WANs for the Cloud takes center stage at Cloud Connect
The critical problems associated with the WAN for cloud service delivery were explored at UBM/Tech Web’s Cloud Connect 2012 conference last week in Santa Clara, CA. The WAN is a huge issue because VPN tunneling over the public Internet will not provide the security, availability, reliability, and performance needed by many large cloud users. Hence, most users will require some combination of a shared private network (IP-MPLS, Carrier Ethernet or Content Delivery Network) for their private clouds. Moreover, most companies must connect their IP-MPLS private networks to the public Internet or a managed IP network to get access to public or hybrid Cloud Computing services (unless the Cloud Service Provider is willing to make it’s Cloud Data Center a node on the customer’s private network). And what about mobile access from all the 3G/4G smart phones and tablets being sold? Will it be just wireless IP or something more to provide QoS?
Although WAN standards for cloud computing are urgently needed, they are severly lacking. We’ve been pounding the table on this point for years and have been very disappointed with the “castles in the sky” approach taken by the ITU-T Cloud Focus Group (FG-Cloud). We have written summary reports for each of their meetings, but they’ve wrapped up their work and have disolved. Please see these two seminal reports:
- What’s UNI, NNI and Network Infrastructure Needed for Cloud Computing
- ITU-T FG Cloud Wraps Up: 7 Deliverable Documents + Jan 9, 2012 Workshop on Cloud Computing and Smart Grid
Here are a few of the unanswered questions for the Cloud delivery WAN:
- What is the UNI and NNI for WAN used to deliver Cloud Computing services?
- How does a private IP-MPLS network interface to a public cloud network?
- How do two private/managed cloud networks interconnect (e.g. Intercloud NNI)?
- How does the provider do cloud service creation and end to end provisioning?
- What about Service orchestration and service management?
- OSS/BSS: modify or redo?
- Federation and Identity Management? (Needed for single sign on to a federated cloud that is actually composed of multiple interconnected clouds)
- And common APIs to access various cloud services? (today, Amazon and Rack Space have their own provider specific APIs to access their cloud service offerings)
It appears that these important cloud networking functions will all be proprietary for some time with no serious standards work being done at this time. Making IP-MPLS VPNs and Carrier Ethernet services into on-demand offerings to match the on demand nature of Cloud Computing is not a simple proposition. Nor is adding on-demand capabilities to existing portals.
Will Telcos be Viable Cloud Service Providers?
Is it an oxymoron to ask if network operators can be successful as cloud service providers?
For quite some time, we’ve believed that network operators have a great opportunity to leverage their network by getting in the cloud service business, especially partnering with a cloud software/web hosting company to offer Infrastructure as a Service. That would enable them to reap some of the value from their network that other public and private cloud providers are monetizing. After all, the carriers own the network, they own the subscriber, they are used to delivering on five-nines availability/reliability, they have Data Centers (primarily for OSS/BSS) which have excess capacity, and they could provide turnkey apps and services ready for consumption by SMBs and larger enterprises. Some pundits say that there’s no reason telcos can’t deliver compute services like Amazon’s EC2 or storage services like Amazon’s S3, providing all of the necessary management infrastructure for migrating workloads from private to public clouds and even between multiple public clouds. But are they up to the challenge?
Two carriers -Verizon and NTT America – spoke at Monday’s Cloud Carrier Forum about their end to end Cloud Services. Savvis, a leading provider of a converged cloud offering, also participated in a panel session. We’ve previously written about Savvis and their cloud network proposition was presented at the October 2011 IEEE ComSoc SCV meeting. All of those presentations can be downloaded for free from: http://www.ewh.ieee.org/r6/scv/comsoc/ComSoc_2011_Presentations.php
AT&T, which was a no show at the Carrier Cloud Forum on Monday (cancelling at the last minute with no reason given), announced a virtual private cloud offering that same day. It allows customers to securely migrate their VMware-based workloads between their own data centers and AT&T’s cloud. It’s called, “AT&T Synaptic Compute as a Service” and works along with VMware vCloud® Datacenter Service. The new enterprise-class cloud service combines virtual private networking (VPN) technology from AT&T with cloud infrastructure software from VMware.
