Listen to the podcast. Find it on iTunes. Read a full transcript or download a copy. Sponsor: VMware.
Rapid adoption of server virtualization that enabled desktop virtualization at a large Israeli insurance and financial services group then lead to rapidly modernized IT infrastructure that has in turn spawned cloud and mobile computing benefits.
Clal Insurance Enterprises Holdings, based in Tel Aviv, both satisfied current requirements and built a better long-term enterprise architecture from this path, clearly illustrating the multiplier effect of value and capabilities from such IT transformation efforts.
In the latest BriefingsDirect podcast discussion, learn how Clal’s internal IT organization, Clalbit Systems, translated that IT innovation and productivity into significant and measurable business benefits for its thousands of users and customers.
Haim Inger, the Chief Technology Officer and Head of Infrastructure Operations and Technologies at Clalbit Systems, discusses the journey with BriefingsDirect's Dana Gardner, Principal Analyst at Interarbor Solutions. [Disclosure: VMware is a sponsor of BriefingsDirect podcasts.]
Here are some excerpts:
Gardner: One of the things that’s interesting to me is the speed and depth of how your organization has embraced virtualization. You went to nearly 100 percent server virtualization across mission-critical applications in just a few short years. Why did you need to break the old way of doing things and why did you move so quickly to virtualization?
Inger: The answer is quite simple. When I got the job at Clal Insurance four years ago, everything was physical. We had about 700 servers, and to deploy a new server took us about two months. The old way of doing things couldn't hold on for much longer.
Regulations in the new businesses that we needed to implement required us to do such things as deploy servers as quickly as possible and simplify the entire process for demanding a new server -- to deploying it and giving it the full disaster recovery (DR) solution that the regulations require.
The physical way of doing things just couldn’t supply the answer for those requirements, so we started to look for other solutions. We tested the well-known virtualization solutions that were available, Microsoft and VMware, and after a very short proof of concept (POC), we decided to go with VMware in a very specific way.
We didn’t want to go only on the development side, the laboratory side, and so on. We saw VMware as a solution for our co-applications and for a long-term solution, not just for islands of simple virtual servers, where we decided from day one to start using VMware on SQL servers, the Oracle servers, and SAP servers.
Full speed ahead
If they held on there very well, then we could, of course, also virtualize the simpler servers. It took us about four months to virtualize those initial servers, and those were very simple. We just pushed the project ahead full speed and virtualized our entire data center.
Gardner: As you’ve gone about this journey, why does it seem to be paying off both on the short-term and setting you up for longer-term benefits?
Inger: That’s very simple to answer. Today, to provision a new server for my customer takes about 20 minutes. As I said, in the past, in the physical world, it took about two months.
DR was the main reason for going into this project. During a DR test in the old days, we had to shut down our production site, start up all servers on the DR site, and hope that everything worked fine. Whatever didn’t work fine, we tested one year after that initial DR was done.
Using VMware with Site Recovery Manager (SRM), I can do an entire DR test without any disruption to the organization, and I do it every three months. Watching our current DR status, if anything needs to be fixed, it’s fixed immediately. I don’t have to wait an entire year to do another test.
So those simple things are enabling us to give our organization the servers that they need, when they need them, and to do the regression in a much simpler way than we did in the past.
Gardner: Tell us a bit more about Clal.
Inger: Clal is a group that contains a very big insurance company and another company that is doing trading on the Israeli international stock market. We have a pension company and insurance for cars, boats, apartments and so on. We even have two facilities running in United States and one in the UK.
We're about 5,000 employees, and 7,000 insurance brokers, so that’s about 12,000 people using our datacenter. We have about 200 different applications serving those people, those customers of ours, running on about 1,300 servers.
Gardner: That's obviously a very large undertaking. How do you manage that? Is there a certain way that you’ve moved from physical to virtual, but have been able to manage it without what some people refer to as server sprawl.
Inger: I know exactly what you mean about over populating the environment with more servers than needed, because it’s very easy to provide a server today, as I said, within one hour.
The way we manage that is by using VMware Chargeback. We've implemented this module and we have full visibility of the usage of a server. If someone who requested a server is not using it over a period of three months, we’ll know about it. We’ll contact them, and if they don’t require that server, we’ll just take it back, and the resources of that server will be available once again for us.
That way, we're not providing servers as easy as could be. We're taking back servers that are not used or can even be consolidated into one single server. For example, if someone requested five web servers based on Microsoft IAS and we’re sure that it can be consolidated into just one server because CPU utilization is very low, we’ll take it back.
If an application guy requires that the server have eight virtual CPUs, and we judge it's use on peak time is only two, we’ll take six virtual CPUs back. So the process is managed very closely in order not to give away servers, or even power, to existing servers that are not really needed.
Gardner: Tell me how you’ve been able to develop what sounds like a private cloud. Do you consider what you’ve done a private cloud, or is that something you’re looking to put in?
Inger: We do consider what we've done a private cloud. We're actually looking into ways of going into a hybrid cloud and pushing some of our systems to the public cloud in order to control the hybrid one. But, as I said, we do consider the work we've been doing in the past three years as fully partnering a private cloud.
