Recently Arista released a white paper surrounding the idea that having
deeper buffers running within the network can help to alleviate the incast
congestion patterns that can present when a large number of many-to-one
connections are happening within a network. Also known as the TCP incast
problem. They pointedly targeted Hadoop clusters, as the incast problem can
rear its ugly head when utilizing the Hadoop Cluster for MapReduce
functions. The study used an example of 20 servers hanging off of a single
ToR switch that has 40Gbps of uplink capacity within a Leaf/Spine network,
presenting a 5:1 oversubscription ratio. This type of oversubscription was
just seen in the recent release of the Facebook network that is used within
their data centers. So its safe to assume that these types of
oversubscription ratios are seen in the wild. I know I’ve run my fair share
The internet has been buzzing about Facebook’s redesigned datacenter
architecture. Facebook, which is used by more than 1.35 billion people,
recently restructured their infrastructure to increase flexibility and
agility to rapidly adjust to application requirements. Our own Marten
Terpstra shared his take on the redesigned infrastructure this week on the
Plexxi blog—it’s definitely worth a read. Below we share some of the
articles that covered Facebook’s new datacenter architecture, as well as
other happenings in the networking space this week.
In this week’s PlexxiTube of the w... (more)
Depending on who you ask, the primary objective of SDN is to improve
management and operations. By centralizing control, things like provisioning
and troubleshooting can be streamlined, reducing the time and effort it takes
to perform tasks. From that central point of control, a number of interfaces
exist that make automation more effective, or at least more accessible.
But there is more to operational efficiency than the commands and keystrokes
required to get something done.
Friction in any ecosystem is greatest at the boundaries between elements.
When two things must work toge... (more)
I always enjoy reading the IPspace blog and as Ivan has stated about our
blog, I don’t always agree with his opinion, but they are informative and
cover just about everything networking. So this may come as a surprise, but
in response to his “Do we have too many knobs” post from about a week ago
I have one simple response: “Amen”.
Networking is unnecessarily complicated. We have written several blogs on
this topic and related items. I used to run the sustaining organization for
all data products at my previous company and when you do the analysis of the
customer reported issues ... (more)
A few weeks ago Facebook announced their new datacenter architecture in a
post on their network engineering blog. Facebook is one of the few large web
scale companies that is fairly open about their network architecture and
designs and it gives many others the opportunity to see how a network can be
scaled, even though the scale is well beyond what most will need in the
foreseeable future, if not forever.
In the post, Alexey walks through some of the thought process behind the
architecture, which is ultimately the most important part of any architecture
and design. Too often we ... (more)