AWS now renting monster HPE servers, even in clusters of 7,680-vCPUs and 128TB Amazon Web Services usually stays schtum about the exact disposition of the servers it rents in its Elastic Compute Cloud, but made an exception on Tuesday with the announcement it is offering instances based on a single HPE server: the Compute Scale-up Server 3200. HPE bills the server, launched in 2023, as ideal for …
it is something only needed by companies which could easily run their own datacentres
If they've already migrated just about everything to the cloud, and that's the one piece they need to be able to decommission their datacenter it might make sense to them.
Of course Accenture or whoever will come along in five years and recommend bringing everything back off the cloud to save money, gotta keep that wheel of consulting fees spinning after all!
Well, whoever signs up for it can turn around and say they reduced their capital expenditure costs! Means they don't have to budget for depreciation any more, or pay the staff that maintained their physical estate...
Doesn't seem to matter to a lot of corps that it actually ends up being more expensive in the long term.
If your SAP instance has to interact with other stuff that is aleady (or that you want) at AWS, it may benefit you to have the HPE instance there. Also, these companies can strongly negotiate with amazon. Which in turn hetsa sweetheart deal from HPE. So, while not cheaper than on prem, it will be competitively priced.
So, while not cheaper than on prem, it will be competitively priced.
It would have to be, for now at least.. I'm no server expert, but..
Installing the HPE servers solves both problems – it's well and truly ready for SAP and is the same hardware HPE offers on-prem
I was expecting this beast to be at least a 9' rack with lots of blinking lights.. But it's kinda small, and says it'll scale up to 4 chassis, probably in a single rack. May need a sparky and plumber to keep it fed and cooled though. But if you're the kind of business that needs this kind of grunt.. why would you want to trust it to AWS instead of buying a chassis scaled to your business requirements and upgrade as needed?
I'm sure by 'competitively priced' will mean priced to look like it'll be cheaper than running it on-prem. For now. But then AWS will have you locked into a 3yr deal, and probably with an option to relentlessly jack up the prices, or charge for additional services? I have no idea how much these beasts cost, but if you need one, I'd suggest it's a good time to be looking very carefully about migrating stuff out of the 'cloud'.
I worked on the Superdome's predecessor once - the HP V2600. It was a box that sat on the floor that was basically a 1m x 1m x 1m cube. No namby pamby racks for it! They (there were two, of course) were connected to a Symmetrix array and a big Storagetek tape library that were likewise much too fat to fit in any sort of rack.
Ah the good old days when the real iron was sitting out on islands between the boring rows of pedestrian gear (yeah they had mainframes too but they were in a separate section of the datacenter and I didn't deal with them)
Ah the good old days when the real iron was sitting out on islands between the boring rows of pedestrian gear (yeah they had mainframes too but they were in a separate section of the datacenter and I didn't deal with them)
Yep. It's strange the way technology goes in circles. I started my career in the days of big iron, wrangling the newtorks of an Amdahl 5990-1400 (probably less compute than my phone has now), VAX8850 cluster and a genuine IBM AT with both VYT and 3270 emulation. Then ended up at an ISP and attempting to argue that an IBM AS/400 was a better webserver solution than racks full of Sun boxen. And then leaning nonchalantly on a 'cloud' while amusing myself at weekends freefalling through real ones.
It always puzzled me why clients would convince themselves that renting space on someone else's HPE servers was better/safer than using their own. Especially when the rental model would often replace client's HPE tin with a 'solution' that charged them back per Mbit, byte, flop and anything else that marketing could figure out a way to meter & rate. But I guess when regulators started cracking down on dodgy timeshare rentals, those sales types had to end up somewhere.. And pretend computing hasn't gone from mainframe-> peer-peer -> thin client and back to mainframe again.
I'm just glad I became independent and can try and convince clients that 'cloud' isn't always the best solution, and the 'cloud' just hides multiple sins that their shareholders & customers might not thank them for when the 'cloud' has a bad day. Again.
these companies can strongly negotiate with amazon. Which in turn hetsa sweetheart deal from HPE. So, while not cheaper than on prem, it will be competitively priced.
That's not necessarily true. This service is aimed at people who:
(1) are hostage to legacy applications that can't be migrated to cloud in the regular way; and also
(2) are hostage to bosses who say "you must put this in the cloud!"
It will be priced at whatever the market will bear. The fact that AWS themselves are buying the tin on the cheap doesn't necessarily trickle down into the end-user pricing.
You've investing $$$$$$$ on a massive server, SAP, in the cloud, putting your entire business at the mercy of AWS who have 1000s of customers larger than you, so when something goes wrong, you're on the phone to AWS support, not onto your own IT manager who you know will be doing everything in his power to get the system or network back up and running.
This is where Cloud goes wrong, there are some large systems that a business is so reliant on, that they should stay in their full control - either on premise or in a data centre local to the business with support teams they can call and get immediate action.
The rental is also going to be more expensive than traditional business finance.
"putting your entire business at the mercy of AWS who have 1000s of customers larger than you, so when something goes wrong, you're on the phone to AWS support, not onto your own IT manager who you know will be doing everything in his power to get the system or network back up and running.”
And this is an argument I have had a number of times wit clients who simply want to ‘move to the cloud as it’s the future’. OK fine, if and when something breaks what do you do? Well I’ve listened to answers along the lines of ‘it can’t / won’t ever break because it’s ‘the cloud’; to ‘well we call Microsoft (and/or) Amazon and they’ll fix it!
Oh for Christ’s sake!
Now oddly enough when I sit them down and tell them ‘well your company is important to you, but to Microsoft / Amazon (insert other cloud services here), you and your company are nothing - they simply don’t care, gone out of business, sorry about that? Tough. You did read the small print before signing up; no?
Sounds like AWS is now competing with SAP's own Private Bare Metal Cloud for HANA...at least in "size" if not in bare metal.
In the past AWS-hosted HANA was used by SAP's "small customers" as AWS couldn't offer large enough machines.
I used to work on the infrastructure for SAP's Private Bare Metal Cloud, extremely impressive hardware (machine counts, CPU counts, RAM sizes, huge SSDs) was used for customer deployments.
S/4HANA does not run on a cloud-native microservices architecture. As such it provides no clear cloud benefits, yet it typically costs more in the public cloud and runs slower in the public cloud. Running this type of workload in a public cloud adjacent, colocation private cloud IaaS model would be a less expensive and better performing option. Many CXOs are not aware of this but their G&A budgets are bursting as they go all-in on public cloud for non cloud-native workloads.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
