The Future of The Network

What lies beyond Gigabit Ethernet? Will IPv6 go mainstream anytime soon? How can companies reduce the wastage of computing resources in server farms and data centres? Prashant L Rao peers into a crystal ball and tells you what the future holds for networking 

Ethernet has come a long way since it was first implemented in the 1970s. In its 30 years of existence it has become ubiquitous; a widely standardised, plug-and-play technology that is used in over 90 percent of corporate LANs. Ethernet originally ran over thick co-ax and provided users with a shared 10 Mbps bandwidth connection. It soon progressed to running over unshielded twisted pair and offering dedicated 10 Mbps connections using switches. Today, switched Fast Ethernet enables dedicated 100 Mbps to the desktop with 1 Gbps trunks.

Gigabit Ethernet

 
Both 10 Mbps and 100 Mbps Ethernet have evolved over the last 20 years to become the local networking standard because of their reliability, low cost and ease of installation. However, streaming multimedia applications are becoming commonplace in the industry, and 100 Mbps doesn’t quite cut it for all applications. That’s where Gigabit Ethernet comes into the picture. With Fast Ethernet you had 10/100 switches; in the case of Gigabit you have 10/100/1000 switches. 10 Gigabit Ethernet is currently in development. This technology will work only with fibre, unlike Gigabit that also works over copper.

Enterprises have been using routers to connect to WANs through E1 pipes (2 Mbps). Analysts predict that while Ethernet switches are capable of delivering 1 Gbps, routers that support a maximum speed of 45 Mbps could become the main bottleneck on enterprise WANs. Some even predict that routers may not be used in future.

The physical performance of Gigabit Ethernet is most often compared to ATM, but the debate continues on Gigabit Ethernet taking on duties formerly handled by ATM. Gigabit Ethernet has made substantial strides towards ATM capabilities with the introduction of frame-based quality-of-service (QoS) features and IP switching.

Beyond Gigabit

“10 Gbps is very close to memory speed. Putting 10 Gbps won’t improve performance for that reason. 10 Gbps will be used to consolidate bigger networks,” says Pramod S, systems specialist at Apara. S Vishwanathan, S E manager at Wipro Infotech says, “The chances of any other technology replacing Ethernet are dim.” There’s a roadmap all the way from Gigabit to 10 Gbps to 100 Gbps. But as Vishwanathan points out, “No application demands this kind of bandwidth—plus you can have multiple links on 10 Gbps.”

Even Gigabit Ethernet is just barely present in the desktop segment. Vishwanathan estimates that barely 1-2 percent of the Indian market is using Gigabit on the desktop. Within a few years, industry experts predict 1 Gbps to the desktop and 10 Gbps trunks.

IPv6

Internet Protocol Version 6 (IPv6) has been designed to fix many shortcomings in today’s IP, IPv4. It has features such as automatic routing and network reconfiguration. Smaller packet headers also allow for faster processing (IPv6 headers have seven fields vs 13 in IPv4). IPv6 replaces IPv4’s 32-bit addressing with 128-bit addressing.

IPv6 simplifies many operations done by the existing IP in patched-up or afterthought implementations. Also, some of the capabilities of IPv6 are simply not there in IPv4. However, the cost factor in overhauling the existing IPv4-based Internet infrastructure will delay the deployment of IPv6. Eventually however, IPv6 will prevail.

Teevra Bose, national product manager, NBU at Apara says, “IPv6 is still on the test-bed. The lack of IPs in IPv4 is not a potential threat; people have learned how to work around the problem.”

VoIP

 
“In the corporate market, VoIP could overtake traditional telephony,” says Vishwanathan. There are a lot of quality standards that offer the highest priority to voice. “QoS has been taken care of,” he adds. Codecs exist that let you compress a voice channel to anything between 5 and 64 Kbps. 8 Kbps is the present standard. “The future of VoIP will parallel developments in IP technology. Everybody understands IP. The customer only has to maintain a single set-up. IP PBXs already support 60-70 percent of the functionality of a traditional PBX; it will take 3-5 years for IP PBXs to match the analog market,” concludes Vishwanathan.

Cisco has first-hand experience in all this, having built the largest VoIP network in the world in China; it has grown to transmit 500 million minutes a month.

iSCSI

 
iSCSI has to be implemented on a Gigabit infrastructure; this is cheaper than a fibre channel network, but not as cheap or as easy as a 10/100 network. Says Pramod, “iSCSI competes with fibre channel as an access protocol. The number of users is low at present, but by the end of 2003 it should catch up. iSCSI’s limitation is that it is not a complete standard. However, by the end of 2002 or early 2003 it should be one. In iSCSI TCP/IP performance overheads are high; the solution is TCP/IP Offload Engines (TOE). Security is another issue; SCSI was designed for dedicated storage. With iSCSI you have data flowing over the WAN, so security is now a requirement.”

