When most people hear the phrase “operating system in the cloud” they usually think of a really cool client-side, Web-based desktop like EyeOS or CloudMe or even Chromium OS. Perhaps that is the future of client operating systems, but when cloud infrastructure providers talk about operating systems, they are making reference to which OS your cloud infrastructure will run on. And, it’s not always limited to just one in many cases.

At GoGrid, we provide a variety of operating systems including:

• Windows Server
• Ubuntu
• Red Hat Enterprise Linux
• Debian
• CentOS

Operating systems in the cloud are very important because they are what your applications and infrastructure are built upon. Whether you’re using the cloud to deploy test & development environments, act as your data center or run your company’s business critical applications, the operating system plays a vital role in cloud infrastructure.

Most IT professionals are pretty passionate about what operating system they prefer. For instance, search for “Windows vs. Linux” on Google – over 109 million results have some sort of opinion on the topic. But, since actions speak louder than words, we wanted to determine which operating system was used more by the IT industry.

In early 2011, as part of the GoGrid Cloud Survey Report, we polled over 500 CTOs, developers and IT professionals and asked them a variety of questions regarding cloud computing trends, best practices and use cases. Because there is such a wide variety of operating systems that serve different use cases, we decided to ask: Which OS do you use for your critical business applications?

The numbers exceed 100% because most businesses use several different server operating systems to run their critical applications (businesses either run several physical, virtual or mixed environments to accomplish this). When we look at the results, it isn’t terribly shocking that Windows Server is used by 80% of the sample – they’ve been dominating the server scene since Windows Server 2003 or earlier. However, the amount of Linux and Solaris servers is much higher than anticipated. In fact, if you do a search of various Linux distros, you can see how Ubuntu, for example, is making huge advances in the space.

Just last year, ZDNet shared stats from the International Data Corp. (IDC) that found 73.9% of servers were Windows and only 21.2% of servers were Linux based. Perhaps our 2011 survey data is showing Linux use in the cloud is on the rise.

Do you have any strong thoughts or opinions about operating systems in the cloud? Which server operating system do you use? How do you think cloud computing will affect server operating systems? Does the Operating System even matter? We look forward to your responses.

With only one more blog post left in the series, the GoGrid Cloud Survey Report is coming to a close. Our next post will also be our last survey results article, but it covers the most interesting cloud computing question thus far: Why aren’t you using cloud computing? Stay tuned…

For more information on our survey methodology or to see all of our results, please download the Cloud Survey Report.

In my previous post, “Set Up A Cacti Monitoring Server in Minutes with this GoGrid Community Server Image”, I showed how to use a Community GSI to quickly and easily set up a monitoring server on GoGrid running Cacti. In this second part of the Cacti GoGrid Community GSI blog series, I would like to walk you through how I connected my Cacti server up to an Ubuntu server (a node in my Glusterfs file storage array). First we will review the objectives we are looking to achieve and then dive into each one.

Objectives:

1. Confirm private network configuration on both Cacti server and Ubuntu server, and test connectivity.
2. Enable SNMP server and configure SNMP rocommunity string on Ubuntu servers.
3. Establish SNMP agent listening IP address
4. Create “Device” in Cacti console and confirm SNMP connectivity to Ubuntu server
5. Create Graphs – CPU usage, Load Average, Memory Usage, PING Latency, Processes, Eth0 Traffic, Eth1 Traffic
6. Repeat process for other Ubuntu servers in your network.

Confirm private network configuration on both Cacti server and Ubuntu server, and test connectivity

On GoGrid, you have the ability to network your servers together over a private network allocated to your account. (Note: all private networking within GoGrid is free.) We need to take advantage of this secure communication method to allow your Cacti server SNMP access to your servers. I recommend you use the private network IP range that is specified in your account – under the “List” tab then “Network”.

In this example we will be using the 10.129.151.0/24 network (yours will be different).

First, log into the Cacti server and assign a private IP address to the eth1 interface. To do this open up the /etc/sysconfig/network-scripts/ifcfg-eth1 file and configure the IPADDR parameter and NETMASK parameter. Now save the file and toggle the interface /etc/init.d/network restart or ifdown eth1 and ifup eth1.

