Monthly Archives: December 2017

Sizing Up the Dell-EMC Merger, One Year Later

Still the biggest merger in the IT industry to date, Dell’s acquisition of EMC was the union of a server empire and a storage kingdom. On the surface, the market synergies of the joined pair far outweighed the product overlaps. It’s been more than a year since the deal was finalized and it’s time to check see if it has lived up to its promise.

Bear in mind that the acquisition was a response to the changing demographics of the industry, basically to defend against the race to the bottom in systems hardware as IT continues to migrate to the cloud. Simply put, cloud service providers buy gear in huge volumes from original design manufacturers typically in Taiwan or China, at much lower prices.

These commercial off-the-shelf units are becoming mainstream amid a recognition that they are functionally the same as more expensive units from long-established brands. This cheap equipment has helped create an “internet-priced” approach to buying SSDs and hard drives. For example, the list price of an EMC drive can be as much as 14X the price of one from online distributor Arrow Electronics.

The bottom line is that both companies, and perhaps especially EMC, saw the writing on the wall. The decline of hardware revenue will take a few years, but it’s inevitable. With this as background we can consider whether the union of the companies, now called Dell Technologies, is a success.  

Financial performance

The obvious first place to look is financial. Tom Sweet, the CFO of Dell Technologies, handled the bookkeeping merger very well indeed. Decision-making has been wisely delegated as much as possible, which should make for a lean and agile company going forward.

What are the results so far?  Dell Technologies closed its 2017 fiscal year in March with $62 billion in revenue (with 53 days of EMC not included) and in the second quarter of fiscal year 2017, it posted  posted $19.3 billion, looking good. Dell carries a high debt burden from the acquisition, but has paid down $9.6 billion already, including selling off its service/software operation for $5.3 billion. The company’s third-quarter results are also strong on the revenue side at $19.6 billion, but while cash flow is $1.6 billion, it lost $533 million. Cash is king, obviously, but this shows that a lot of work still remains to make the combined entity cost efficient.

There are some red flags on the fiscal horizon. The company grew only $34 million year over year  in total enterprise storage sales, losing market share in the process. HPE has a small lead, but also lost share. The big winner was the ODM segment with collectively the largest revenue and a whopping 55% gain in market share, which seems to add credence to those fears that triggered the union of the companies. Server sales mirror this, though there is less of a growth gap with ODMs. In fact, it looks like the ODMs are in fact achieving the market penetration Dell and EMC feared and this is a long-term systemic change.

Technology challenges

The combined entity also faces some technical challenges. First, it still must consolidate overlapping storage systems. I expect one major move will be lowering the emphasis on RAID arrays, EMC’s signature product. All of the iSCSI product lines will need rationalization; there are just too many to allow the salesforce to be efficient, just like the auto industry with all of its brands back in the 1970s.

These changes imply a  threat to internal engineering and marketing empires. It will be interesting to see how this plays out in a environment that’s seeing the most change in four decades. Will protection of product families interfere with retooling the design effort to accommodate trends such as modular computers, hyperconvergence, and solid-state drive solutions?

Likely, we will see some strain between the established Dell units and the need to adopt new technologies, as already indicated by the number of OEM deals fueling new product areas Dell is pursuing, such as the partnership with hyperconvergence leader Nutanix. I would add, though, that Dell is nimble in the OEM space and appears to be signing up a deeper and more first-rate portfolio than its competitors.

The cloud

For all the traditional system vendors, the elephant in the room is the cloud. There is no question that IT is pulled to the cloud and that the percentage of IT run in the cloud will continue to increase. Dell and the others have so far not done well in competing with the ODMs for cloud sales to the top three cloud-service providers — Amazon, Google and Microsoft — and it’s unlikely they will penetrate the rapidly expanding Chinese cloud market either.

The reason is business cost models. Chinese ODMs run very lean indeed, without big planning staffs and minimum HR. Most of their R&D is based around reference designs from vendors such as Intel, rather than designing from scratch. The reality is that the traditional US suppliers such as HPE, Cisco and Dell have followed the cloud providers to the ODMs, which supply many parts to them, if not complete branded systems.

Dell will have to look to add value to what is a Chinese-sourced commodity server or storage appliance. This is a place for integration and support, and signals more rapid devaluation of the EMC side of the business, as proprietary RAID units are replaced with inexpensive COTS products.

Source link

IPv6 Auto-Configuration in Linux |

In Testing IPv6 Networking in KVM: Part 1, we learned about unique local addresses (ULAs). In this article, we will learn how to set up automatic IP address configuration for ULAs.

