Why is VMWare considered so often
-
@coliver said:
@Breffni-Potter said in Why is VMWare considered so often:
Well this is new.
http://blogs.vmware.com/vsphere/2016/05/goodbye-vsphere-client-for-windows-c-hello-html5.html
They've been talking about this for awhile now. We ran into it in the past year where they told us they wouldn't be updating the desktop client to work with TLS1.1 or 1.2.
You've got a ton of options for management of hosts.
HTML5 native, SSH, PowerCLI (PowerShell), RESTFUL API's with SDK's for Python, C, Java and a bunch of other languages, SOAP API. If you want a thick client just use Fusion/WorkStation (What I use at my house for quick VM/console access). There's other interfaces you can use too (vRA, VCD have their own).
-
@John-Nicholson said in Why is VMWare considered so often:
@DustinB3403 said in Why is VMWare considered so often:
@Carnival-Boy I must've missed that part.
But still the justification of cost for such a tiny deployment of a single host and 2 VM's to have to pay even the $500 seems insane.
Why do you need the Backup API's for 2 VM's. Just use in guest agent backup tools (Unitrends, ShadowProtect). There are backup options for ESXi Free (Trilliad, and GhettoVCB).
The backup API's (CBT/VADP) are only valuable at scale, and given that Hyper-V hasn't had a CBT API Until 2016 (which isn't wildly supported yet by backup products) Arguing that you NEED essentials license to get a comparable functionality isn't actually true. Xen doesn't have a CBT based backup API (Doing external backups still requires doing a full read). XO does have a non-application consistent forever reverse incremental snapshot system, but that's still going to require a full read of IO for synthetic roll ups, and the lack of application consistency makes it un-usable for transnational workloads.
For some reason everyone on SW and ML seems to think that the Backup API's are CRITICAL when your talking about a tiny SMB (they aren't) and ignores their value at scale (Where they are critical to reducing backup windows by 95%). I've never really understood this...
I'm not sure to understand if you understand (^^) XO delta backup.
- We are using quiesce snapshots by default (if the VM support it, it will be consistent on VM level).
- What do you mean about "require a full read of IO for synthetic roll ups"? We are exporting only the changed blocks between 2 snapshots (therefore, generating a small VHD file). Then, the oldest delta VHD will be merged into the full VHD, on the remote storage (so without the XenServer assistance).
- We don't have any report of issues in restore even for DBs.
edit: you are a vendor from which company?
-
@olivier He works with VMware...
-
-
Disk based quiescence alone isn't enough for support for Microsoft applications (Microsoft demands you do full app level quiescence). You could do scripts to stun the service and trigger log flushes but at this point your starting to have to aggressively track changes in every patch and version (this is why backup software is expensive, all the QA work).
-
Didn't realize you could do a remote synthetic roll up. If your not doing compression /dedupe that's still a lot of IO over the wire. I've seen smaller shops keep 20TB's backed up using a 1Mbps upload link with Veeam. CBT with a agent that does data reduction local, then hits a WAN accelerator cache can reduce transport by a lot. (You could buy a silver peak VM to run stuff thru but that's not the cheapest solution to this). Bandwidth isn't free yet for everyone sadly.
-
Crash consistency isn't as ugly as it used to be. The bigger issue is time to restore can take forever if SQL/Exchange have a lot of logs that need to be replayed. I've had to stare at a crash consistent DB restore for 4 hours before it came back online (was fine, no corruption, just SQL didn't see itself as being properly backed up). Also, without the backup software being application integrated your not doing a log flush (and transaction logs will eventually build up and fill the file system). You can work around this by doing local application level backups to a VHD (Then backing that up), but that's a LOT more data/IO to move/store (especially without dedupe).
-
-
@Dashrender said in Why is VMWare considered so often:
Wow.. that maintenance costs seems a bit high, but I guess it's only $311 per processor. I'll have to lookup what my customer just paid for their maintenance.
When its 3AM and you have to restore something and need help you don't question what you paid for backup vendor support. It's one of those things I couldn't imagine going without.
Years ago, I had a DRDB cluster go split brain on me, and not having real enterprise support to deal with the issue (and mess of sorting the data back together) made me realize why storage/backups are normally something you normally have enterprise support vs. build your own. I lost 3 days of my life to that mess and still want it back...
-
@John-Nicholson There is probably a lot of things you don't realize about XO, but I don't bother to be forgotten in a niche market
-
Even when quiesce is calling the VSS service (via the guest tools of XS)? if it's not enough, it's broken from my point of view. Hint: I'm a Linux/BSD guy, and I'm often baffled by (dis)order of magnitude when you are dealing with MS world. But that's my turn to not realize a lot of things about this, you are the pro on that! PS: about why backup is expensive: is it your opinion as a VEEAM or VMWare developer?
