Facebook has made it official that Intel is all set to provide low consuming processors in coming days. The social media giant has also disclosed that instead of using servers with more CPU sockets, Intel’s new processors will witness a reduction in number of them to meet the 11KW limit per rack budget set by it in 2011.
The details are as follows- When we need more compute power, generally we either add more CPUs or go with more powerful processors. Intel till date has come up with more and more powerful CPUs each time around. Although, this is good, the tradeoff was that each generation of CPU was coming with higher power consumption ability, up to 120W.
Facebook wants its computing supplier to come up with low power consuming chips, in order to meet green computing standards. The world’s number one social media giant began to work with Intel a few years ago to develop a single socket CPU, Broadwell-D, which was minimized to exactly fit specific needs.
While Intel focused on the CPU, Facebook redesigned its server infrastructure. Facebook designed a 4 CPU/sled server that would accommodate the new 65W CPUs. That means in Facebook’s 30 server rack they could go from 60 120W CPUs to 120 65W CPUs that hit their needs specifically—Twice the CPUs per rack at similar power consumption.
Intel’s new line of products is called the Xeon-D line. Facebook has designed the Mono Lake server, which it calls the server embodiment of Xeon-D and the building block for its SOC based server designs. It is highly simplified board that allows for the SOC to be placed with its associated memory and storage for boot and log, currently 32GB RAM and 128GB of storage. In future, as the workloads increase in Facebook’s server farms, more RAM and storage (for boot & log) will be added.
Facebook is using Yosemite to incorporate Mono Lake Servers into its racks. Yosemite has Sled and associated NIC support and so connects the four Mono Lake Servers to the NIC through PCle lanes, to give an independent IP address to each server. This design requires the use of a multi-host NIC to aggregate the bandwidth from the four servers. The output of the NIC is currently designed to be 2×25 Gbps, which is dynamically shared across the four CPUs. Aggregating ports at the sled level makes for more efficient usage of the top-of-rack (TOR) ports.
Using low powered CPUs seems counterintuitive to get more performance, but instead of using high powered CPUs for all applications, differentiating the use as per needs makes more sense in Facebook’s perspective. It also gives Facebook a better overall performance with low power consumption.
Hope, Facebook doesn’t face any challenges with such usage.