When to Use Unique Local Addresses

Unique local addresses use the fd00::/8 address block, and are similar to our old friends the IPv4 private address classes:,, and But they are not intended as a direct replacement. IPv4 private address classes and network address translation (NAT) were created to alleviate the shortage of IPv4 addresses, a clever hack that prolonged the life of IPv4 for years after it should have been replaced. IPv6 supports NAT, but I can’t think of a good reason to use it. IPv6 isn’t just bigger IPv4; it is different and needs different thinking.

So what’s the point of ULAs, especially when we have link-local addresses (fe80::/10) and don’t even need to configure them? There are two important differences. One, link-local addresses are not routable, so you can’t cross subnets. Two, you control ULAs; choose your own addresses, make subnets, and they are routable.

Another benefit of ULAs is you don’t need an allocation of global unicast IPv6 addresses just for mucking around on your LAN. If you have an allocation from a service provider then you don’t need ULAs. You can mix global unicast addresses and ULAs on the same network, but I can’t think of a good reason to have both, and for darned sure you don’t want to use network address translation (NAT) to make ULAs publicly accessible. That, in my peerless opinion, is daft.

ULAs are for private networks only and should be blocked from leaving your network, and not allowed to roam the Internet. Which should be simple, just block the whole fd00::/8 range on your border devices.

Address Auto-Configuration

ULAs are not automatic like link-local addresses, but setting up auto-configuration is easy as pie with radvd, the router advertisement daemon. Before you change anything, run ifconfig or ip addr show to see your existing IP addresses.

You should install radvd on a dedicated router for production use, but for testing you can install it on any Linux PC on your network. In my little KVM test lab, I installed it on Ubuntu, apt-get install radvd. It should not start after installation, because there is no configuration file:

$ sudo systemctl status radvd
● radvd.service - LSB: Router Advertising Daemon
   Loaded: loaded (/etc/init.d/radvd; bad; vendor preset: enabled)
   Active: active (exited) since Mon 2017-12-11 20:08:25 PST; 4min 59s ago
     Docs: man:systemd-sysv-generator(8)

Dec 11 20:08:25 ubunut1 systemd[1]: Starting LSB: Router Advertising Daemon...
Dec 11 20:08:25 ubunut1 radvd[3541]: Starting radvd:
Dec 11 20:08:25 ubunut1 radvd[3541]: * /etc/radvd.conf does not exist or is empty.
Dec 11 20:08:25 ubunut1 radvd[3541]: * See /usr/share/doc/radvd/README.Debian
Dec 11 20:08:25 ubunut1 radvd[3541]: * radvd will *not* be started.
Dec 11 20:08:25 ubunut1 systemd[1]: Started LSB: Router Advertising Daemon.

It’s a little confusing with all the start and not started messages, but radvd is not running, which you can verify with good old ps|grep radvd. So we need to create /etc/radvd.conf. Copy this example, replacing the network interface name on the first line with your interface name:

interface ens7 {
  AdvSendAdvert on;
  MinRtrAdvInterval 3;
  MaxRtrAdvInterval 10;
  prefix fd7d:844d:3e17:f3ae::/64
                AdvOnLink on;
                AdvAutonomous on;
                AdvRouterAddr off;


The prefix defines your network address, which is the first 64 bits of the address. The first two characters must be fd, then you define the remainder of the prefix, and leave the last 64 bits empty as radvd will assign the last 64 bits. The next 16 bits after the prefix define the subnet, and the remaining bits define the host address. Your subnet size must always be /64. RFC 4193 requires that addresses be randomly generated; see Testing IPv6 Networking in KVM: Part 1 for more information on creating and managing ULAs.

IPv6 Forwarding

IPv6 forwarding must be enabled. This command enables it until restart:

$ sudo sysctl -w net.ipv6.conf.all.forwarding=1

Uncomment or add this line to /etc/sysctl.conf to make it permanent:

net.ipv6.conf.all.forwarding = 1

Start the radvd daemon:

$ sudo systemctl stop radvd
$ sudo systemctl start radvd

This example reflects a quirk I ran into on my Ubuntu test system; I always have to stop radvd, no matter what state it is in, and then start it to apply any changes.

You won’t see any output on a successful start, and often not on a failure either, so run sudo systemctl radvd status. If there are errors, systemctl will tell you. The most common errors are syntax errors in /etc/radvd.conf.

A cool thing I learned after complaining on Twitter: when you run journalctl -xe --no-pager to debug systemctl errors, your output lines will wrap, and then you can actually read your error messages.

Now check your hosts to see their new auto-assigned addresses:

$ ifconfig
ens7      Link encap:Ethernet  HWaddr 52:54:00:57:71:50  
          inet6 addr: fd7d:844d:3e17:f3ae:9808:98d5:bea9:14d9/64 Scope:Global

And there it is! Come back next week to learn how to manage DNS for ULAs, so you can use proper hostnames instead of those giant IPv6 addresses.

Learn more about Linux through the free “Introduction to Linux” course from The Linux Foundation and edX.

Dell to Disable Intel’s Insecure IME » Linux Magazine

Intel’s IME (Intel vPro Management Engine) came under fire recently when security researchers found serious bugs that allowed a remote attacker to take control of the affected systems.

“The exploitation allows an attacker to get full control over business computers, even if they are turned off (but still plugged into an outlet). We really hope by bringing this to light, it will raise awareness about security issues in firmware and avoid possible issues in the future,” wrote Embedi, the security firm that discovered the bug.

Intel doesn’t share any information about these “secretive” Management Engine technologies. ME modules sit above the operating systems and users have no access or control over the technology. Organizations like EFF are calling for more transparency around ME modules. EFF asked Intel to “Provide a way for their customers to audit ME code for vulnerabilities. That is presently impossible because the code is kept secret.”

Because Intel doesn’t provide any such information, PC vendors and users don’t have any means to audit or fix such vulnerabilities. Now one PC vendor has taken steps to protect its users. Dell is now disabling IME in all new systems, and users will have to pay to enable the service.

In a statement to ExtremeTech, Dell said, “Dell has offered a configuration option to disable the Intel vPro Management Engine (ME) on select commercial client platforms for a number of years (termed Intel vPro – ME inoperable, custom order on Some of our commercial customers have requested such an option from us, and in response, we have provided the service of disabling the Management Engine in the factory to meet their specific needs. As this SKU can also disable other system functionality it was not previously made available to the general public.”

PC vendors, especially those selling Linux preloaded systems, are following the suite and disabling ME by default. Dell is the biggest PC vendor, and if other vendors start disabling the engine, Intel might be compelled to either open source the technology or offer more transparency around it.

Source link

KubeCon Concluded in Austin, Texas » Linux Magazine

Kubernetes has become the Linux of the cloud. It has seen massive adoption in the last three years. The first release of Kubernetes was announced in 2014. All three major cloud providers, including Google (the creator of Kubernetes), Microsoft, and AWS now support Kubernetes. Even Docker started offering Kubernetes as an orchestrator along with its own orchestrator Swarm. Cloud Foundry has adopted Kubernetes as Cloud Foundry Container Runtime, and OpenStack vendors have already adopted Kubernetes to deploy OpenStack as an application. All major Linux vendors, including Red Hat, SUSE, and Canonical offer Kubernetes distributions.

The adoption and growth of Kubernetes was the theme of KubeCon, the Kubernetes conference that was held between December 6 and 8 in Austin, Texas. During the conference, Oracle open sourced its Kubernetes tools for serverless deployment and multicloud management.

Microsoft announced that Azure would bring new serverless and DevOps capabilities to the Kubernetes community, and Bitnami launched a new in-cluster Kubernetes Application Consol.

The Kubernetes community announced the 1.0 release of CoreDNS, a cluster DNS for Kubernetes. JFrog and Baidu joined CNCF (Cloud Native Computing Foundation), the home of Kubernetes, as Gold members.

Source link

7 Enterprise Storage Trends for 2018

Enterprises today are generating and storing more data than ever, and the trend shows no sign of slowing down. The rise of big data, the internet of things, and analytics are all contributing to the exponential data growth. The surge is driving organizations to expand their infrastructure, particularly data storage.

In fact, the rapid growth of data and data storage technology is the biggest factor driving change in IT infrastructure, according to the Interop ITX and InformationWeek 2018 State of Infrastructure study. Fifty-five percent of survey respondents choose it as one of the top three factors, far exceeding the need to integrate with cloud services.

Organizations have been dealing with rapid data growth for a while, but are reaching a tipping point, Scott Sinclair, senior analyst at ESG, said in an interview.

“If you go from 20 terabytes to 100 terabytes, that’s phenomenal growth but from a management standpoint, it’s still within the same operating process,” he said. “But if you go from a petabyte to 10 or 20 petabytes, now you start taking about a fundamentally different scale for infrastructure.”

Moreover, companies today see the power of data and understand that they need to harness it in order to become competitive, Sinclair said.

“Data has always been valuable, but often it was used for a specific application or workload. Retaining data for longer periods was more about disaster recovery, having an archive, or for regulatory compliance,” he said. “As we move more into the digital economy, companies want to leverage data, whether it’s to provide more products and services, become more efficient, or better engage with their customers.”

To support their digital strategy, companies are planning to invest in more storage hardware in their data centers, store more data in the cloud, and investigate emerging technologies such as software-defined storage, according to the 2018 State of Infrastructure study. Altogether, they’re planning to spend more on storage hardware than other infrastructure.

Read on for more details from the research and to find out about enterprise storage plans for 2018. Click on the row of buttons below or on the arrows on either side of the images. For the full survey results, download the complete report.

(Image: Peshkova/Shutterstock)

Source link