Dell's PowerEdge server line is once again trying to get out in front of Intel, announcing that its PowerEdge-C family of microservers are revved up with the new Ivy Bridge Xeon E3 processors, which the chip giant is launching soon. Dell might plunking the forthcoming E3-1200 v2 processors, which are baked using Intel's 22nm …
Dell keeps making these ultra-dense servers with four DIMM slots. What is the average Dell customer going to do with these servers? You are not going to run a VMware server with four DIMM slots. Memory is the bottleneck, not CPU. IBM has this right with eX5. Increase the DIMM outs to the moon. No one needs the CPU cycles.
I supposed if you were running some rendering application or another workload that is highly CPU intensive you could use these things, but how many of Dell's customers are running rendering clusters?
Well, DDR3 DIMMs now go up to 32GB per stick, although you don't want to know what they cost. 16GB sticks are much more reasonable, giving 64GB in four slots, which is quite adequate for a VMWare server running a handful of VMs.
However, VMWare is not the only game in town. For a high-traffic web site, for example, running a rack of these with Apache on, with a load balancer on the front end, would make a very useful solution.
I think you have fallen for the hype that everything should be run as a virtual machine ....... (always makes me laugh when vmware tell me I can get better performance as a vm than as a physical :-))
Hyperscale customers HATE hypervisors and "advanced" management, adds cost and complexity for no gain (we are talking 10,000+ physical boxes performing the same function).
You will find that most of these servers will be run with a stripped out linux core running apache or some other open source product acting as a huge scale out farm (usually web facing). In that case 4 DIMM's is more than adequate, throw in a bit of I/O (couple of 1Gb ports will do nicely thanks) and maybe a cheap 3.5" disk to run the OS on ....... and strip out everything else (no DRAC's, iLO's etc) to keep cost low.
For management you only want 2 things, ability to deploy an OS (simple PXE will do) and the ability to reboot the server. Troubleshooting these things goes as follows ........ if its broken turn it off, then turn it back on ....... if its still broken throw it in the bin put a new one in!
Oh and you are right that CPU's are too powerful. Most of these customers are investigating ARM or Atom based platforms (even 1 CPU and 4 DIMMs is too big!) as they are even smaller and more efficient (see the HP Moonshot announcement for what they are doing in this space and it should give you a good feel for what these people actually want and need)
- Batten down the hatches, Ubuntu 14.04 LTS due in TWO DAYS
- FOUR DAYS: That's how long it took to crack Galaxy S5 fingerscanner
- Did a date calculation bug just cost hard-up Co-op Bank £110m?
- Feast your PUNY eyes on highest resolution phone display EVER
- Wall St's DROOLING as Twitter GULPS DOWN analytics firm Gnip