A Scalable, Commodity Data Center Network Architecture
Download as PPT, PDF5 likes2,832 views
The document provides an overview of scalable data center network architectures, focusing on the fat tree topology and the Monsoon solution, which utilizes a layer 2 flat routing design to improve performance and reduce costs. It discusses the benefits of the fat tree approach, such as supporting high throughput while maintaining packet ordering and addressing issues related to traditional topologies. Additionally, the Monsoon strategy emphasizes the elimination of spanning trees and the use of cost-effective components to facilitate efficient host communication within the data center.
A Scalable, Commodity Data Center Network Architecture
- 1. A Scalable, Commodity Data Center Network Architecture
- 2. Overview • Structure and Properties of a Data Center • Desired properties in a DC Architecture • Fat tree based solution • Monsoon: layer 2 flat routing
- 3. Common data center topology Internet Data Center Core Layer-3 router Aggregation Layer-2/3 switch Access Layer-2 switch Servers
- 4. Problem With common DC topology • Single point of failure • Over subscript of links higher up in the topology – Trade off between cost and provisioning
- 5. Properties of solutions • Backwards compatible with existing infrastructure – No changes in application – Support of layer 2 (Ethernet) • Cost effective – Low power consumption & heat emission – Cheap infrastructure • Allows host communication at line speed
- 6. Cost of maintaining switches
- 7. Need for Layer 2 In DC • Certain monitoring apps require server with same role to be on the same vlan • Using same ip on dual homed servers • Allowing growth of server farms.
- 8. Review of Layer 2 & Layer 3 • Layer 2 – One spanning tree for entire network • Prevents looping • Ignores alternate paths • Layer 3 – Shortest path routing between source and destination – Best-effort delivery
- 9. FAT Tree based Solution • Connect end-host together using a fat tree topology – Infrastructure consist of cheap devices • Each port supports same speed as endhost – All devices can transmit at line speed if packets are distributed along existing paths – A k-port fat tree can support k3/4 hosts
- 11. Problems with a vanilla Fat-tree • Layer 3 will only use one of the existing equal cost paths • Packet re-ordering occurs if layer 3 blindly takes advantage of path diversity
- 12. FAT-tree Modified • Enforce special addressing scheme in DC – Allows host attached to same switch to route only through switch – Allows inter-pod traffic to stay within pod – unused.PodNumber.switchnumber.Endhost • Use two level look-ups to distribute traffic and maintain packet ordering.
- 13. 2 Level look-ups • First level is prefix lookup – Used to route down the topology to endhost • Second level is a suffix lookup – Used to route up towards core – Diffuses and spreads out traffic – Maintains packet ordering by using the same ports for the same endhost
- 14. Diffusion Optimizations • Flow classification – Eliminates local congestion – Assign to traffic to ports on a per-flow basis instead of a per-host basis • Flow scheduling – Eliminates global congestion – Prevent long lived flows from sharing the same links – Assign long lived flows to different links
- 16. Results: Heat & Power Consumption
- 17. Draw Backs • No inherent support for VLan traffic • Data center is fixed in size • Ignored connectivity to the internet • Waste of address space – Requires NAT at border
- 18. Monsoon approach • Layer 2 based using future commodity switches • Hierarchy has 2: – access switches (top of rack) – load balancing switches • Eliminate spanning tree – Flat routing – Allows network to take advantage of path diversity • Prevent MAC address learning – 4D architecture to distribute data plane information – TOR: Only need to learn address for the intermediate switches – Core: learn for TOR switches • Support efficient grouping of hosts (VLAN replacement)
- 19. Moonson
- 20. Monsoon Components • Top-of-Rack switch: – Aggregate traffic from 20 end host in a rack – Performs ip to mac translation • Intermediate Switch – Disperses traffic – Balances traffic among switches – Used for valiant load balancing • Decision Element – Places routes in switches – Maintain a directory services of IP to MAC • Endhost – Performs ip to mac lookup
- 21. How routing works • End-host checks flow cache for MAC of flow – If not found ask monsoon agent to resolve – Agent returns list of MACs for server and MACs for intermediate routers • Send traffic to Top of Router – Traffic is triple encapsulated • Traffic is sent to intermediate destination • Traffic is sent to Top of rack switch of destination
- 23. Forwarding
- 24. Other Work in the Data Center Space • Network Security – Policy aware switching • Data Center Cabling – 60GHz Data-Center Networking: Wireless
Editor's Notes
- #4: Before describing the problems with current middlebox deployment approaches, let me first describe the commonly used 3-tier data center network topology. At the top is the core-tier, whose layer-3 routers connect the data center to the Internet or to the rest of the campus network. At the bottom is the access tier, containing the layer-2 switches into which servers are plugged in. In between the access and core tiers are the layer 2/3 switches of the aggregation tier. Middleboxes are commonly deployed at the aggregation tier. Multiple redundant links connect together the various switches and servers. To prevent forwarding loops, we use mechanisms like spanning tree construction to block out some of the links. For example, the topology as shown here.
- #7: Netgear ~ 3K Procurve – 4.5K
- #9: In this talk, I shall next explain the problems with current middlebox deployment mechanisms. I shall then describe how our solution, the policy-aware switching layer simplifies middlebox deployment and achieves the properties I mentioned earlier. I shall briefly discuss related work, and our prototype implementation and evaluation of the policy-aware switching layer.