Fitting 40 logical cores and 400GB of RAM into one half-height blade allows me to achieve the density in a cloud environment I aim for. What's more is it handles workloads of 25 VMs and more without any noticeable penalty performance-wise. Despite having 20 physical cores running into two sockets on one board, the vCPU-ready times remain good throughout, while averaging 4 X vCPU for each physical core per blade.
Improvements to My Organization
The density achieved by the M630 while maintaining good performance allowed us to cut our hardware footprint by more than half. This makes for a huge cost saving, not only for hardware to maintain and rack space rental costs, but also reduces the required overall hypervisor licensing, which doesn't come cheap.
Room for Improvement
For a half-height blade, Dell did well with this one, not leaving much room for improvement. The ever-increasing boot time due to ever-increasing POST checks with every new generation makes maintenance tasks somewhat of a headache. That and the occasional chip creep causes one to have to re-seat DIMMs from time to time. Other than that, not much can be said in terms of cons.
If Dell can reduce the chip creep specially on the RAM modules that would e great. This does not occur that often but often enough to try and improve on. Each blade having each own iDRAC instance makes sense for connecting to a single blade on an ad-hoc base but becomes pain when having to access the local console for all blades in a chassis one after the other.
Each connection requires downloading the java aplet from the blade and jumping though a few hoops before it runs. When closing the aplet and one needs to connect to that same blade again the entire process needs to be repeated again. Performing this same task on an IBM blade chassis is a breeze. The CMC provides one with a drop down list listing all slots in the chassis. This allows one switch between local console mode for all blades in the chassis without any repetitive aplet downloads or even having to configure a management IP address for each blade.
Even though this is more of a blade chassis feature than a blade feature Dell would do well in stealing this page from IBM's book.
Use of Solution
I have been using it for almost one year.
These units have proven very stable. Even with component failure, in most cases, the unit keeps running, allowing one to migrate workloads off of it first, before attending to any break\fixes.
This solution has proven very scalable.
Customer Service and Technical Support
Technical support generally is acceptable. Having to jump through hoops can sometime be frustrating. Especially when your firmware has to be on the latest version before further assistance will be given. Even though I understand the logic behind it, keeping an entire environment on the latest version is nearly impossible at the rate firmware, drivers and other patches are released.
I have used IBM blade chassis before. The change to Dell at the time was more a financial one - one that I now fully support from a technical perspective.
Setting up the solution is fairly straightforward. If you have any experience on rack-mounts historically, it should all be familiar. Some networking and storage zoning is to be expected as well, though.
Pricing, Setup Cost and Licensing
The price tag on the M630 can easily be compared to that of most beefy rack mounts out there. And that with very little internal storage. If the aim is raw processing power in a dense and scalable solution with shared storage available, little out there will rival it. Keeping an eye on the exchange rate can make for some huge cost saving, as well, especially when looking to purchase a fully populated 16-blade chassis.
Other Solutions Considered
Having used predecessors like the M610 and M620 over the years, refreshing with M630s was the obvious choice.
When increasing the core count of the CPU when requesting a quote, be wary of the big jumps in price going from an 8-core CPU to 10-core or more etc. Dell doesn't give a price per component breakdown on their quotes. Getting two quotes for the same blade where only the CPUs differ will clearly indicate the huge price they associate per core.
Disclosure: I am a real user, and this review is based on my own experience and opinions.
Sep 01 2016