Podcast: Play in new window | Download
Subscribe: Apple Podcasts | RSS
Welcome, Rick Altherr (@kc8apf)
- Rick works on the platform team at Google, designing firmware and software for the servers that go into hyperscale datacenters. He does not discuss specific Google stuff.
- Chris was surprised when Rick walked into a meetup carrying the Zaius server under his arm.
- The standard 1U rack size is 19″. This is the first of many standardizations that have happened over the years.
- You can purchase a slot, a rack, an entire cage or a building. Each comes with an increasing amount of logistical questions and potential for customization.
https://www.youtube.com/watch?v=_uvdPDXaBPc
- Engineers are following “Scale out” architecture these days.
- Kubernetes
- Gluster
- Reed-solomon codes
- MTBF of hardware
Your [server hardware] product is not truly shipped until it has caught fire at a datacenter
- These servers are part of the Open Compute project, which includes many large companies in the tech space.
- Two of the popular designs are the Leopard and the Zaius
- The Zaius has 2 Power9 processors, 48V rails, point of load converters that take the voltage down to 0.8V, and slots for 32 sticks of RAM. You can view all the design files on GitHub.
- PUE – power usage effectiveness
- Hot isle/cold isle
- Lights out management type system
- OpenBMC – Chassis manager
- NCSI – connect the ethernet
- IMPI – controlled by DMTF
- Project olympus motherboard microsoft
- You can just…BUY…these servers from the Open Compute project.
- Go on a 360 degree tour of the datacenters at Google
- Read more about Google’s datacenters here.
- Decisions are made on metrics like “People per megawatt” (how much management is required) and “Build vs buy” (is there a solution already in the marketplace).
- Finding a job: search for terms like “Datacenter”, “Platform”, or ping Rick via DM for mentoring and/or practice interviewing
- Rick gave a great talk at HDDG about monitoring engines called “Knock knock: Add more fuel“
- You can find Rick on Twitter or GitHub
Occasional barks during the show provided by Rick’s new puppy! 😀
https://twitter.com/kc8apf/status/901932715754217472
Looks like the puppy crashed 🙂
s/IMPI/IPMI
I’ve always wondered about hard drive space… are data centers using just off the shelf 8TB hard drives? And for a typical high-volume text website, say github or reddit, how much data space do they grow by each day?
Would also love to hear more about the acronym soup that is Infrastructure as a Service (IaaS) , Platform as a Service (PaaS) and Software as a Service (Saas).
Very cool episode. This is a topic that I can related to in a big way having worked for 9 years on HP Enterprise Storage, Fibre Channel, FCoE, iSCSI, Ethernet connectivity, Storage Target Drivers and NVMe SSD drives. I has been interesting to see the transition from the big Server and Storage companies to more modular rack based Open servers and storage as well as even more virtualization.
Thanks for sharing the great info.
So, where be Dave?