-
Yes, despite XO is a recent backup products, we are able to execute quite fast
And in fact, it's in the pipe to have "proxies" on remote site to improve the global traffic flow in a distributed architecture. Adding compression into it could be a good idea indeed. Thanks for the tip, this will be trivial to implement I'll open an issue to not forget that! (WAN accelerator is also a cool name for this!) -
Bigger issue + MS apps in the same sentence: I trust you on this. For my curiosity, cf my first question in 1
Anyway, if you are from VMWare, I'll be glad to ask you some questions regarding licensing, segmentation is crazy and I have some difficulties to compute in head to head scenario the cost of XO/XS vs VMWare/VEEAM. I'll send you a message in private if you have a bit of time
Thanks! -
-
@John-Nicholson said in Why is VMWare considered so often:
Years ago, I had a DRDB cluster go split brain on me
This, totally agree. Before adopting a product in a real prod env, you have 2 choices:
- pay for a turnkey solution (packaged with support)
- install it by yourself ONLY after having enough knowledge on how it works.
As you said earlier, you can't master something in 4h. You need to practice, validate, crash it, restore it etc.
-
@scottalanmiller said in Why is VMWare considered so often:
@TAHIN said in Why is VMWare considered so often:
VMware needs to realize that their pricing model is losing thousands of potential customers.
They realize and they don't care. I mean that literally. They basically don't want to deal with the "freebie" market and use their price model as a means of eliminating it. They only sell VMware, so going to something free like their competitors would leave them with essentially nothing to sell.
The pricing model is actually pretty damn cheap if you think about it...
The free product works fine for people who use it and an Essentials Plus kit can easily run up to 200 VM's for 6K up front and ~1K a year for 24/7 support on 3 hosts. That works out to ~$3 a day per host for 24/7 support of a hypervisor, build in backup and replication software (With agents for SQL/Exchange/Sharepoint/Dedupe etc), central monitoring and email alerts, historic performance monitoring and a HTML5 interface. The management software doesn't require a Windows Server or SQL database anymore (less licensing than SCCM/SCOM) .Let's quit being dramatic and Lets go over thing's that are more expensive than the daily support cost per host here is...
- My wife's Starbucks addiction.
- My household booze budget/Sams Bar Tabs.
- What you spend to go to a single IT conference.
- Less than 1 minute of my time a day billed at my standard rate when I was consulting.
If you think a 6K capital purchase and 1K a year for 3 hosts support is "OUTRAGEOUS".
Assuming your a standard sysadmin paid 70K a year ($35 an Hr, with a 20% overhead fringe cost, so closer to $42 an hour cost to the business) your looking at the cost in your time being 10 minutes. If it can save you close to 10 minutes a day with all the management/features (Ignoring other values like Backup software, support to help cut the time of an outage in 1/2 etc) then it pays for itself. If you try to price out 24/7 engineering support (who can get developers to issue a hotfix) for other platforms (Hyper-V, Xen/KVM from RedHat etc) I'm fairly certain that VMware has the cheapest option here.
-
@olivier said in Why is VMWare considered so often:
@John-Nicholson said in Why is VMWare considered so often:
Years ago, I had a DRDB cluster go split brain on me
This, totally agree. Before adopting a product in a real prod env, you have 2 choices:
- pay for a turnkey solution (packaged with support)
- install it by yourself ONLY after having enough knowledge on how it works.
As you said earlier, you can't master something in 4h. You need to practice, validate, crash it, restore it etc.
See we can all get along. One thing that Xen/XenServer has to go up against is the massive amount of training and operational experience that is in the field. It reminds me of the linux desktop. Linux makes a decent desktop, but without 3rd party vendors seeing value (and porting apps) and existing staff seeing value in re-training) it's hard to get larger adoption. You'll have to not only reach feature parity (and cost) but actually have "killer app" incentives to switch (of which drives VMware R&D to adapt). This is the arms race of our industry (and honestly we all benefit from this, vendors, customers, and end users...).
-
@John-Nicholson I think you are mixing Xen and XenServer here.
-
@olivier said in Why is VMWare considered so often:
Anyway, if you are from VMWare, I'll be glad to ask you some questions regarding licensing, segmentation is crazy and I have some difficulties to compute in head to head scenario the cost of XO/XS vs VMWare/VEEAM. I'll send you a message in private if you have a bit of time Thanks!
