFS volumes
[ 02/17/15 11:54 AM] Cormac Hogan (VMware): nice
[ 02/17/15 11:54 AM] Jorgen.Sandtorv: so esx is still good
[ 02/17/15 11:54 AM] Cormac Hogan (VMware): So what else can I tell you in the last 5 minutes?
[ 02/17/15 11:54 AM] Catalin.Bostan: ok
[ 02/17/15 11:55 AM] Cormac Hogan (VMware): Oh – there’s a nice new feature for failing over a path if there are too many I/O retries
[ 02/17/15 11:55 AM] Jorgen.Sandtorv: You recomend keeping all the vsan nodes on the same ToR?
[ 02/17/15 11:55 AM] Cormac Hogan (VMware): you might know that we failover a path based on certain scsi sense codes
[ 02/17/15 11:55 AM] Cormac Hogan (VMware): and that certain errors do not cause a failover
[ 02/17/15 11:55 AM] Catalin.Bostan: when infiniband stack will be supported properly on esxi..
[ 02/17/15 11:55 AM] Jorgen.Sandtorv: I found that it can be tricy if one of the hosts is located in another rack..
[ 02/17/15 11:55 AM] Catalin.Bostan: and get rid of FCoE..
[ 02/17/15 11:55 AM] Cormac Hogan (VMware): this new enhancements allows you to set a limit, and it if is surpassed, we fail over
[ 02/17/15 11:56 AM] Jorgen.Sandtorv: havent figured out what couse the troubble though
[ 02/17/15 11:56 AM] Cormac Hogan (VMware): @Jorgen – yes to make multicast easier to setup, but its not always possible
[ 02/17/15 11:56 AM] Cormac Hogan (VMware): @Catalin – no insight into that sorry
[ 02/17/15 11:56 AM] Catalin.Bostan: well nor vmware nor mellanox..too bad
[ 02/17/15 11:57 AM] Cormac Hogan (VMware): see that a few times now
[ 02/17/15 11:57 AM] Catalin.Bostan: nfs over infiniband rdma–yumm
[ 02/17/15 11:57 AM] Cormac Hogan (VMware): haha
[ 02/17/15 11:57 AM] Jonathan.Sherman: I work with a service provider who would love to use infiniband on the back-end. That would be great.
[ 02/17/15 11:57 AM] Cormac Hogan (VMware): ok – a few minutes left
[ 02/17/15 11:57 AM] Catalin.Bostan: i have that under linux..amd solaris
[ 02/17/15 11:57 AM] Jorgen.Sandtorv: ahh.. that could be it… We’re in a silobased compay so network gusy are the enemy
[ 02/17/15 11:57 AM] Catalin.Bostan: i want under vmware
[ 02/17/15 11:57 AM] Jorgen.Sandtorv: cant wait to put in NSX
[ 02/17/15 11:57 AM] Cormac Hogan (VMware): are there are more questions about storage in 6.0?
[ 02/17/15 11:58 AM] Cormac Hogan (VMware): VSAN? VVOLS? CORE stuff?
[ 02/17/15 11:58 AM] Jorgen.Sandtorv: thx for the chat @Cormac
[ 02/17/15 11:58 AM] Cormac Hogan (VMware): my pleasure – hope there were some useful nuggets there
[ 02/17/15 11:58 AM] Wouter.Kursten: indeed, thanx @Cormac
[ 02/17/15 11:58 AM] Jorgen.Sandtorv: sure was
[ 02/17/15 11:58 AM] Cormac Hogan (VMware): i put out stuff from time to time on my blog
[ 02/17/15 11:58 AM] Catalin.Bostan: yep nfs 4.1
[ 02/17/15 11:59 AM] Cormac Hogan (VMware): cormachogan.com has a bunch of new stuff for 6.0 if you get a chance to check it out
[ 02/17/15 11:59 AM] Cormac Hogan (VMware): ok – thanks everyone

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s