Direct from Development – PowerEdge MX-Series Optimizations for the Software Defined Data Center
Tue, 10 Nov 2020 23:17:51 -0000|
Read Time: 0 minutes
The Software Defined Data Center is emerging as one of the leading architectural approaches for customers who wish to cut costs, increase agility and improve reliability without incurring vendor lock-in. The new MX-Series Modular solutions from Dell EMC were designed specifically to enable SDDC by integrating key optimizations for Software Defined Storage and Software Defined Networking with the industry leading Dell EMC PowerEdge Server family.
The term Software-Defined Data Center (SDDC) defines the extension of virtualization to all data center resources. According to Wikipedia, an SDDC virtualizes “all elements of the infrastructure – networking, storage, CPU and security and delivers them as a service.” The benefits include reduced acquisition cost, reduced operating cost, increased levels of automation with each component potentially provisioned, operated and managed through an application programming interface (API). Most customers have started the journey towards SDDC with the implementation of server based hypervisors like VMWare, Hyper-V and KVM for the compute layer and many have extended this concept to storage with Software Defined Storage (SDS) solutions. For these customers, the next stage in the journey will often be the virtualization of Network Services (SDN). The Dell EMC MX-Series has been designed specifically to assist customers with this journey.
MX Optimizations for SDS
Software Defined Storage (SDS) solutions aggregate disk storage local to each server into a highly reliable, extremely scalable, high performance storage pool that is easy to deploy and manage. This approach costs less, performs better and has helped many customers reduce the time it takes to deploy new solutions but has historically been a poor fit for blade environments due to their low disk counts.
The complexity of managing large numbers of servers led to the development of blade systems where multiple servers could be enclosed and managed from a single chassis. The challenge for many customers as they evolve to SDS based storage is that these systems were designed for the SAN based storage technologies available and simply do not offer the local disk capacities necessary for SDS.
With drive subsystems optimized primarily for boot functions, most blade designs offer only 2 drives. In designing the MX-Series, Dell EMC Engineers took the opportunity to rethink the entire architecture and to design a solution that not only exceeds the management efficiencies of existing blade solutions but adds in key design elements that make it an ideal solution for SDS environments.
MX SDS Capacity Enhancements
A key element of SDS optimization is capacity. Unlike other 2 socket blade solutions on the market, the Dell EMC PowerEdge MX740c offers up to 8 drives including 6 front-mount hot pluggable 2.5” drives and 2 internal SSD’s installed on the optional Boot Optimized Storage System (BOSS) controller.
For customers desiring the cost efficiency of traditional disks, this allows for up to 5 x 2.4TB SAS drives for a total of 12TB of raw capacity, 1 SSD for caching and then offers the optional BOSS controller to provide space for the operating system and log files.
For customers desiring maximum capacity and performance, the system can be configured for “All Flash” operation with raw capacities up to 23TB using 6 x 3.84TB SSD’s. The optional BOSS controller can again be utilized maintain the operating system and log files allowing all hot pluggable drives to be utilized for SDS.
For customers with even more demanding storage requirements, the optional MX5016s storage system can be added. With support for up to 16 additional hot plug SAS drives, this device can be used add drive slots to 1 server or to divide the drive slots between other servers in the enclosure to increase the capacity of an SDS solution.
Other considerations for optimized SDS
SDS solutions are designed to accommodate network bottlenecks however, it is logical that reducing latency and increasing bandwidth can also increase the efficiency of the storage pool by reducing the replication time required to protect the pool. “All Flash” environments in particular can deliver significantly higher Input/Outputs per second (IOP’s) than traditional disks and the subsequent increase in disk activity can more quickly be processed in an environment where network latency is reduced and/or network bandwidth is increased.
Dell EMC Engineers addressed both of these elements in the MX-Series. First, all compute devices have standardized on 25Gb/s Ethernet which more than doubles the throughput available with existing 10Gb/s technologies. Second, the MX-Series offers network switching options specifically designed to accommodate the full bandwidth of the solution with no oversubscription. With industry-leading latency rates of sub-600ns for the MX5108n and Sub-500ns for the MX9116n, the MX-Series also processes network transactions faster.
Note: an SDS solution can survive the failure of multiple disks but not the failure of all the disks. For this reason, Dell EMC Engineers designed the MX5016s to support a maximum of 16 drives and to allow the addition of multiple MX5016’s in a single enclosure to avoid the risk of creating a single point of failure for the entire SDS pool. This design dramatically reduces the impact of a storage sled failure. Each MX5016s has redundant, hot plug expanders and is connected to redundant SAS switches in the enclosure.
