8 Warning Signs You Need A Better Data Center Interconnect
Fady Masoud is Senior Advisor for Technical Marketing within Ciena's Portfolio Solutions Group responsible for data center interconnect (DCI) solutions. He is co-author of the Experts Guide to OTN and brings 20 years of expertise and insights on architecture and requirements of next-gen optical platforms. This article originally appeared in Network Computing.
As cloud applications, devices, and traffic grow, the networks between and within your data centers may be showing their age.
According to Ovum, there are more than 6,500 data centers worldwide. The top 20 global cities have more than 2,200 data centers, some containing more than 100 facilities within their metropolitan boundaries alone. These data centers are the heart of our digital economy, and the connectivity within and between them is the key to application performance. Here we outline eight key reasons it might be time to revisit that infrastructure.
1. You're planning to move critical business functions to the cloud.
According to Forrester Research, the public cloud market is estimated to reach $191 billion by 2020. Cloud services enable convenient, secure and on-demand network access to a shared pool of configurable computing resources that can be rapidly provisioned and released with minimal management effort or service provider interaction. However, from software testing and customer relationship management to virtual machine migrations, those mission-critical cloud applications will dictate how networks are built.
2. Routine storage backup to your off-site data center takes days instead of hours.
In current cloud deployments, the network is rigid, static and often un-optimized for the type of connectivity required by the various building blocks of the cloud network. Link capacity is pre-set between endpoints based on the estimated amount of traffic to be carried across the network. This capacity remains unchanged regardless of the state of the cloud network (time of day, congestion on certain links, etc.). For spans that require some high-bandwidth connectivity, network planners tend to "over engineer" these links with extra capacity that is often not needed.
Moreover, almost all network operations -- such as setting up and tearing down a connection between two sites, increasing the bandwidth of a certain span, etc. -- require manual and cumbersome operations that negatively impact the performance of the cloud network. To address these challenges, the network has to be a dynamic pool of resources for the various applications in the cloud environment, operating in a "client-server" mode of operation where the hardware and software (application layer) can automatically and autonomously interact with the network to perform a specific task.
3. Last week's VM migration brought the entire corporate network to a standstill.
Virtual machine migration can be used for workload balancing of hot spots to less busy sites or move processing power closer to the applications requesting it. For example, an occasional VM migration would take 1.5 8-hour days to migrate 10 VMs with approximately 500G of data between two sites using a 40 Mbps connection in a static transport network. The flexibility provided by a more dynamic network allows the creation of a temporary high-bandwidth connection (such as 1,000 Mbps) for the duration of this task, reducing the migration time from 1.5 8-hour days to 66 minutes in a fully automated environment.
4. You've got corporate data stored in five different data centers throughout the metro, and it would be faster to take a cab to the data center to transport data than to send it over your existing connection.
The proliferation of services and applications that require instant access to content has been the main driver behind the ever-growing demand for data centers in metropolitan areas – with the goal of pushing content closer to the end user. The enterprise migration to cloud has also fueled the deployment of data centers across a variety of applications, such as site mirroring or backup. Other services, such as email or web pages, do not require instant access, so they can be hosted from data centers in lower cost, remote areas.
Your network manager should not need to camp out in the data center for a week when he needs to provision a new connection between data centers, when he could just use a mobile device to configure a modern solution. Data center networks are constantly changing, resulting in traffic trends that are difficult to predict, given the spontaneous access to the resource pool by a wide variety of users and applications. If your network manager is spending most of his time in camping gear, a better option would be to look at ways to automate operational tasks through the use of APIs and associated applications.
5. Your data center is crowded with bulky hardware and cabling, and your energy bill is through the roof.
Rather than bleed capital and operational dollars, it's time to look at data center interconnects with smaller footprints and lower power consumption that can scale.
6. Information is the lifeblood of your business, and it must be stored, protected, and easily accessible.
The ever-increasing rate of data breaches has put significant emphasis on ensuring that data is safe wherever it resides, be it at rest or in flight between data centers. While disk encryption and stringent rules for access of stored data are widely deployed to protect data from intruders, protection is needed for information leaving one data center to the moment it enters another data center over the interconnecting network.
7. Your need for timely data is constant, but capacity requirements are unpredictable.
Traffic trends on cloud-based networks are difficult to predict, given the spontaneous access to the resource pool by a wide variety of users and applications. A network control plane that can react to network changes -- such as one of multiple simultaneous failures, a change in network topology, or an increase in latency in some of the network's critical spans -- and handles and executes daily operations -- such as bandwidth increase requests, setting up a new connection between two end-points, or modifying an existing connection -- in real time, without requiring any human intervention, is the next step.
8. You need your back office tools and scripts to talk with your data center, and it's just not possible with your current network.
Data center interconnects need to be open and programmable to allow operators to leverage existing software tools. That way they can automate manual tasks and extend the IT practices currently used in servers toward the networking function. An open application development environment allows data center operators to create, test, and fine-tune customized applications for their own needs.