Open Networking with Whitebox Switches
As Cloud technology evolves (to implement most functions in software), what we need from the networking equipment is simplicity, reliability and high-performance. That is the reason the largest data centers in the world have embraced the whitebox model, based on merchant silicon.
Traditional networking is too complex!
Somehow, we are used to deploying and operating complex networks, with a different design each time, with huge deployment challenges, and with extremely expensive equipment and professional services.
Because of this complexity, traditional networking is not suitable for hyper-scalable data centers:
- Complicated protocols do help to grow the network is a simple way.
- Manual configuration is a limitation that blocks innovation, and relies on human intervention.
- VLANs do not scale, and cannot be reused across data centers.
- STP is unstable, sometimes comes with proprietary extensions that makes it incompatible, and keeps a lot of ports blocked.
- Troubleshooting is a pain, as certified personnel (in that specific brand) is required.
- We are captive to a single vendor, with no simple way to try other vendors.
- The network is (most of the time) a stopper for our business needs.
The Cloud changes the game. Most of the complexity is adopted by the hypervisor, making the plumbing much simpler.
Networking is getting simpler!
As cloud technologies evolve, most of the advanced networking is moving to the compute nodes.
The hypervisor is in charge of:
- Network Virtualization: creating private virtual networks across data centers, transparently for the cloud operator. By using the industry broadly adopted VXLAN protocol, hypervisor can create on-demand tunnels from VM to VM, without requiring fancy VTEPs (Virtual Tunnel Endpoints) in the switches.
- Security: Creating distributed security rules that apply on the VM interface level. Rather than investing in a physical security device (that does not see traffic between VMs), cloud operators increase security and reliability by adopt the FWaaS approach, which applies security features at the interface level in a more granular and more robust manner.
- Load Balancing: Per-Service load balancing (LBaaS) implemented by software, instead of a (likely to be point of failure) physical device.
Therefore, switching devices hava a single critical function:
To move traffic at the highest speeds, without drops and with low latency & jitter, with simplified troubleshooting and without single points of failures (SPOFs).
Whitebox Switching
Whitebox switching is a new networking model, where hyper-scalable data centers adopt commoditized networking, by using compact devices with identical form-factors, allowing to rely on more than a vendor.
Whitestack uses commoditized switching in its Cloud deployments, providing customers with investment protection by avoiding the traditional vendors’ lock-in.
Powered by Merchant Silicon
Whitebox switches leverage the most widely used networking chipsets (the same technology used by most of the traditional vendors).
Under the new Merchant Silicon paradigm, vendors have adopted the most powerful chipsets available in the market, from specialized vendors (such as Broadcom or Mellanox) rather than design their own ASIC (Application Specific Integrated Circuit).
This new industry trend reduces the hardware cost, provides higher performance boxes, reduces development time, and allows vendors to focus on software differentiation.
WS-SW-4810 (Leaf/ToR/EoR/Access Switch)
Our WS-SW-4810 is a Leaf/ToR/EoR switch with deep buffers for Data Center Networks as well as for Carrier network deployments, which leverages the Broadcom Qumran-MX chipset.
It is an ideal Leaf, Top-of-Rack, End-of-Row switch, but is also applicable for Metro aggregation. It provides 48 flexible SFP+ slots (1G / 10G) and 6 high-capacity QSFP28 slots (40 / 100G) uplinks.
- Architecture
- ASIC: Broadcom Qumran-MX
- Processor: Intel Rangley C2538
- Memory: 8GB DDR3
- Switching capacity: 800Gbps switching bandwidth
- Forwarding rate: 1607 Mpps
- Jumbo Frame support: Up to 9Kbyte
- MAC Address: Max 750K
- Ports
- 48x 10GbE SFP+ and 6x 100GbE QSFP28 ports
- 1 RJ-45 10/100/1000Mbps out-of-band management port
- 1 RJ-45 console Port
- 1 USB port
- Packet Buffer
- 4GB GDDR5
- 2 External T-CAM
WS-SW-32100 (Spine/Backbone Switch)
The WS-SW-32100 is a high-performance, high-density next-generation switch with target application for data center networks and high-capacity backbones, which leverages the Broadcom Tomahawk2 chipset.
It is an ideal Spine and Backbone switch, providing 32 high-speed QSFP28 slots (40 / 100G).
- Architecture
- ASIC: Broadcom Tomahawk
- Processor: Intel Rangley C2538
- Memory: 8GB DDR3
- Switching capacity: 3.2Tbps switching bandwidth
- Forwarding rate: 4762 Mpps
- Jumbo Frame: Up to 12Kbyte
- MAC Address: Max 136K
- Ports
- 32 100GbE QSFP28 ports
- 1 RJ-45 10/100/1000Mbps out-of-band management port
- 1 RJ-45 console Port
- 1 USB port
Our Open Networking solutions delivers the highest performance and reliability when used for implementing Spine & Leaf data center fabrics, following the most robust design patterns.