I would recommend using ifdown (which brings a specific network interface in this case we would specify eth1 right after command) and following the completion of that command execute the ifup (which will bring up a specified interface). This serves the same function as restarting networking but doesn’t touch the other interfaces that we don’t need to restart.

Once this is done – confirm you now have a private IP address on the eth1 interface.

While we are at it – let’s add the host name of the Ubuntu server to the /etc/hosts file. You need to open up the /etc/hosts file and add a line containing the 10.x.x.x address you are going to give the Ubuntu host. In this case, the Ubuntu server has a hostname equal to Gluster2.

Renaming our Cacti server is the last step before we log out. The hostname of a GoGrid system is a unique string after initial deployment. Let’s change this to something that is more descriptive of what the host is. While in the Cacti server – modify the line HOSTNAME in /etc/hostname to “Cacti”, see image below. After saving and closing this file, reboot your Cacti system for this hostname change to take effect.

The second part of this objective is to bring up private network connectivity on the Ubuntu server – in this case Gluster2.

Log the Ubuntu server via SSH and open up the file – /etc/network/interfaces. In Ubuntu, the networking configurations for all interfaces are contained this file unlike the Cacti server which is running on CentOS. The screen shot below shows how I changed the eth1 inet from DHCP to Static and added the address and netmask parameters.

After the file /etc/network/interfaces has been updated – the network interface needs to be restarted.

You should now be able to ping the Cacti server via its private IP address. You may also want to add the Cacti server to the /etc/hosts file on your Ubuntu server. This is the same file and location as on a CentOS server (The Cacti image is saved on CentOS 5.5). After you have confirmed private network connectivity between the two servers, you can move onto the next objective.

Enable SNMP server and configure SNMP rocommunity string on Ubuntu servers.

This objective requires us to log into the Ubuntu server via SSH. From the SSH login, we need to install the necessary packages to get SNMP running on your Ubuntu server. Run the command below.

snmpd will be installed and configured to start on boot automatically

After successful installation of the packages for SNMP, it is time to edit the snmp.conf. Here we need to add the unique read-only community (rocommunity) string identifier and the host (Cacti) that will be accessing this server via SNMP inquires.

If you are following along on your own – add your own unique string anywhere in this file and reference your Cacti server’s private IP address which should be different than mine.

rocommunity <unique string> <private IP of your Cacti server>

Establish SNMP agent listening IP address

Next edit the file /etc/default/snmpd. This file contains default SNMP configurations. We are interested in modifying the default interface the SNMP daemon will be listening on. By default, this interface is 127.0.0.1. We need to change this setting to the local 10.x.x.x interface of the Cacti host.

Modify this entry to reflect your private IP address (your private IP address will be different than mine):

Restart SNMPD

Check to make sure SNMPD is listening on the right IP with netstat –lpn | grep snmpd

We are now ready to create the device in Cacti for monitoring. If you don’t see the snmpd process listening on your correct private IP – check the configuration in /etc/default/snmpd again. If you don’t get anything back from this command then confirm that snmpd is started and running properly with /etc/init.d/snmpd status

Create “Device” in Cacti console and confirm SNMP connectivity to Ubuntu server

Our next objective requires us to log into the Cacti graphical interface as admin. In the last Cacti blog post – I had you deploy the server and do the initial log in which will require you to update the admin login password to a password of your own. I use my web-browser and point to the public IP address of the Cacti server via http://<public_IP>/cacti/

Once you have successfully logged in – go to the “Create devices” on the main page. Refer to image below.

On the “devices” page click on the “Add” button.

Here is where we will set the Description, Hostname, Host Template, SNMP version, SNMP Community, and Notes. Once this information is filled out appropriately, click on the “Create” button.

Create Graphs – CPU usage, Load Average, Memory Usage, PING Latency, Processes, Eth0 Traffic, Eth1 Traffic

If the device was created successfully, the SNMP information will be displayed for the device near the top of the device page. If you get an SNMP Error message there, then it is possible your private networking on one or both of the hosts is not configured properly or SNMP was not configured properly on the Ubuntu server.

Next, click on “Create Graphs for this Host”

This will take you to a new page where you can select the graphs available from the template used for this device. Remember, the template we set was the ucd/net SNMP host template. Put a check box next to the items you want to graph (CPU usage, Load Average, Memory Usage, PING Latency, Processes, Eth0 Traffic, Eth1 Traffic) and click “Create”.

**Note – it will take 5 minutes for the Cacti poller to start gathering data and populating into the rrd file This is the default /etc/crontab setting for poller execution. (An rrd file is essentially the data base the Cacti poller creates and uses to house data gathered from the device. )

After a few minutes Cacti will have built out the rrd file for this device and gathered initial data to be viewed in graphs. To get there go to “Graphs” in the upper left and then to “Lists” on the upper right side.

From here we select the host “Gluster2” and filter for the graphs available for this device. See image below.

Click on each graph and you should see data being graphed (If you receive a “broken image” wait for a few minutes as Cacti may not have enough data to present a graph yet).

Repeat process for other Ubuntu servers in your network.

Repeat this process on other Ubuntu servers on your network. Shortly, I will follow up with a blog post on how to monitor your Windows server via Cacti! Remember to look for my Community GoGrid Server Image (CGSI) within GoGrid to get you up and running with Cacti in a matter of minutes!

Questions? Leave a comment!

A couple of weeks ago we released a new version of GoGrid which included a variety of exciting enhancements and features. Most notable are:

• GoGrid Dedicated Servers
• List View of GoGrid Objects
• Edit F5 Load Balancers via the API
• New Login Page
• Self Service Support Links

More details about these features can be found on the following GoGrid blog post. For a quick overview of the new features in the form of a screencast, please watch the video below. I quickly cover many of the items listed above including:

• Walking through the new login page
• Deploying a GoGrid Dedicated Server
• Overview of the new GoGrid List View
• Dedicated Servers showing in the Jobs Tab
• Showing of Self-Service Links in the portal
• Editing Load Balancers using the GoGrid API

If you have more questions about these new GoGrid features, I recommend that you attend our webinar (on February 24, 2010) where you can get more information about this important release. Be sure to register for the webinar and come prepared with questions!

More news raining from the Cloud:

• Microsoft Azure, Server teams form new cloud division
  “In a much anticipated move, Microsoft announced the combination of the Windows Azure group with the Windows Server and Solutions group into a new organization, titled the Server and Cloud Division. The new division, headed by Senior Vice President Amitabh Srivastava, will be a part of the Servers and Tools Business, headed by Bob Muglia.” (source: cnet – the Wisdom of Clouds)
  • Commentary: As always, James Urquhart breaks the news on his great blog, the Wisdom of Clouds. While it seems that this announcement has been some time in the making, it is good to see it materialize. It only makes sense to me that Microsoft blend the Server/Cloud environments. It’s important that synergies exist from the get-go since many of the features can be leveraged cross-departmentally and should be views as a “single solution.” I’ve mentioned this numerous times before but this is why we at GoGrid developed Cloud Connect which is a Hybrid Hosting solution of physical and cloud servers all connected via private dedicated physical connections. Buried in James’ article is the mention that CTO Ray Ozzie is no longer in charge of the Azure team which signifies Microsoft’s move toward making Azure an important part of their business offerings.
• Microsoft, Cisco, IBM and Others Form Cloud Computing Group
  “A group of companies is starting up an Enterprise Cloud Buyers Council in hopes of removing barriers to enterprise use of hosted cloud computing. Initial members include companies that offer hosted cloud computing as well as enterprises that use such services, including Microsoft, IBM, HP, Cisco, AT&T, BT, EMC, Deutsche Bank, Alcatel-Lucent, Amdocs, CA, Nokia Siemens Networks, Telecom Italia and Telstra. Two industry organizations, Distributed Management Task Force and the IT Service Management Forum, are also involved. The TM Forum, an industry association that helps information and communications companies create profitable services, came up with the idea of the council.” (source: PCWorld)
  • Commentary: It’s great to see so many big names joining together to form a “Council.” My only fear is that egos within each of the organizations will come into play as each company tries to promote their own agenda. I can only hope that this (and other similar organizations that have been set up) will actually do something constructive with their efforts. Working on avoiding “vendor lock-in” and “standards-based solutions” is great and all, but these are overly hyped buzz-words that have been discussed throughout 2009 (and back some). Perhaps by throwing some big names behind a council will help to mitigate the fear that so many enterprises have about the cloud (e.g., security, reliability, compliance, standards, etc.) Time will tell if this new organization will produce any effects or results whatsoever. I’m a bit skeptical myself.
• Top 5 cloud computing predictions for 2010
  “Cloud computing standards and major cloud computing outages top Linthicum’s list. Evidently it’s a requirement that all of those in the cloud computing world must chime in with their cloud computing predictions for 2010, so here are mine…” (source: InfoWorld)
  • Commentary: Oh boy, I’m already behind on my predictions. Guess I need to start on that really soon! David’s Linthicum’s top 5 is a pretty sensical list actually, although numbers 4 and 5 might indirectly go hand-in-hand. Hint, if you are thinking about a successful new-business to start, do a cloud computing startup since you will be snapped up by a large player once you are mildly successful. Well, that is a dramatic oversimplification of #’s 4 and 5. Read through his list though as they are distinct and good thoughts.

Okay, let’s face it. I know at least one Enterprise organization that has migrated their traditional infrastructure to one based on Cloud Computing. That would be The North Pole. Their CEO is none other than Santa Claus (Saint Nicolas, Kris Kringle & Father Christmas are some of the other aliases he goes by). He runs an incredibly efficient organization with Mrs. Claus as CFO and the Head Elf as CIO. They have a very predictable business cycle, which includes global, year-round information gathering and management that catalogs the “naughty” and “nice” people world-wide coupled with a robust transportation and distribution system active only one day annually. How is this all accomplished so efficiently year after year?

I recently spent some time with Santa and had the opportunity to ask him a few questions about his organization. Below is a transcript of our discussion:

Me: Hello Mr. Claus.

Santa: Ho ho, please call me Kris.

Me: Ok Kris. Thanks for spending some time with me today and granting me some insight into your traditionally “closed” organization.

Santa: But of course. Transparency seems to be quite important nowadays. Obviously I can’t disclose all of my secrets; that would take away my magic!

Me: Ok, I understand. So let’s dig into this. For me, and many other technologists out there, your IT infrastructure must be amazing! Can you give me some insight into how you accomplish what you do?

Santa: Sure. So let’s take a look first at how things were done previously. Long before computers and servers were mainstream within our organization, we relied heavily on word-of-mouth and traditional delivery methods. It was a very decentralized environment. For example, children around the world wrote letters that were delivered via “snail-mail” to our corporate headquarters. Letters were read, manually sorted, categorized and audited against the NoN (Naughty or Nice) list. During the holiday season, we were forced to hire many temp workers and managers to oversee the process. Prior to that, we had a distributed workforce around the world essentially spying on children to ensure their compliance and meeting our strict NoN requirements. As the data was manually tabulated, it was physically transported back to Corporate. The holiday season represented a manual merging of the two data sets: letters from children and the NoN audits.

Me: Well, that certainly sounds complex, time intensive and very inefficient! What made you decide to re-engineer your business processes?

Santa: Yes, it WAS very intensive. My beard and hair aren’t just white from old age alone! Ho ho ho. Well, as the world population grew and became a bit more complicated, we realized that we could not scale efficiently using manual methods. It simply wasn’t working. Believe it or not, we did have several mid-year strikes and walk-outs by the elves. It was not a pretty scene. Luckily we employed an incentive-based strategy and coupled that with a complete implementation of an IT infrastructure. I quickly became the unofficial CTO of the organization. Here’s a little secret that I will probably get in trouble for saying: You know all of those children who asked for computers for Christmas and never got them? Well, they were produced by us but we opted to exercise eminent domain and requisitioned them for our own infrastructure. Those children that were affected all received the second item on their list, however.

Me: So, you essentially set up a large server farm using those bare-metal servers. What did that do for you?

Santa: That’s right. It was a difficult decision to make, especially with how we undertook the acquisition process. Ultimately though, it was a wise choice. The next challenge that we had was in the conversion from a manual process to one that was much more automated. We considered outsourcing for this, but the concerns around security and privacy made us keep all of our hosting in-house. We changed many of our manual processes to incorporate much more automated and integrated ones. We opted to keep our remote workforce for the NoN list as our field offices needed to have first-hand insight into how children were acting. The transmission method was changed as we started using a CRM solution to manage this, so, no more hard-copies of field reports. All updates were handled through our CRM software.

Me: Interesting. If I may ask, who did you choose for your CRM?