TOE hasn’t really caught on in the market. M S Sidhu, managing director of Apara believes that iSCSI will only catch on along with 10 G Ethernet networks. That said, iSCSI has fundamental advantages: messaging (file) and storage (block) data can be sent over the same wire. This simplifies SAN configurations. The fact that iSCSI builds upon SCSI, TCP/IP and Ethernet is expected to reduce the steepness of the adoption curve.

Unified storage

 
“Unified storage is the ultimate goal,” says Banda Prasad, practice head for storage services at Wipro Infotech. Unified storage is defined as storage that supports SAN and NAS as well as emerging protocols such as iSCSI, either natively or through gateways. This new technology is being hailed as a panacea for all the ills of today’s storage networks.

According to Gartner, on an average an enterprise spends $3 managing storage for every $1 spent on storage hardware. Unified storage solutions eliminate constraints as they accept both file and block requests simultaneously over the same wire.

DAS and NAS offer limited scalability. Once a DAS or NAS installation reaches its storage capacity limit, additional servers must be deployed with their own islands of storage. SAN can scale, but SANs are complex and costly. It is tough to add storage in all the above technologies without some degree of service-disruption occurring.

Unified storage simplifies scalability without disruptions because it uses a common management layer that automatically reconfigures the underlying storage subsystem without administrative intervention. Capacity is simply added to the pool of available storage resources. With unified storage, a single resource manages everything and automatically reconfigures new storage or hardware elements. The storage pool, accessed by multiple topologies, can be scaled with significantly less management overheads than with traditional technologies.

Researchers estimate that 80 percent of today’s installed storage infrastructure is DAS. Customers with heavy DAS infrastructure face a number of challenges: the nature of storage and access is changing as digital images and video co-exist with database applications, and disk storage associated with a given DAS server is accessible only through that specific server, thus creating islands of data.

SANs have a high entry price, require special equipment and management, handle only block data, and are cumbersome to integrate with existing infrastructures. NAS solutions are suited for file-only applications, have limited scalability, and are difficult to manage and cluster.

Unified storage offers a fresh approach to data storage. It lets companies store file or raw block data directly on existing IP networks. The raw block transfer capability of NUS solutions enables distributed storage volumes to be seen as directly attached disks on the network. These solutions are protocol-agnostic, which enables them to integrate easily into existing heterogeneous storage environments. In addition, they can be purchased incrementally, allowing customers to grow storage in a modular and scalable manner. A virtualised cluster of unified storage devices can be partitioned and allocated as a raw block device running in conjunction with an Oracle application, or a simple network-attached file server.

Virtualisation over a SAN is another way of fooling 30-40 different applications into thinking that a storage box is dedicated to them. This is done by virtualisation of the switch.

One expected development in enterprise storage is a self-bootable device—essentially storage that starts itself up.

Cabling

 
Experts believe that it will take 3-4 years before fibre to the desktop catches on. Cat6, which supports Gigabit Ethernet, will become the preferred means of cabling using copper. The migration from existing Cat5 to Cat6 will be a phased one. Only those users who need greater bandwidth, perhaps those using CAD or multimedia authoring tools, will benefit significantly from this technology. Changing to a Gigabit backbone doesn’t come cheap; the cost works out to around Rs 1 crore for changing switches and rewiring.

Network management software

 
Experts feel that despite driving factors, it will be some time before the network management software (NMS) market in India matures to the level of commanding significant volumes. The cost of the software is currently on the higher side, restricting the market to larger enterprises; it will take another two years for NMS to make deeper inroads. Meanwhile, there is a thrust from the big three server vendors—IBM, HP and Sun—towards improving resource utilisation in data centres and server farms.

N1

 
Sun’s N1 is an effort at letting customers assign computing tasks to pools of servers, storage systems and network equipment without needing to allocate jobs to individual pieces of hardware. At the same time, N1 has the goal of improving resource utilisation in data centres. N1 competes with HP’s Utility Data Centre and IBM’s autonomic computing projects. “N1 builds the computer out of the network,” says Anil Valluri, director of systems engineering at Sun Microsystems India.

N1 will take advantage of virtualisation to let many computing jobs be handled by a pool of computing equipment rather than by a particular server. For this, Sun has acquired two start-ups, Terraspring and Pirus Networks. Terraspring’s software keeps track of computing equipment in a data centre and lets administrators provision systems for new jobs. Pirus sells hardware and software that virtualises storage systems from various manufacturers.

N1 will debut in Sun’s blade servers and Sun’s services that help customers prepare for the technology.

IBM’s autonomic computing

 
IBM’s initiative is a response to the problem of dealing with the consequences of the decade-long reign of Moore’s Law. Demand for skilled IT workers is expected to increase by over 100 percent in the next six years, and to help cope with that comes the idea of relieving humans of the burden of coping with computing complexity and passing it back to the computers. Say Big Blue’s researchers, “We need to develop autonomic computer systems that self-regulate, self-repair and respond to changing conditions without conscious effort on our part.” IBM is investing $2 billion into developing autonomic systems.