Gardner: Have there been any hardware benefits when moving to a private cloud, perhaps using x86 hardware and blades? How has that impacted your costs, and have you moved entirely to standardized hardware?
Inger: Of course. When we saw that those 20 servers that we initially did in late 2008 and everything worked okay, we decided to do standards. In one of the standards that was decided upon was if it doesn’t work on VMware it won’t get on our data center. So a lot of applications that run on Itanium microprocessors were migrated into Linux and on top of VMware running on x86,
We managed to save a lot of money, both in supporting those legacy systems and developing in those legacy systems. They’re all grown. Everything that we have is virtual, 100 percent of the data center. Everything is run on x86 blades, running Windows 2008 or in Linux.
All these systems we have used to run on a mainframe. It’s Micro Focus COBOL running on top of Red Hat Linux latest version, on top of VMware, and x86 blade.
Gardner: Let’s take the discussion more towards the desktop, the virtualization experience you’ve had with servers and supporting such workloads as SQL Server, Oracle, and SAP. This has given you a set of skills and confidence in virtualization that you’ve now taken out, using VMware View, to the desktop. Perhaps you could tell us how far you’ve gone in the virtual desktop infrastructure (VDI) direction?
Inger: After finishing the private cloud in our two data centers, the next step within that cloud was desktop. We looked at was how to minimize the amount of trouble we get from using our desktops -- back then it was with Windows XP Desktop -- and how to enable mobility of users, giving them the full desktop experience, whether they’re connecting from their own desktop in the workplace or if they’re using an iPad device, connecting from home, or visiting an insurance broker outside of our offices.
We looked at the couple of technologies that would fit in VMware View. Again, after a short POC, we decided to go ahead with the VMware View. We started the project in January 2012 and right now w're running 600 users. All of them are using VMware View 4.6 which is being upgraded, as we speak, to version of 5.1.
It enables us to give those users an immediate upgrade to a Windows 7 experience, by just installing VMware View, instead of having to upgrade each station of those viewers, and without going to those 600 users who are on Windows 7 right now.
And we're delivering it on every device that they're working on. If they’re at work, at home, outside of their office, their devices, iPad as we said earlier, are getting the same experience. The plan is that by the end of next year, all of our employees could be working on VMware View.
The ability to give the user the same experience on each device that he works on is sometimes priceless. When I fly from Israel to the United States and have a wi-fi connection in the plane, I can use an iPad and then work on my office application as if I were in the office. Otherwise, if it’s a 12 hour flight, I'd be 12 hours out of work.
If you take into account the entire ecosystem that you’ve built surrounding VMware View, it’s actually priceless, but it’s very hard to quote exactly how many dollars it save us on a daily basis.
By end of the year 2012, our plan and budget was for 1,000 users. So we're on the way to meet our goal in December this year. For next year, 2013, our goal is to add 2,000. So it will cover almost the entire organization. It leaves something like 500 power users. I’m not sure that VMware View is the best solution for them yet. That will be tested in 2014.
Of course VDI as a stepping stone is an essential element in implementing a bring your own device (BYOD) policy. That’s something we're doing. We're in the initial steps of this policy mainly with iPad devices, which a lot of employees are bringing to work and would like to bring when they're on site, offsite, or at home. Without VDI, it would be impossible to give them a solution. We have tons of iPads today that are connecting to the office via VDI with a full Window Server experience.
Gardner: I'd like to get your thinking around virtuous adoption. As we started talking about DR, your full virtualization of your server workloads, your being able to go to standardized operating systems and hardware, moving to VDI, then moving to hybrid cloud and also now mobile, it truly sound as if there is a clear relationship between what you’ve done over the years with virtualization and this larger architectural payoff. Maybe you could help me better understand why the whole is perhaps greater than the sum of the parts.
Inger: The whole is greater than sum of the parts, because when I chose VMware as a partner combined with EMC on the storage side and their professional services, I had actually done a lot of the work together with my people.
Life gets easier
Life gets easier managing IT as an infrastructure, when you choose all those parts together. An application guy could come to you and say, "I didn’t calculate the workload correctly on the application that's going to be launched tomorrow, and instead of 2 front-end servers I need 15."
Some other person could come to me and say, "I have now five people working offshore, outside of the Israel and I need them to help me with a development task that is urgent. I need to give them access to our development site. What can you do to help me?"
I tell him, "Let’s put in our VDI environment, and they can start working five minutes from now." When you put all of those things together, you actually build an ecosystem that is easier to manage, easier to deploy, and everything is managed from a central view.
I know how many servers I have. I know the power consumption of those servers. I know about CPU’s memory, disk I/O and so on. And it even affects the decision-making process of how much more power I'll need on the server side, how many disks I'll need to buy for the upcoming project that I have. It’s much easier decision making process. Back in the physical day, when each server had its own memory, its own CPU, and its own disk, there was much more guessing than deciding upon facts.
Listen to the podcast. Find it on iTunes. Read a full transcript or download a copy. Sponsor: VMware.
You may also be interested in: