Posts by Lost_Signal 27 publicly visible posts • joined 24 Mar 2015
Houston's a 61 today and partly cloudy. The hot months out of the year we:
1. Live in Air Conditioning
2. wear as little as possible when outside (or go to the large shaded parks, or do water activities).
3. We go out to bars/restaurants later. Patio bars are everywhere and in some extreme cases like Midtown they will outright air condition the outdoors. (As well as deploy giant fans, shade tarps and misters).
We have very mild winters as a pro.
Disclaimer VMware employe, posts are my own blah blah blah.
EMC could say what they wanted but that hasn't been a VMware position since the 2nd release of the product (on the 6th version now!).
EMC leadership has publicly admitted they got position wrong (Likely while they blew out their sales quota's for VxRAIL last year).
Chad Sacak's clears the air by saying...
We totally got the initial positioning wrong. We thought of HCI Appliances like VxRail as “SMB/Robo/Enterprise Edge” and HCI Rack Scale systems with network/SDN in scope as “Data Center”. This is a bit of a “legacy” way to think, and we got schooled pretty quickly by the customers and the market.
http://virtualgeek.typepad.com/virtual_geek/2017/02/happy-birthday-vxrail-what-a-year-one.html
Market positioning changes, and what customers want always wins out. I have to give Chad respect for admitting that the positioning was off, and Dell-EMC like any company follows the customers...
Full disclosure current VMware employee, but my posts are my own and this is my 2 cents on what happens...
So when I worked in the channel for a partner we could sometimes get our storage partners to go negative on a deal if there was a business case made that there would be expansion in the next 6 months that would bring the deal positive (Sell array below cost, and then 87% discount vaporizes on expansion, or support renewal in year 4 spikes).
My understanding of Hyperflex is unique in that you can't renew the hardware support without paying Opex pricing for the software (also the software MUST be licensed to run a bit like Meraki in that you can't keep running it without paying a renewal). In this case for customers who's TCO calculating skills involve looking at PURELY capex will be able to buy it below the cost of the hardware up front and make it up (The razor blade model) on renewals. On top of this there is so much profit in a Nexus 7K refresh that if you bundle it with a FI/5K/7K refresh you can sell a 50K Hyperflex bundle for 40K because it's bundled in with a 400K Routing and Switching refresh that's still full of profit and massage the numbers so that they get less discount if they don't buy the hyper flex with it. I'll also note that storage array vendors have been doing this for YEARS. Array renewal time? If you go with a competitor you'll miss out on 80% off on the backup software, or dedupe target renewal savings that we can't give you. Hell even Microsoft and oracle get in on the fun of making your ELA more expensive to renew if you don't agree to book some Azure credits so they can show cloud growth.
I've spoken with a few Cisco partners, SE's and customers and in my (anecdotal experience) it sounds like aggressive bundling, or account seeding accounted for every booked deal I'd spoken with.
I'm not going to argue that bundling is bad for the customer (More discounts, fewer vendors to talk to), but when the costs get hidden and fudged together like this eventually there is a reckoning If they realize for the TOTAL value they are getting it's not what they could do if they spilt it up and went best of breed or wholesale went to another shop. Either way, the biggest losers in these deals long term are always the small and up and coming guys who don't have anything else to bundle (Say your only storage, and don't have a VDI platform or a hypervisor) and don't have enough differentiation to show why a premium should be paid for "unbundling".
Full disclaimer vSAN Tech Marketing here (@Lost_Signal on twitter).
Would it help if the version numbers were reduced to x.x.y releases for major features so you can feel like it's a "mature release" but you get the new hot goods?
While vSAN does have file system, I'd like to point out that it does NOT have or use a clustered file system. It's an object storage layer built on top of local file systems (Disk groups). This allows for retooling under the hood, and scaling to the moon while maintaining low overhead and incredible resiliency.
The alternative is you buy a storage product either at the beginning of it's life cycle (Bleeding edge!) or at the end (Your deploying SAS based SSD's while everyone else is talking about NVMe). I'd argue with vSAN you have a lot more flexibility in that you can choose your hardware (Stay a version behind, or deploy the new 3D XPOINT now!), you can choose your software vintage (The 6.0 branch of ESXi, or the 6.5 branch) depending on what you are comfortable deploying.
If you like a slower release cadence, and a more "fixed" storage array like consumption model have you looked at Dell-EMC VxRAIL, or HDS's UCP-HC. They both add extra lifecycle management on top of vSAN, and have their own vendor specific QE cycles added to the mix with hand holding hardware/software single number support. With 2 (maybe more?) OEM appliance offerings, and 15 server vendors and hundreds of certified devices you've got a LOT of options (depending on what your looking for).
"HyperFlex is the only hyperconverged solution that allows native snapshot integation with Veeam"
Not exactly...
vSAN and Veeam are supported together and Veeam leverages Native VMware vSAN snapshots (introduced in 6.0, a replacement for the VMFS Snapshots). There is also additional integration (Agents poll the object store directly to do data proximity on backup jobs to lower networking overhead) as well as SPBM integration (Veeam can backup and recover SPBM policies with vSAN).
I'm curious if this this is LUNs only to EC2, or are there plans to get vVols into VMW on AWS?
While it's easy to point to the array's uptime, what about the SLA of the network. I vaguely remember someone at AWS saying it was impossible to hit 5 nines of uptime with networking gear within an AZ.
You talk about de-risking and making troubleshooting easier, but is their any type of performance visibility on the hops between you and the host (Something like Thousand eyes but for layer 2, maybe some SNMP/NetFlows?).