Autonomic computing aims to introduce self-configuring, self-healing, self-optimising and self-protecting capabilities in the whole range of computing systems, from desktop computers to mainframes to software. Its early efforts are focused on storage and software (Tivoli, DB2 and WebSphere). Autonomic computing capabilities are being built into IBM’s Shark product line through software that lets users configure and manage data across large server farms. IBM calls Tivoli Risk Manager 4.1 the first “autonomic security management software” capable of automatically monitoring a network’s health, protecting it from attacks, and healing it. 

Risk Manager’s ‘self-healing’ features include the ability to integrate with software distribution tools, including Tivoli Configuration Manager and similar third-party products. This feature lets Risk Manager push out security patches and software updates to devices under its management. WebSphere 5.0 server contains a number of self-optimising and self-healing features. DB2 8 has a feature called ‘Health Center’ that automatically updates a database administrator on system performance, offers advice about database problems or applications, and sends alerts when a fix has been generated.

HP Utility Data Centre

 
On an average, data centres have utilisation rates of 35 percent. Businesses would love to push that number up to, say, 75 percent, and reduce data centre costs without jeopardising service levels. That’s where HP UDC (self-adapting, self-healing and policy-driven) comes into the picture, allocating and managing computing power automatically. UDC supports multiple hardware vendors and operating systems, and lets users graphically assign servers to jobs as demand arises by booting servers in the data centre from a shared disk network that includes multiple operating systems and applications. The target segment? Telecom, financial services, managed service providers and infrastructure application service providers. The core of UDC is HP’s utility controller software that simplifies the design, allocation and billing of IT resources for applications and services.

UDC allows data centre infrastructures to be wired once, then provisioned virtually and managed on the fly-deploying new applications and services, activating new customers, offering usage-based billing.

Internet2

 
The Internet2 project kicked off in 1996. The US government, colleges, universities and companies such as IBM, Lucent, Cisco, and Nortel are involved in this project. 185 universities and research labs are working on Internet2. The goal behind creating Internet2 was to deploy advanced network applications and technologies to speed up the creation of a faster Internet. Internet2 is not available to the public. It uses a high bandwidth backbone, but supports a much smaller number of users than the public Net, three million versus several hundred million. Internet2 is fast; the slowest connection on Internet2 is 155 Mbps. It is designed to minimise the number of hops between routers to further speed things up.

Internet2 has been used to test futuristic applications such as 3D Brain Mapping, Remote Medical Specialist Evaluation (real-time interaction among non-physician experts and supervising physicians sharing large image data sets), Digital Video (including interactive two-way digital video), Remote Instruction and Teaching, and Remote Telescope Manipulation-to name a few.

Internet2 has been compared to a time machine showing where the [public] Internet will be in three to five years.

The network uses two high-performance optical backbones: MCI Worldcom’s very-high-performance bandwidth network service (vBNS), and Abilene, a 10,000-mile backbone built specifically for Internet2. Methods are being developed to give some transmissions higher priority. By marking data as “urgent,” researchers can ensure that real-time data (say, a video stream of a surgery) will cross the network before less time-sensitive data (e-mail). Multicasting allows a single data stream (such as a live video broadcast) to travel across the Internet and then split off copies of itself to multiple destinations. Compared to this, on the public Internet, the originating server must transmit a separate data stream to each user, greatly increasing congestion.

Once you have your CCNA certification the obvious next step is making sure you are employed in one the many CCNA jobs.  New graduates with a CCNA certification, sometimes find it hard to immediately walk into CCNA jobs, but there are jobs out there.  Having a good plan of attack is the best approach to dealing with this problem.

As with any job search, when looking for a CCNA jobs, be clear on the jobs you are qualified to do and don’t aim for those completely out of your league.  If you have just passed your CCNA and have limited prior experience you are simply not going to get a Network team leader role in a large organization.  You need to be realistic.

It Is Important To Research CCNA Jobs

Being realistic means you need to research what CCNA jobs are out there and what indicative salaries are being paid for these roles.  Finding out salary ranges is always the hardest part of any evaluation, but clues provided by online job search engines and industry publications will give you some indication.  Then, concentrate on applying for the CCNA jobs that are within your reality.
Work on tailoring your resume/CV for any CCNA jobs that you apply for.  If a company is looking for a particular skill make sure this is highlighted or prominent in some way on your resume.  There is no point in hiding your light under a barrel when it comes to getting a job.

CCNA Jobs – Prepare For Your Interview

Be well prepared for any job interview or any company interaction you may have.  It is all well and good to have the qualifications for c c n a. but in the end it is likely to be your personality and how you present yourself at the interview that

may swing the job decision.  Your resume will have told the employer you are worth interviewing, along with everyone else that is interviewed.  Your personal approach may be the difference between candidate A and B in the eyes of the employer.    Listen the questions that are asked, and answer what is asked – don’t just tell them what you want them to hear.

Use Time Wisely If The CCNA Jobs You Want Are Not Available

If it is taking some time to find the job you are looking for, use your time wisely.  Continue to build on the skills you have now.  The more qualifications or experience you have, the better off you will be on a comparison with others who are applying for the same CCNA job