Santa: Sorry, I’m under a non-disclosure agreement with that. We need to ensure that all of our operations remain as concealed as possible. But I can tell you that it was web-based which enabled our field-offices and remote branches to be much more mobile. In fact, our accuracy and efficiency showed dramatic increases and our operating expenses, specific to our shipping and receiving, dropped. We were still, however, receiving physical letters that needed to be reviewed. For this, we opted for some OCR and cataloging software to increase efficiency. Thus, when it came to the actual data crunching at the end of the year, we were better able to link the NoN database with the Gift Requests database, which resulted in a huge time and cost savings and other unforeseen efficiencies on our part.

Me: Wow, that sounds like a great success story!

Santa: Yes, but it goes on. With success comes failure as well. We found a number of issues after a few years running “digitally.” For one, we just couldn’t scale quickly enough. Also, our energy consumption was much higher than anticipated. We noticed that much of the polar ice cap around our datacenter was starting to melt and that the average temperature around the facility had risen a few degrees which was definitely concerning. We never stopped adding servers and it wasn’t fair to keep acquiring all of those servers from tech-inspired children. Our CapEx was really starting to grow almost exponentially. We only utilized our equipment to capacity toward the end of the year and then, after that, our utilization dropped dramatically. This wasn’t ideal. AND, Mrs. Claus was saying that our margins were shrinking dramatically. We obviously didn’t want to reduce the quality of the toys we produced so a year ago, we undertook yet another strategic technology change.

Me: Oh? And what was that?

Santa: We started migrating our business over to Cloud Computing and what a difference that made! First, we were lucky enough to be allowed in on a private beta of a Cloud Computing Infrastructure provider. What an amazing technology! Not only were we able to deploy a complete, load-balanced scalable infrastructure, but we were also able to control our costs dramatically. I even worked with Rudolf to develop an interface to the Cloud provider’s API. This interface monitored the influx of NoN reports, the number of days until Christmas and Gift Request volume and it programmatically provisioned new servers ahead of the anticipated demand. It even automatically scales back after the holiday season.

Me: Sounds simply like the ideal solution. Can you tell me more about it and who they are?

Santa: Again, I’m under an NDA, as well as part of another private partner program which gives me some enhanced functionality. But let me tell you this, they offer Windows Server 2003/2008, Linux, free Load Balancing, Cloud Storage, a web GUI, and free support and inbound bandwidth. The free inbound bandwidth is particularly nice since we have our entire field offices transmitting data regularly inbound to corporate.

Me: Hmm. That provider sounds REALLY familiar!

Santa: Ho ho ho. Let me tell you. Just being able to dynamically scale and only be billed for what I use is fantastic, and their Cloud Storage is perfect for all of those scanned letters. When it comes to the December analysis, we just spin up a series of Database servers and do the data crunching. It used to take us weeks to do the same amount of work which we can now do in a mater of days. Not only that, our carbon footprint has been dramatically reduced. In this particular instance, outsourcing made the most sense to us. Many Clouds are built using some sort of virtualization or hypervisor technology which means that one larger server takes the place of multiple older ones. Power consumption is reduced (we used to have hundreds of unused servers sitting around doing nothing during the off season). Our actual RAM/CPU/Drive utilization on those older bare-metal servers was around 10-15%. Now we have amazing efficiencies using the Cloud. I must say, Mrs. Claus is very happy…she even gives me extra cookies…her cookie budget has increased due to the reduction of our CapEx.

Me: Well it seems like you have come a long way. Congratulations! What do you plan on doing with your new free time and money savings?

Santa: I plan on buying a nice little bungalow on a very sunny island and take Mrs. Claus there for some rest and relaxation. Oh, but it must have internet access…the nice thing about my Cloud Computing solution is that I can control my infrastructure from anywhere via a web browser.

Me: Wow, that’s great. Lucky her! Any last remarks you want to make?

Santa: Just one, all of you non-Cloud-enabled companies and Enterprises out there, I’m setting a “naughty” flag next to your entry. You have no excuse! Take it from me, the CEO of the most efficient and on-time Enterprise on the planet. The Cloud is the way to go!

---

So there you have it, straight from the mouth of the jolly, white-bearded CEO of The North Pole Corp. Need I say more? Happy Holidays from us at GoGrid!