MX Optimizations for SDN
All MX switch options come enabled for SDN. Delivering on the promise of “Open Networking,” both the MX5108n and the MX9116n ship pre-configured with Dell EMC Networking OS10 but include ONIE, enabling the option for third party SDN solutions from Dell EMC Networking partners.
The Dell EMC MX-Series was designed specifically to help customers facilitate their move to a Software Defined Data Center. From compute, to storage through to networking, the solution has been designed to work seamlessly and deliver uncompromised performance, reliability and efficiency. Tying all of these elements together is the new Open Manage Enterprise management framework which is built on the same API’s that Dell EMC provides to customers for custom development. This open approach allows customers to integrate the MX-Series seamlessly within their own management tools and workflows or, use the interface provided by Dell. Most importantly, this “open” approach allows customers to avoid costly vendor lock-in.
Related Blog Posts
Direct from Development – PowerEdge MX-Series Fabric Management
Tue, 10 Nov 2020 19:37:43 -0000|
Read Time: 0 minutes
Storage technologies have been evolving at an accelerated rate driven primarily by changes in workloads. Solutions like virtualization, big data and in-memory databases utilize storage in very different ways. This evolution results in customers needing flexibility in their server infrastructure to meet these diverse needs.
The PowerEdge MX-Series has been designed specifically to reduce this complexity while at the same time, deliver the robust, flexible storage architecture that customers need to optimize for all of these environments.
In many cases, workloads are now optimized for specific storage solutions. Virtualization environments benefit from the low latency, high capacity capabilities of Software Defined Storage (SDS). Scale-up In-memory Databases benefit from locally attached SSD’s for faster Database loading and Scale-out In-memory Databases require low latency shared storage solutions like SAN for shared access to data and fast database loading.
Adding to this complexity is wealth of options available for connection to storage environments. Traditional storage attach would include Fibre Channel (FC) or iSCSI and in recent years, Fibre Channel over Ethernet (FCoE). Conversely, newer technologies, like SDS, benefit from large pools of locally attached SAS storage using internal drive and/or JBOD enclosures.
MX storage connectivity options
To meet these diverse needs, the MX-Series has been designed to deliver a wealth of storage connectivity options. Traditional SAN connections are available through a pair of redundant Brocade 32Gb FC switches. In environments where SAS is required, the FC switches can be replaced with a pair of redundant SAS switches that connect to the optional MX5016s drive sled, with future support for external SAS JBOD enclosures. Traditional iSCSI, FCoE, and FC NPIV support is built into the integrated network switches and customers have the option to also use these devices to connect directly to a Dell EMC or 3rd party SAN.
FC direct connect allows a Fibre Channel storage array to connect directly to the MX9116n Fabric Switching Engine (FSE) without the use of external FC switches. In this mode, SAN access becomes available to all servers in the connected chassis as well as all servers installed in the chassis that are connected to the MX Scalable Fabric.
Internal Storage expansion
For customers looking to take advantage of SDS or simply to utilize SAS as their primary disk storage, the new MX-series offers enhanced local drive support with up to 6 hot plug drives in the MX740c and 8 hot plug drives in the MX840c. Both models also support 2 internal mirrored M.2 drives with the optional BOSS card. Additional expansion is available with the MX5016s storage sled. This sled houses up to 16 x 2.5” dual port SAS drives which connect through a pair of redundant SAS switches to provide additional storage internal to the enclosure. Drives hosted in this sled can be dedicated to a single server, be divided amongst all of the servers or, have their capacity aggregated and shared amongst all of the servers in the enclosure in operating environments that support this mode. For more demanding environments, multiple MX5016s sleds can be added to a single enclosure to increase the number of disks available to the servers.
External Storage expansion
For customers looking to connect to an existing FC SAN, the optional redundant FC switches support connection speeds of up to 32Gb/s and integrate into existing SAN fabrics without compromise.
Integration with existing FCoE or iSCSI environments can be accomplished natively through the optional Ethernet switch options for the MX. For FCoE, these switches can be configured to pass traffic directly through to an external gateway. Additionally, the MX9116n FSE can be configured as an NPIV gateway to an existing Brocade or Cisco MDS SAN.
As noted above, the MX9116n FSE can also be directly connected to a Dell EMC or 3rd party FC storage array. This mode provides access to shared storage that is dedicated to the devices connected to the MX Scalable Fabric and reduces complexity and latency to serve demanding application environments like big data.
