Defiant Broadcom calls for tech to go back where it belongs: On-premises Broadcom has opened its VMware Explore conference in a defiant tone, declaring it now offers a superior user experience compared to public clouds. CEO Hock Tan took the keynote stage in Las Vegas and asserted that the VMware Cloud Foundation (VCF) private cloud suite offers better security and cost management than hyperscale …
Broadcom - like Oracle - has no say in my systems or how they're deployed.
Precisely because of their previous behaviour.
And everything goes in cycles.
Consolidated, distributed, thin client, fat client, external service, on-prem, it's all going to go on the same 10-20 year cycle as everything else, as people hop from one to the other because the thing that's "bad" at doing X will make them move onto the thing that's "bad" at doing Y, and then vice versa, rather than accept X or Y or find a middle ground.
Honestly, I lose track of how many times some things have been through that cycle now.
The biggest original selling point for Cloud was that you only paid for the servers when you needed them. Obviously, the providers added costs by packaging the tools to deploy, which took the edge off any savings (or completely removed any cost advantage!)
In some respects, unless you run diverse workloads with different time based load requirements, on-prem cloud looks to me like "You pay for the servers when you buy them, then you pay for the cloud-like tools to manage the workloads and servers".
This doesn't sound to be much benefit to me, not unless you're wedded to the tools to manage and deploy a cloud-based environment.
I wonder how long it will be before we cycle back to provisioning for the peak workload like we used to, and either shut down unused infrastructure except for when it is required, or just ignore the fact that it's spinning for no benefit (intelligent monitoring of the use of cores may assist in this). I'm sure someone must be doing the cost-based analysis.
I am not in [your] field but wouldn't hybrid cloud mostly solve this equation? On-prem for (time proven) base load, spinning up cloud VM's for "spiky" needs. This lowers cloud costs, you are only paying for high demand, and moves most ROI analysis, security and availability to on-prem.
I am not sure why your think that is not being done already. In both sides, on the internal cloud side and the the VM clustering side both in prem exist solutions to dynamically manage servers and resources currently. The debate is more about what integrated managed services you get from external provider versus you managing it internally. Those include general maintenance and software provisioning, availability accross world regions, etc. The paradox is that every cloud provider service usage ways are diverting as time goes by and cloud vendor lock in is growing every time that pass and those vendor costs are growing in some cases uncomfortable way. That is why most of cloud users are trying to use their party to deal with public cloud companies that deal with multiple cloud providers even when they do not deal with direct server provisioning but use microservices paradigm. In general not all should be going to public cloud unless you are a huge company with billions of users like big tech and specially when your users are geo located where the company can have its own internal cloud. Or VM cluster.
Dell already offer a solution to that conundrum with APEX, you get a bunch of hosts and only PAYG for the ones that go over 5% CPU utilization. Allowing you to have a sizeable standby capacity to meet increase of demand without going through procurement. More importantly it's all structured as OpEx, which is music to the beancounter's ears.
Sadly you still need to pay for the VCF licenses regardless.
But if you have big enough ELA with Microsoft and go down the Azure Local route with Dell APEX (or similar PAYG offering from other vendor) you kinda get the best of both worlds
It's interesting that the CPU usage is so low. 5% is almost nothing, and could well be consumed in the basic OS and resource costs of the tools to monitor the usage.
We've seen similar capacity on demand features before. IBM still has the facilities to provide capacity-on-demand for Power systems, where you pay for part of the delivered system, with capacity turned off, but which can be dynamically enabled as needed, and for a cost, when it is required.
I don't know many companies who actually used this feature, but I think it's still there.
Yes - with cloud you only pay for servers when you need them. But in terms of comparing public cloud vs-onprem costs, that's really only a factor for very small environments or places with very dynamic sizing requirements.
If your business basically consists of two 24x7 VMs, and you sometimes to fire up a third for testing/reporting/whatever - then cloud will save you money. But if your business consists of 200 24x7 VMs, and you sometimes need to fire up number 201... on-prem is going to be hugely cheaper, even with some spare (aka wasted) capacity built in.
On the other hand, if your business consists of 200 VMs in normal operation, but there is some crazy peak thing that happens (christmas shopping, tax return season, whatever) and you sometimes need to scale dynamically up to 2,000 VMs... obviously the cloud is a great choice. But despite this being one of the main cloud marketing points, most businesses don't have that kind of cycle.
Ultimately if you have a non-trivial compute need, and your need for dynamic scaling is small compared to your always-on load, on-prem will always win. And the bigger your environment the more true that becomes.
I was not promoting cloud over on-prem. I think it has it's place, as you point out, but for the most part, cost is not it's main advantage.
I am mostly in the on-prem camp. I feel that for many companies, this provides the control, cost certainty, and data sovereignty that the IT departments want. But it does mean that the company carries significant capital assets in the systems themselves and the environment that they run in, and the headcount to run it.
This does not often match what the bean-counters want. Oh yes, they say that they want everything, they want cost control, they want data security, and they want to be able to manage the systems, they just don't believe what their own IT departments say when compared to the sales people and consultants of the cloud companies, who can often be likened to snake-oil salesmen.
It seems to me that this just plays into the capital vs. operational costs debate. Cloud normally fall onto the operational costs side, whereas on-prem falls onto the capital side (I'm ignoring leasing capital assets, the previous way of shifting costs to operational expenditure for this argument). So, even if you can point out that in the long-term, on-prem is less expensive, less risky and provides higher controllability, the shareholders, the board, and the beancounters want these costs on the op-ex side, and are really just interested in the bottom line on the annual financial statement. This is a huge hurdle to overcome.
Crudely, the cycle is as follows:
What you are doing NOW is WRONG ... change to XXX NOW!!! it is wonderful, cheaper, faster and makes Flapjacks as well !!!
Wait 18 Months until the boom has slowed ...
What you are doing NOW is WRONG ... change to YYY NOW!!! it is wonderful, cheaper, faster and makes toasted Bagels as well !!!
Wait 18 months until the boom has slowed ...
Repeat ad nauseam changing the extra 'Breakfast snack' as needed !!!
Broadcom is probably sensing that there is push back against the cloud movement. And they are trying to remind customers that they still offer products that go beyond their portfolio of networking and virtualization offerings.
Smart thinking actually. But let's see if thy can find a niche for buggy whips in an EV market.
Its well known that BC wanted to let go their smaller customers in favour of their global 1k.
VCF9 with on prem private cloud capabilities would be appealing to the smaller operators outside of that global 1k.
Biggest issue with the cloud are teh vampire costs, where some dev has spun up a load of stuff that is still running long after they've gone and much of it is possibly not needed but no one knows so teh company is charged for things they may not be needing & the cloud provider is the benefactor.
Same issue can & will happen on prem but the cost is effectively just some electric & no cloud provider profits.
You're right. They are concentrating on the 1000 biggest customers. Massively profitable for 3 to 5 years. Unsustainable in the long run .... but that is long after the current batch of C-level have cashed out and gone.
And overnight, they have created a market to service the other 99.9% of customers that aren't in that small niche.
This post has been deleted by its author
When they first bought VMWare I tried to buy a new license for a 3 host backed onto a SAN setup. I couldn't get a license from Dell or Dicker Data or through Broadcom's website.
How does one actually purchase VCF licenses if they aren't a 3000+ CPU site? Or has that changed?
VMware isn't even relevant anymore, who is even considering using VMware either in cloud or private cloud ? You'd need t be paid to do that, a lot.
And Broadcom, what was it people said.... how can you tell when Trump is lying ? His mouth is open.... same for Broadcom.
It's over.
Goodbye VMware.
