This is the second of a two-part blog series developed in association with Tom Edsall, a Cisco Fellow and CTO of Insieme Networks, recently acquired by Cisco Systems. While Part 1focused on the role of SDN overlays and deployment considerations, this one delves into the benefits derived from such a model in an ACI implementation.
First, an insight into these design principles and the focus on applications from some of the top Insieme engineers-
Benefits of Overlay Integration in ACI deployments
If you have a small amount of state to update when an application moves or is added to the data center you will scale better than if you have a lot of state to update. With the ACI implementation, we are pretty fortunate in that the amount of state required by the mapping database is relatively small. It is a simple binding of identity to location. There is other state not related to the overlay such as access policy which may be larger, but that will be discussed at another time as it is not strictly required for the overlay. The amount of state that changes for a single move does not change with the size of the data center.
At the heart of the transition to cloud computing is on-demand provisioning of a wide variety of applications, linear scalability of resources, and non-stop operation at lower total cost. With the increasing frequency of rapid provisioning of data-intensive applications in the cloud, organizations are increasingly challenged to better scale and manage network and storage environments without business disruption. This necessitates a network that provides uniform latency, high bandwidth, full utilization of all paths, and configuration simplicity.
The Cisco Nexus® 9508 40GbE data center Ethernet switch was recently tested by Lippis report and turned in remarkable performance results, while supporting 288 40GbE ports for the highest 40GbE port density of any switch tested to date. Cisco Nexus 9508 performed with the best overall store-and-forward latency of core switches tested to date, while providing consistent latency across all packet sizes at line rate. In addition, it demonstrated 100% throughput (i.e. without dropping a single packet!) across all 40GbE ports for a wide range of packet sizes. This is key to public and private cloud providers seeking aggregation and core networking technology that underpins large-scale, highly virtualized data centers and converged storage systems with support for disparate workloads having a wide range of performance requirements.
The industry-leading 40GbE density and performance of Cisco Nexus 9508 enables data center IT to upgrade aggregation network infrastructures from 10GbE to 40GbE to complement the shift in server networking from GbE to 10GbE. Having extremely impressive cross sectional bandwidth and latency numbers, the Cisco Nexus 9508 can also excel for aggregation and core infrastructure applications in traditional, cloud data centers as well as hyper-scale data center environments. Cisco Nexus 9508 is also optimal network infrastructure for high performance cluster computing applications, for example, for large-scale data analytics and low-latency trading applications.
For unicast traffic, Cisco Nexus 9508 delivered store-and-forward latencies ranging from 1.6 microseconds for 64B packets used in transaction workloads to 3.5 microseconds for 9KB packets used in data-intensive, large-file applications. The latency variation ranged between 1 to 3 ns, allowing consistent latency across all packet sizes at line rate. These are by far the lowest latency measurements observed by Lippis Report in core switches to date (the previous record for modular switch latency was 2.2 to 11.9 microseconds, at the same packet range, however at much less density).
For IP multicast traffic, the Cisco Nexus 9508 demonstrated store-and-forward latencies ranging from 1.6 microseconds for 64B packets to 3.5 microseconds (3465.3 ns) for 9KB packet forwarding IP multicast traffic faster than any other core switch observed in Lippis core switch tests.
The Cisco Nexus 9508’s congestion management is excellent at nearly 78% of aggregated forwarding rate as percentage of line rate during congestion conditions for L3 traffic flows, but when considering the density of ports supported and sheer magnitude of the traffic flow, the Cisco Nexus 9508 achieved congestion management at a scale never before attempted.
The Cisco Nexus 9508 also demonstrated 100% throughput as a percentage of line rate across all 288 40GbE ports for unicast traffic. In other words, not a single packet was dropped while the Cisco Nexus® 9508 was presented with enough traffic to populate its highly dense 288 40GbE ports at line rate.
The full report can be found here:
Following are links to webcasts providing highlights of unicast and multicast support of Cisco 9000:
Nexus 9000 Unicast forwarding by Lilian Quan
In less than 48 hours, on Nov 6th, Cisco will officially announce details of its approach to Application Centric Infrastructure based on innovations from Insieme networks.
[To join our Webcast with John Chambers and Insieme Soni Jiandani on Wednesday November 6 at 10:30 am EST/7:30 am PST register here ]
No doubt that there are great expectations riding on this announcement. In part due to Insieme’s pedigree, and in part due to the promise it brings. The former I covered in my previous blog – Application Centric Infrastructure gets ready to rumble. As for the latter, it is important to put the promise in perspective.
As businesses of all sizes continue to pump billions of dollars of investment into data center and cloud IT and consumption models, the expectations for IT to generate a credible RoI in terms of business agility, productivity and efficiency has never been higher. IT today epitomizes a very powerful business function that directly impacts agility and contributes to bottomline and customer experience. It is but natural to think of data centers are futuristic, glamorous environments, quietly humming away transforming businesses at their core. Many new data centers are in fact that, and seem to come out for sci-fi movie.
However, in a majority of cases, many data centers have just “grown-up” too fast over the last decade that saw adhoc spurts in data center consolidation and server virtualization. While both phases have provided tremendous benefits in terms of efficiencies and economies of scale, they have also contributed their mite to increasing operational complexity. From cabling sprawl, to network and server sprawl, to VM sprawl – the rapid growth has in some cases negated cost efficiencies gained through server virtualization. The same can be said of software stacks with complex licensing and version control issues. The affinity to applications in this chaos has somehow been either taken for granted or partially lost in the complexity, making it harder to bring predictability into application deployments or troubleshooting, leading to both time and cost overruns. How can this be simplified? How can infrastructure be better linked to the needs of applications? How to make life simpler for data center operations and facilitate a better application experience?
Per a survey conducted by ESG last year, 63% of IT pros say new app deployments take a month or more, 50% say upgrades take just as long. 77% of enterprise IT pros say they would manage more than 150 applications over the next year. This is a lot! For IT to deliver high performing apps, they need a way for infrastructure to automatically respond to the needs of the application, and to have excellent visibility when something goes wrong and requires troubleshooting during application deployments or upgrades.
True Innovators are not easy to find. In fact, people who build next-generation innovations and succeed are certifiably rare. Repeat successes with these innovations are rarer still. Hatricks are legendary. Anything beyond is best relegated to a rarefied stratospheric atmosphere and dismissed as fiction.
Fortunately for us at Cisco, one need not look too far for such innovators. The team of Mario Mazzola, Prem Jain, Luca Cafiero and Soni Jiandani (affectionately called by some in the industry as the MPLS team) has both the reputation as well as the track-record of not just building world-class innovations, but for successfully converting them into multi-billion dollar global businesses within the Cisco fold. So, when they talk about networking, a subject which they know a thing or two about, most people find it worth their while to listen up.