As workloads have evolved, storage technologies have been dramatically affected. Customers now find that they often require multiple storage options in their environment to meet this need. Optimizing storage for a virtualization environment often means deploying a Software Defined Storage solutions like Microsoft Storage Spaces, VMWare vSAN or Dell EMC ScaleIO. This approach requires large pools of server based storage devices. Optimizing for solutions like in-memory database can be more complex with scale-up solutions requiring extremely fast server based storage and scale-out solutions benefiting from SAN based storage solutions. Similarly, big data solutions may require extremely large pools of centralized storage to deliver the capacity necessary for large data-sets.
Regardless of the storage technology required, the PowerEdge MX-Series delivers. From multiple connection options for SAN to a robust, flexible architecture for server based storage, customers can configure the MX system for uncompromised storage optimization.
Direct from Development – PowerEdge MX-Series Fabric Management
Tue, 10 Nov 2020 19:04:14 -0000|
Read Time: 0 minutes
The Dell EMC PowerEdge MXSeries offers a wide range of connectivity options. Optional switch modules provide full L2/L3 switching capabilities and implement industry standard Open Networking technologies for true software defined networking. Additionally, customers can choose to enable SmartFabric Services to greatly reduce the administrative burden of switch management.
The new PowerEdge MX-Series platform offers a wealth of networking options designed to scale from a single chassis to many. Offering a Future Forward design, MX-Series networking offers end to end 25GbE/100GbE connectivity while still maintaining backwards compatibility with 1/10/40GbE networks.
For single chassis requirements, the MX5108n Ethernet switch provides low latency, high bandwidth network access and supports a variety of uplink types.
Figure 1. MX5108n
For more demanding environments, the MX9116n Fabric Switching Engine reduces latency and increases bandwidth even further, while scaling up to 10 chassis. This model adds 32G Fibre Channel and Fabric Expansion capabilities via Dell EMC’s Scalable Fabric Architecture. The MX9116n also supports heterogeneous rack servers and other Ethernet devices, for a total of 104 devices.
Both switches support Open Networking, Full-switch and SmartFabric operating modes.
Figure 2. MX9116n
Each system supports ONIE, enabling the option for third party SDN solutions from Dell EMC Networking partners. Most installations, however, will be deployed with Dell EMC Networking OS10 Enterprise Edition, which can operate in two modes:
Full-Switch mode provides industry standard L2/L3 switching with full customization and can be configured using the Command Line Interface (CLI), API, or DevOps solutions such as Ansible.
SmartFabric Services is the latest Dell EMC Networking innovation and is designed to greatly reduce the complexity of fabric configuration, management, failure remediation and firmware updates. SmartFabric Services also provides I/O aggregation, Active-Active multi-pathing and physical topology validation.
In SmartFabric mode, these devices provide I/O aggregation to eliminate complex upstream switch configuration. By providing Active-Active multi-pathing to all hosts, load balancing and fault tolerance is delivered seamlessly. In addition, error prone tasks like VLAN and priority assignments are attached to server profiles and drive automated Quality of Service (QoS) configuration on the switch. Management is achieved with a single CLI and/or API to manage multiple switches and multiple fabrics.
With an unprecedented level of automation for a Modular design, Firmware updates are managed at the “Fabric level” to ensure consistency and in the event of replacement, the new device will be updated automatically and physical topology confirmed before the new switch is brought on-line.
SmartFabric Services automatically detects misconfiguration and link level failure conditions and notifies the administrator of the error as well as corrective actions required. Upon resolution, the system will automatically “heal” the fabric once the conditions are resolved.
Scalable Fabric Architecture
One of the critical design elements that have made Modular servers so popular is their ability to consolidate administration functions for all servers into a single management console. The new MX Scalable Fabric Architecture extends this concept even further with the development of the new MX7116 Fabric Expander Module.
Figure 3. MX7116 Fabric Expander Module
With this innovation Dell EMC Engineers were able to deliver a robust fabric expansion solution that integrates with the MX9116n Fabric Switching Engine to connect up to 10 MX-Series chassis together into a single, manageable domain. Installed in pairs, full fault tolerance can be achieved without adding additional switches. This device is completely unmanaged, performs no switching functions, runs no OS, and has a latency of only 55ns. It reduces cabling and provides a seamless connection back to the MX9116n with no port to port oversubscription.
The Dell EMC MX-Series was designed specifically to help customers facilitate their move to a more automated, resilient and high performing infrastructure. For those customers not yet ready to make this move, traditional management process can be used to configure the networking elements of the MX-Series infrastructure using Full-Switch mode or Open Networking but for those looking for ease of use, self-healing technologies, automated configuration of VLAN and QoS as well as the convenience and cost savings of reduced administration, SmartFabric Services is the ideal solution.