Another AT&T deal with open stack champion Cloudscaling was alluded to at the Carrier Cloud Forum by another panelist, but it was not formally announced. Nor would Cloudscaling’s CTO answer my question of what work they were doing with AT&T.
Verizon and NTT America detailed the many steps they are taking to capitalize on their natural advantages as carriers in an effort to garner attention from enterprises looking to benefit from the cloud. Mike Palmer, VP of Product Management for Verizon Enterprise seemed to be very pragmatic and focused on doing what it takes for Verizon to succeed as a Cloud Service Provider. The company has taken a hard look at operations for the cloud. Rather than tweak their existing OSS/BSS’s they are revamping and starting from scratch to meet the new cloud business model. Cloud is a “service model” according to the company. Verizon is building a new billing system to enable attribute-driven OSS/BSS that lets it bill for third-party products, as well as its own. They’re 18 months into that process, with a couple more years of ongoing investments in the OSS/BSS space.
More about that in this video: Verizon Revamps BSS for Cloud
Regarding “pay for what you use” access to cloud services, Mr. Palmer said, “it would be much easier to deliver what customers want if their facilities were more conveniently located on the existing networks.” Again, this spotlights the problem of interconnecting the customers private IP MPLS network with the cloud service provider network or the public Internet.
NTT America’s keynote talk, “Cloud Reality for the Global Enterprise,” CTO Doug Junkins discussed the unique requirements global companies face and provide an example of how one company has made the successful transition. “The cloud is not the cloud without the network,” said Mr. Junkins and we couldn’t agree with him more. Large enterprises, especially multi-national companies are looking at cloud as a viable IT option and this is the area NTT is focusing on with its global virtual data center offerings. NTT is seamlessly interconnecting their data centers and making them available to large cloud users all over the world via their global IP backbone. NTT America will be able to deliver End-to-End Service Management, according to Mr. Junkins. Their global Data Centers (DCs) appear as one DC to the customer from a network perspective. The cloud offering provides Any-To-Any Connectivity as well as Virtual Machine Portability.
The problem NTT America has in providing this packaged virtual global DC service is they don’t own the access network to the customer (outside of Japan). Therefore, they must interconnect with other network operators to deliver their on demand cloud services to the enterprise customer on as add needed/ pay as you go basis. Here again is where the missing Cloud NNI standard is urgently needed. IEEE had started work on such an Inter-Cloud standard, but it stalled after the first meeting last summer.
For more on NTT-America’s planned Cloud offering, please refer to this video:
NTT Builds Global Virtual Data Centers
We’ve observed that most carriers haven’t developed a well thought-out strategy to deliver cloud services. Cloud Scaling’s CTO & Founder- Randy Bias said that instead of thinking through what’s really needed for cloud, telcos taking the “spaghetti-on-the-wall” approach of seeing what sticks. “That creates a lot of froth but it doesn’t really get us where we need to get,” he said.
Mr. Bias believes that telcos aren’t recognizing the value of putting product marketing people behind their cloud offerings. “I’m not seeing carriers take traditional models that startups use and have an owner for a new service or capability, and do the research before it comes to market,” he said during a Carrier Cloud panel session on Monday afternoon.
Since carriers own the network, one would think that for priotitized real time applications they could control bandwidth, latency and jitter. A lot of cloud-ready apps like those from Zynga and Netflix, have huge latency issues. In response, Bias said, “carriers ought to be building applications, for mobile in particular, where they can actually use their latency and all their natural assets as leverage.”
So in the end we’ll have to wait and see which telcos will succeed in the cloud computing business. From what we’ve seen, we have a short list of only three that are on target: Savvis, Verizon and NTT America. But last mile access may still be a problem for any competitive (Savvis) or global carrier (like NTT) that has to make a deal with another carrier to reach the cloud customer’s premises. We are very curious about what AT&T and Cloud Scaling were supposed to announce, but didn’t at Cloud Connect 2012.