Dave, note that Horizion view advanced includes VSAN for free. VSAN for desktops is the SKU for Citrix if you go down that path. I've built some VSAN/VDI clusters and it never took me more than a few days all in (rack, stack, build cluster, build VDI templates, setup VDI, create GPOs etc). Realistically this is all something some good PSO can take care of..
You don't need to be a VCP to deploy ESXi on a stand alone host, and replicate it back to your main datacenter. It comes pre-installed when you buy a server from Dell or most OEM's, and there is a pretty HTML5 interface now on the hosts to manage it.
As far as replication, vSphere Replication, or Veeam (or a dozen other easy to use systems) can replicate that VM from the site back to your main site, and in the case of ROBO you get the uber cheap ROBO licensing to cover it (pay per VM for small sites).
Last time I checked It still requires a Windows Kernel to run a VM in (and pass IO through) much like the DOM0 in Xen. In hyper-V server its stripped down, but still there (just really limited).
I would argue KVM is actually a more elegant design (What's the point of having a hypervisor if you must hairpin all IO through a management VM?!)
So how do all these hyperconverged people handle an outage of the entire cluster?
Can only speak for VMware's VSAN but...
1. Fault Domains (Split cluster across cabinets or data center pods).
2. Stretched Cluster (VMware HA fires off VM's on remote site)
3. vSphere native Replication with compression and SRM to orchistrate the failover.
4. One of the 100x other products that does virutalization backups or replication...
I'd argue Physical disasters (That are limited in local scope) can be handled by HCI and replication systems. I'm not arguing HCI is the solution to everything and some idea's for DR are kinda silly (Push backups to AWS seems to be a weird one that people think can quickly be turned into something useful).
In recent years it's messed up sizing for VSAN?
The only case of sizing for VSAN I've seen change was a change from a 10% of RAW, vs 10% of actual data (Which would imply if you followed the advice from the beta period when this was changed, you'd potentially have more cache than you need which isn't really a bad thing and was always listed as a rule of thumb).
"which doesn't seem to be called VSAN any more"
Technically the product name has been "VMware Virtual SAN" since launch. (VSAN if you want to be pedantic is a bit like a VLAN for a fibre channel fabric). You don't have product names that you can't trademark. VSAN is a very common abbreviation for it as it does roll off the tongue quite a bit better than a 3 word phrase....
Ken in Houston, expanding LUNs with vSphere VMFS and Windows (2008r2 at least) doesn't require downtime. 10 year old EVAs could do this.
I agree that LUNs suck, but NFS mount points can still cause issues for management. If you don't want LUNs vvols and VSAN with storage based policy management both have you covered (no plugins needed!).
John-In-Houston :)
1. I"m curious who are they getting SLC flash? High endurance flash today (Like an Intel S3700 that has an impressive 10DWPD) is eMLC. I know Micron and some will custom make whatever you want, I'm just curious for endurance reasons why you would go SLC when eMLC can do quite a lot.
2. Essentials Plus is ~6K for the capex and ~1K a year for support. Saying the "vTax" is significant is a bit of a stretch for
3. Having seen a VSA use 8 Core's (and half of a HCI host) I agree that the VSA architecture has problems scaling (Either out or down). Kernel implemented storage makes the most sense for HCI for CPU and memory reasons.
4. I have a lot of respect for the Scale guys. They are not following Go 4 billion or bust plans of other startups in this space, but a methodical customer focused route. I wish them well, as I think this would be a better model for taking care of customers than many of the existing startup models.
3rd party plugins (Everyone would rather write HTML5).
The C# client hasn't been properly updated for new features in years. Seriously if your using it today with 6.0 its time to stop.
The HTML5 embedded host client replaces the C# for fixing a broken vCenter. Its already GA (the newest build can be found on the fling site here).
https://labs.vmware.com/flings/esxi-embedded-host-client
The HTML5 web client for vCenter is on the fling site and can be found here.
https://labs.vmware.com/flings/vsphere-html5-web-client
The team is iterating quickly. The flash client will stick around briefly as it fleshes out, but its an agile team moving rapidly with quick releases. People are sick of flash (Security/Speed) and with the industry is preparing to deprecate flash (and HTTP) this is a serious customer demand.
I've always found that Federal is the #1 target for Hyperconverged. If there is anyone with legendarily slow deployment issues its Gov sector. An easy button for infrastructure will win anytime vs. the 5 sub deep labor at $500 an hour, paid $25 an hour in reality. This isn't just Nutanix, everyone in this space should be selling in the federal space, because its an arbitrage of slow confusing procurement and bloated labor to a SingleSKU and a day to deploy.
Correct. This short article wasn't meant to settle the NO-SQL vs SQL argument for good (or even address it!) it was to talk about reducing and filtering data on ingestion being a good tool that everyone should view more seriously. As always, your mileage may very.
Good catch!
Looks like some 9's got dropped in the shuffle (as well as looking around many secondary sources had it wrong).
Lets try this again shall we?
L1 filtering 40Mhz to ~60-65Ghz (so ~.015% data retained).
L2 filtering 65Khz to 6Khz so (10% of data retained)
L3 filtering 5-6Khz to 500-600hz so (10% of data retained). So 99.99995 % of data is filtered (well before further passes are made at the data). (This math check out?)
As much as I love arguing about percents, the point is a lot of companies are hoarding the inverse (99.99% rather than 0.01, or 0.000001%), something they can "afford" to do at small scale when they don't try to do anything useful with the data, but will not scale as volumes grow.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
