Cisco has a well-proven and well-tested layered Enterprise Data Center design approach that has been improved over the last several years. The Cisco Enterprise Data Center infrastructure is used to implement even extremely large data center environments. The most important features an Enterprise Data Center design must have are:
- Flexibility
- Maintainability
- Resilience
- Performance
- Scalability
The Enterprise Data Center architecture, as defined by Cisco, is a three-tier model that delivers scalability, performance, flexibility, resiliency, and maintainability. The three tiers include the following:
- Data Center Core Layer
- Data Center Aggregation Layer
- Data Center Access Layer
The Enterprise Data Center Core Layer offers high-speed packet switching with 10-GigabitEthernet connections or EtherChannel and backup links. The Core Layer is the switching backplane for all of the data flows going in and out of the Enterprise Data Center. Smaller data centers may use a collapsed Core Layer design that combines the Aggregation Layer and the Core Layer into a single entity.
The Aggregation Layer usually contains multilayer switches and it aggregates a variety of network services. Other functionalities of the Aggregation Layer are as follows:
- It acts as a service submodule integration layer.
- It acts as a Layer 2 domain definition point.
- It provides STP processing.
- It provides redundancy.
- It provides default gateways.
- It provides security functions (e.g., firewalls, IPS, and security policies).
- It provides network analysis.
- It provides SSL offloading.
- It provides content switching.
- It provides load balancing
The Enterprise Data Center Access Layer usually offers Layer 2 and Layer 3 Access Layer services, including:
- Layer 2 access with cluster services
- Blade chassis with integrated switches and pass-through models
- Mainframe services
- Other Layer 3 Access Layer services
Some of the benefits of the three-tier Enterprise Data Center model include the following:
- The Aggregation Layer prevents Layer 2 domain extension across the Core Layer.
- The service submodules offer an overall lower TCO.
- It reduces the complexity of the design because it reduces the number of components that have to be configured and managed.
- A combination of Layer 2 and Layer 3 access models can be used and some of the newer, faster modular platforms can be leveraged.
- It offers support for clustering services and NIC-teaming.
Cisco recommends implementing the Enterprise Data Center Core Layer infrastructure at Layer 3 and positioning the Layer 2 to Layer 3 boundary within or below the Aggregation Layer services. Layer 3 links offer the ability to achieve bandwidth scalability and quick convergence at the Enterprise Data Center Core Layer. It also avoids matters such as path blocking due to the STP algorithm or the downside of having uncontrollable broadcast issues that are typically seen in a Layer 2 domain.
OSPF and EIGRP are two of the routing protocols recommended for use in data center environments because of their capabilities to scale a large number of routers and achieve fast convergence times. The Layer 2 access design approach includes three categories:
- Layer 2 loop designs (triangle loop design and square loop design)
- Layer 2 loop-free designs (loop-free U and loop-free inverted U topologies)
- Layer 2 FlexLink design
The Layer 3 access design involves using Layer 3 (multilayer) switches at the Access Layer, as the Layer 2 to Layer 3 boundary moves down at this level in the Enterprise Data Center architecture. In a Layer 3 design, the Access Layer switches link to the Aggregation Layer switches with a Layer 3 uplink and a dedicated subnetwork. In this type of environment, Layer 3 routing takes place on the Access Layer switches first. The Access Layer switches are connected with Layer 2 trunks that provide support for the requirements of Layer 2 adjacencies in the data center. The most important consideration when choosing this design is that Layer 2 adjacencies are limited to the Access Layer switch ports. Therefore, VLAN extension across the Enterprise Data Center is not possible.
The overall decision of choosing a Layer 2 or a Layer 3 access design in the Enterprise Data Center should be based on a complex analysis of the topology. Each design type has specific benefits and drawbacks.
Some of the advantages of having a Layer 2 access design include the following:
- The NIC-teaming feature and Layer 2 adjacency across a wider area.
- It supports high availability clustering using Layer 2 adjacencies.
- The VLANs can be extended to support the server requirements.
- It provides better support for custom applications.
- FlexLinks can be used.
On the other hand, a Layer 3 access design offers the following benefits:
- Easier loop management.
- Faster convergence of the network compared to using the Layer 2 design and STP.
- It minimizes the broadcast domains, which leads to greater stability and easier management and troubleshooting.
- It provides better support for link utilization on all of the active uplinks.
The process of deciding on a specific Enterprise Data Center Access Layer design should take the following aspects into consideration:
- The organization’s needs
- Whether the staff has expertise in Layer 2 or Layer 3 technologies
- Oversubscription requirements
- Supporting and implementing service submodules
- Scalability issues
- Costs (for migrating from a Layer 2 to a Layer 3 Access Layer design)
Data center scalability issues include the following:
- The STP design is a huge issue because this is the feature that will help deal with a large number of VLANs in the organization and it will determine the ability to extend the VLANs across the Enterprise Data Center.
- Determining the need for VLAN extension in all areas of the network.
- Determining the number of necessary Access Layer switches and the total number of possible Access Layer switches supported by the Aggregation Layer.
- Determining whether to use RSTP or MST as the Spanning Tree Protocol mechanism.
- Preventing and solving scalability issues by adding Aggregation Layer services.
- The possibility of performing manual pruning on trunk links to reduce the total number of necessary active logical and virtual port instances.
- Determining the optimal ratio of VLAN to HSRP instances.
Redundancy should be implemented at Layer 2 in the server farm; the common NIC-teaming options include the following:
- Adaptive Fault Tolerance (AFT)
- Switch Fault Tolerance (SFT)
- Adaptive Load Balancing (ALB)
The main components of a Storage Area Networking infrastructure are as follows:
- Host Bus Adapters (HBAs), which are very fast adapters that connect to the disk subsystem.
- The communication subsystem, which evolved from IDE and ATA connectivity to Fibre Channel communications.
- The actual storage system, which consists of an array of hard drives that is considered “just a bunch of disks” (JBOD).
The most common levels of RAID are:
- RAID 0 (striping)
- RAID 1 (mirroring)
- RAID 3 (error detection)
- RAID 5 (error correction)
Some of the most important SAN design considerations involve port density and topology. Port density considerations include:
- The number of currently necessary ports
- The number of ports needed in the future
- The expected SAN lifetime
Topology design considerations include:
- Whether the data center is on a single floor
- Whether the data center is in a single building
- Whether SAN extension is required (in multi-building scenarios)
Some of the most popular designs include the following:
- Collapsed Core Layer architecture
- Dual-fabric collapsed Core Layer architecture
Cisco Nexus technology was developed to unify the LAN and SAN infrastructures. While most data centers use separate LANs and SANs with separate switches and network adapters in the servers, Cisco Nexus technology allows for the implementation of a unified fabric Enterprise Data Center architecture, with a consolidated LAN and SAN at the Access Layer.
Cisco’s line of products for server virtualization is called the Cisco Unified Computing System (CUCS), and one of the major advancements Cisco introduced in this area is a consistent I/O that provides uniform support for hypervisors across all servers in a resource pool. CUCS products facilitate advanced management and configuration of the throughput across all of the servers.