Jupiter : The Centralised Control Network Of Google Datacenters

[Spartan]

 

 

Google has given everyone a rare look inside its server rooms and detailed how keeps up with the massive growth of its search business. In a blog post, Google Fellow Amin Vadat said that the company's current network, Jupiter, can deliver a petabit per second of total throughput. That means each of its 100,000 total servers can randomly speak to each other at a speed of 10Gb/s, a hundred times faster than the first-generation network it created in 2005. To get there, Google did something surprising -- it built its own hardware from off-the-shelf parts.

 

It was back in 2004 that Google decided to stray away from products by established companies like Cisco and build its own hardware using off-the-shelf chips from companies like Qualcomm. The aim was to put less onus on the hardware and more on software, something that's impossible with off-the-shelf switches. Vadat said hardware switching is "manual and error prone... and could not scale to meet our needs." Using software switching was not only cheaper but easier to implement remotely -- critical for a company whose bandwidth requirements have doubled (or more) every year.

 

Google considers its servers as a key advantage over rivals like Microsoft and Amazon, so why is it talking now? For one, it's recently started selling its cloud services to other businesses, so it's keen to brag about them. It's also being pragmatic -- its data requirements are now so huge that it needs academic help to solve configuration and management challenges.

 

 

That's why it's presenting the paper at the Sigcomm networking conference in London, and if you're in the mood for a (much) deeper dive, you can read it here.

[Spartan]

Our SIGCOMM paper on Jupiter is one of four Google-authored papers being presented at that conference. Taken together, these papers show the benefits of embedding research within production teams, reflecting both collaborations with PhD students carrying out extended research efforts with Google engineers during internships as well as key insights from deployed production systems:
 

• Our work on Bandwidth Enforcer shows how we can allocate wide area bandwidth among tens of thousands of individual applications based on centrally configured policy, substantially improving network utilization while simultaneously isolating services from one another.
• Condor addresses the challenges of designing data center network topologies. Network designers can specify constraints for data center networks; Condor efficiently generates candidate network designs that meet these constraints, and evaluates these candidates against a variety of target metrics.
• Congestion control in datacenter networks is challenging because of tiny buffers and very small round trip times. TIMELY shows how to manage datacenter bandwidth allocation while maintaining highly responsive and low latency network roundtrips in the data center.

 

 

These efforts reflect the latest in a long series of substantial Google contributions in networking. We are excited about being increasingly open about results of our research work: to solicit feedback on our approach, to influence future research and development directions so that we can benefit from community-wide efforts to improve networking, and to attract the next-generation of great networking thinkers and builders to Google. Our focus on Google Cloud Platform further increases the importance of being open about our infrastructure. Since the same network powering Google infrastructure for a decade is also the underpinnings of our Cloud Platform, all developers can leverage the network to build highly robust, manageable, and globally scalable services.

[andhrapradesh@123]

Gp-----> infinity

[Maximus]

[shivakumar577]

saw this on flipboard

[Luppon_Gidigidi]

[Novak_Djokovic]

[aragorn]

uncle edhi nee area ne gaa, paper chadiveva aithe

[compose]

[k2s]