With Licensing I"m reminded of a quote from Balmer.
Sure we could simplify our licensing but SOMEONE would loose out and have to pay more....
Fun fact, there is a Opex SKU for Essentials Plus that everyone forgets about. -
So you edited your post and now I don't agree with you
You can have support for XS and also support for XO (our pricing model is based on a dedicated appliance + support).
-
@olivier said in Why is VMWare considered so often:
@John-Nicholson I think you are mixing Xen and XenServer here.
You are correct. BTW, first time I tried Xen was with Solaris as the DOM0. It was... interesting. It was powerful but operationally it was a mess. Oddly now I see more Xen adoption in the field from Oracle XVM than XenServer (Purely my anecdotal recent meetings with larger customers, who are doing it to deal with Oracle Licensing FUD)
-
@John-Nicholson I think indeed you are not yet leading the licensing FUD, Oracle seems always better than anyone else on earth.
-
@olivier said in Why is VMWare considered so often:
@John-Nicholson I think indeed you are not yet leading the licensing FUD, Oracle seems always better than anyone else on earth.
I've never met anyone at VMware ACTIVELY try to over license people with threats of lawsuits or breach notices (and If you do run into them, please report them as that's not the culture and I can see Pat personally firing anyone who tries that). You can not be "cheap" without being an asshole who's shaking down clients to pay for a new Yauht.
Most of the products are inexpensive IF you are using the functionality. The alternative to NSX is physically attaching a Layer 4 firewall like a ASA to every NIC port coming out of a ESXi Host, Buying Nexus 7K's with OTV licenses to do Layer 2 encapsulation over the WAN, and having to increase the size of your IDS/F5's/PaloAlto's by 20x. Does NSX SOUND expensive at first? Sure. The alternative though so laughable expensive that you historically had to leave gaps in your security strategy for east/west traffic or expect it to take weeks/months to move a VM between data centers vs. seconds...
-
Anyway, my point was:
-
Free software is great and powerful, you just trade this against time to understand how it works (or cross your fingers, but it's not acceptable in production). Note that you could mitigate the risk in different ways but you should understand most of your infrastructure.
-
Support/Service on proprietary software can be useful if you have money and you don't care about what's happening here (ie not your core business)
-
Support/Service on Open Source software is a kind of best of both worlds.
But that's my opinion
-
-
@John-Nicholson I'm not here to attack the product at all (I don't even know what half of the acronyms meant). I'm not building an hypervisor.
I'm just here to try to survive with the crumbs left from server virt market, without leaving my philosophy (making Free software).
-
@olivier said in Why is VMWare considered so often:
Anyway, my point was:
-
Free software is great and powerful, you just trade this against time to understand how it works (or cross your fingers, but it's not acceptable in production). Note that you could mitigate the risk in different ways but you should understand most of your infrastructure.
-
Support/Service on proprietary software can be useful if you have money and you don't care about what's happening here (ie not your core business)
-
Support/Service on Open Source software is a kind of best of both worlds.
But that's my opinion
One of the benefit of open source is that there are multiple parties contributing to it that you are not having to pay. (The challenge is their needs may not align with yours, although this happens with commercial software also). The SMB is perpetually in an awkward drafting of larger enterprises hoping for lower cost solutions that solve their problems to fall off the truck. Sometimes this works, sometimes it doesn't. When I worked for a SMB I'd always grown when I saw new versions come out that supported 10K VM's instead of 5000VMs and so forth...
The VMware VCSA uses Postgres, Photon Linux (you can find on GitHub). The major proprietary "Secret sauce" is the 150MB worth of proprietary VIB's on a host and the ESXi kernel itself. If you include GPL drivers, BusyBox, Linux, Tomcat/Apache, I think your typical vSphere deployment actually has more "Free" code on a per MB basis than not...
-
-
@olivier said in Why is VMWare considered so often:
@John-Nicholson I'm not here to attack the product at all (I don't even know what half of the acronyms meant). I'm not building an hypervisor.
I'm just here to try to survive with the crumbs left from server virt market, without leaving my philosophy (making Free software).
There's a lot of crumbs (maybe even spare slices of bread) in the hypervisor market (especially Xen) for easier to use management stacks and tool chains. Look at what Scale Computing is doing in going after the SMB market. There's a lot of people with 2-12VM's 1-2 IT guys, and they are not going to the cloud.
-
@John-Nicholson I never searched the VMWare contribution into the OSS, so I can be wrong but I never heard a lot people contributing to it (on major projects not those tailored only for VMWare itself)
