In the first post in this article we looked at how you could get high levels of hardware fault tolerance with a “statefull” application such as a database and how it was very expensive. In this post we will look at how the C Infinity private cloud can get the same results at a fraction of the cost and with greater functionality.
The C Infinity private cloud offers a cheaper way to get these higher levels of fault tolerance. We have a running virtualisation platform that is clustered.
Let’s take a customer who wants to run a highly available SQL server as an example. With physical servers we were looking at the “statefull” failover cluster architecture that would have to be implemented just for that customer at a great expense that would be passed on to the customer.However, C Infinity’s private cloud is a pre-existing cluster of N+1 virtualisation host servers. That’s a mouthful so here’s what that means:
- We run a farm of virtualisation host servers based on Microsoft’s Hyper-V. That features great performance of virtual machines and storage. They are all HP Blade servers with 4 hour response time contracts with HP.
- The storage we use is the highly performing, scalable and fault tolerant HP EVA SAN (storage area network). All servers have 2 paths over our fibre channel network to the disks that use RAID.
- Those virtualisation hosts can each run many virtual machines. The customer only pays for what they consume instead of an entire cluster.
- We keep 1 host server free. That means we can lose a physical host server and still be 100% operational. All virtual machines can stay running at 100% capacity.
- We have no “bursting”, “oversubscription”, etc. If you pay for 8GB of RAM then your get all of that 8GB of RAM, 100% of the time.
- The virtualisation servers are clustered. Thanks to Hyper-V and “Live Migration” that means we can move virtual machines between the host servers with no down time! In the case of a hardware failure, sure, the virtual machine will power off … but it powers back up again automatically on another host giving you the same result as with a physical failover cluster implementation at a fraction of the cost!
So this means that a customer who is worried about server fault tolerance and was originally looking at multiple web servers and a database cluster now can look at implementing half the numbers of servers with standard edition operating systems and at a fraction of the over all costs. They also don’t need to worry about the complications of cluster maintenance and upgrades. That all sounds way too good to be true but it is up and running now.
For example, we recently got an alert from our Microsoft System Center Operations Manager monitoring system. One of our host servers in the cluster had detected that a memory board was degrading. We called HP and their best enterprise partner arranged to come out to replace the memory board in less than a couple of hours (I love that!). Let’s have a look at the scenarios:
- Standalone server: the service would have to go completely offline for that hosting customer and affect all of their customers. There goes the 100% SLA! Memory board replacements, CPU replacements, storage backplanes, SCSI controllers all bring that realistic single server SLA down to 98% no matter what any sales man tells you.
- Physical Cluster: Services would be moved around. There would be anything from a few seconds to a few minutes of an outage while services stop and restart. That still seems like a lot of downtime for a huge spend.
- VPS: Every customer hosted on the physical server would have downtime while the board is replaced. Oops!
Ouch! And we only detected this issue because we have an all-encompassing, intelligent monitoring solution, not one of those cheap-ware “we’ll monitor X number things for you” solutions based on SNMP. Here’s how the process went for us:
- We moved the virtual machines from host A to host B. There was no down time.
- We powered down host A so the engineer could remove it from the chassis and perform the memory board replacement.
- Remember, all of the virtual machines stayed running.
- The engineer powered up host A.
- We checked it out and tested it.
- We moved the virtual machines back to host A.
So nobody had any down time while we physically powered down a host and replaced a memory board. I’ve since chatted with a few of the “affected” customers and told them about this. They hadn’t a clue.
Let’s consider another situation. What if you need to double your storage or your server’s RAM?
- Standalone server: There’s a good chance you have already filled the disk or RAM slots in the server. An upgrade means changing servers at big expense and migrating between servers where there will be an outage. And buying in equipment does take anything from a day to a week.
- Cluster: Sorry, all servers should be identical. That means you need to double your spend on the RAM! You also face a similar situation with disk slots in your storage tray.
- VPS: It’s a single server, probably using cheap 5K disks (not the industry standard speed of 15K) and the slots are possible full. Expanding your allocation of storage possibly means powering down your VPS machine and moving your GB’s of storage across the network to another host, assuming the hosting company has one available. There is potentially a lot of downtime there.
- C Infinity Private Cloud: We move a couple of slider controls. Yes, that’s it. We keep spare capacity in our SAN so we always have disk available. If we need to give you more RAM than is available on your current host we use Live Migration to move you to another host with zero down time. Give it 15 minutes or so and your spare capacity is added with little more than a phone call from you.
There’s other things that the C Infinity private cloud offer. For example:
- Need an identical copy of your machine? We can do that in a short period of time.
- Going to do a major upgrade of your application or operating system? We can snapshot your machine. If the upgrade is satisfactory you merge the snapshot and carry on as normal If it isn’t satisfactory we can roll back to the snapshot and carry on as before, totally unaffected by the failed upgrade.
So that’s the C Infinity private cloud for you. You’re getting a lot of bang for your buck and a lot of flexibility. We believe in the concepts of Dynamic IT and we make it an achievable goal for all of our customers. We believe in getting high uptimes and we have succeeded in achieving that for all of our customers.
Please give us a call if you would like to achieve a highly available, secure, reliable, managed and flexible server hosting service.
Related posts:
