Posts Tagged ‘server performance’

Iceland Data Centers – Location, Location, Location

Tuesday, February 14th, 2012

Over the last few years the northern countries have been busy promoting ideal conditions for construction of data centers. Most of them pitch the idea of green data centers, powered by renewable energy sources, free cooling, ideal location and all that jazz.

There is a great site called Data Center Map. They provide detailed information on major data server locations. After only a couple of minutes on the site you will begin to see patterns. Data centers are usually built on international fiber optic crossroads, places with better economy conditions, close to power plants and any place where location can provide advantage in one way or another.

Iceland has been “ignored” as a viable data center location for quite some time. Investors have been skeptical because the country is notorious for its seismic activity. Not to mention the fact 3 banks went bankrupt in 2008, leading to a national crisis. Usually, this is all it takes to scare off potential investments. This, however, did not scare off Verne Global from building a 44.5 acre data center in Iceland.

Presently Verne Global’s campus in Keflavik, Iceland, is 100% carbon neutral and draws its commercial power from Iceland’s dual-sourced renewable energy grid of geothermal and hydroelectric power. This is the first major data center in Iceland. Because of its location and direct access to a modern power grid, they can offer over 50% cost savings to their customers. On top of this, Verne Global took advantage of Iceland’s strategic location between Europe and America. Two submarine cables connect Iceland to Europe and one to North America.

 

Verne Global chose Colt for their modular solution, allowing the owner of the data center to choose when to invest in infrastructure and expand, maximizing the utilization of the facilities over time.

So why not build more data centers in Iceland? Cheap renewable energy, redundant power grid, redundant transcontinental network connectivity and zero cost for cooling. What’s not to like?

If you like to check how servers in Iceland perform from your location, we might be able to help.

One popular MMORPG is based in Iceland – Eve Online. On eve-offline.net you can get IP addresses for different servers. Try 87.237.38.200 on port 26000. Our latest test from Seattle shows less than one second response time. In fact, most server monitoring targets from Iceland would return good figures, based on what we have seen. If you run a server in Iceland, go ahead and try how well it performs from Europe and the United States. Don’t forget to let us know what you found out.

DDoS Threat Growing

Thursday, November 3rd, 2011

Apparently the DDoS threat is growing to a point where it is becoming a major concern for data center managers as firewall products’ efficiency is failing.

The security testing organization NSS Labswhich recently discovered that 3 out of 6 firewall devices stop operating when tested for stability. DDoS has been a major threat for network operators for over ten years since their appearance, but recently these attacks have become more aggressive and have increased in frequency and impact.

DDoS is a “distributed denial of service” and is a violation of the policies of all Internet service providers. The way it works is by sending a great load of requests or ‘attacks’ to the targeted computer. These attacks then force the computer to reset itself or to consume its own resources. As a result, the machine is no longer able to provide its intended service and drops the communication with its user. DDoS targets are mainly sites hosted on high-profile servers such as credit card companies or banks.

When DDoS attacks are successful, they lead to significant outages, OPEX (increased operational expenditures), revenue loss and frustrated customers. Unfortunately, the capacity of security products such as firewalls and IPS is limited and the attackers are well aware of it. They can easily exhaust the application layer resources and cause significant downtime.

According to a recent study conducted by Arbor Networks, the volume of DDoS attacks has reached out 100Gbps barrier or, in other words, the DDoS attacks are growing in number and strength.

In order to reduce risk, specialists suggest that large state-exhaustion attacks must be stopped in the ISP/MSSP because this is where the attacks occur. A packet-based detection and protection against all kinds of DDoS is required as well.

Are You Looking at the Right Metrics?

Wednesday, August 17th, 2011

The most dangeours type of downtime is the one you don’t know about. It is discturbing how true that one line actually is. Should it occur, website/server downtime can and will cause problems and ripples throughout your organization. Before we get tarred and feathered for making such a bold statement, let us build our case.

The Problem

In this realworld situation, a business lost roughly 30% of their leads for July. Apart from their initial loss, they simply handed out a good portion of the market to their competitors, in high season. When the figures arrived, all hell broke loose. All major markets felt the downturn. In search for a logical explanation, hours of daytime were invested in finding the reason. After it was made clear that the traffic was stable, the management went on to search for answers somewhere down the line. The marketing team had to pull out detailed reports for their activities in the last three months. Seasonal sales people got numerous tests calls. A full-scale internal audit took place. This caused a ripple effect and the normal workflow was seriously disrupted.

Locating the Issue

Upon request, the IT department emailed external statistics on the webserver’s uptime. They had employed the services of a website monitoring service (not ours). According to their information, the server only went down for 20 minutes that month during the scheduled maintenance. What they failed to notice is that the service they used only gave them figures of the network availability of the hardware device, not the server software. The machine was available nearly all the time, but was doing what it was supposed to (serving web pages) only ~80% of the time. We were able to find that out only after we began tracking the server ourselves.

We were able to locate the problem, because the service we chose to test with, actually tested the website itself. We tried loading all major application forms from multiple locations over a given period of time. It wasn’t long before we got the first alert about a page not loading. It turned out that the server was failing to deliver the pages after a certain number of concurent connections. With some modest server upgrades and clever workarounds by the IT department, all website returned to normal. Simply the hardware couldn’t take the load and the server software decided to drop a number of queries in order to serve the rest.

The website service, employed by the business, worked exactly as it should. What was referred to as website uptime was actually server uptime.

Quick Tips

  • Network availability is only a prerequsite for a website to function properly. Even if the site loads sucessfuly it is not clear if the forms on the site will be 100% functional.
  • One good sign to look for, when trying to find the exact cause for problems with your traffic and conversion rate, are the traffic sources. If you notice significant decrease accross all mediums, then it is most likely that your website is not performing as expected.
  • Make sure you are using the service you need. You can test our range of website and server monitoring services completely free.

SSD vs. HDD for Business

Wednesday, July 6th, 2011

Is SSD the solution for the ever widening gap between current hard drive technology and CPU advancements? Is this the next step in data storage and are SSD drives here to stay? How reliable are they? Are they actually worth it?

This is just a handful of questions from a huge, huge batch. SSD are still pretty expensive for everyday use. They are still the domain of computer enthusiasts and early adopters. Leaving the money question aside, let’s check whether they are a good solution for business workstations and high-end server hardware.

While SSD drives might not be the best choice for personal computing, they might be great for server hardware. In an average laptop, you might be better off with a 5400rpm drive. Most users don’t see any battery life improvement. In fact, the 5400rpm drives can drain the battery even less than a SSD drive. Unlike the SSD drive, HDD can actually spin down and actually reduce the battery drain to a minimum. SSD drives might use more power when running idle, but that is only an issue for personal computers and laptops. Server hard drives rarely stay idle.

SSDs are in fact great for database servers. Most requests are extremely small in size and are often random in nature. With SSD drives there is no mechanical latency to limit the performance. They offer great speed improvements even over 15000rpm drives. The lack of moving parts reduces the heat coming from the drive, thus requiring less power to cool down a server rack.

Many data centers cut costs from cooling. This way they stress their hardware more and need to replace it more often, but it pays off when you look at the power bill. You can have the best of both worlds with SSD. They generate almost no heat at all. Claims of 50% lower electricity bill might not be too farfetched, when you consider the less power required to cool down the server racks.

The life expectancy of a SSD drive is said to extend to 50 years, which is pretty hard to believe and most likely not applicable to servers. It must be somewhat close. Unfortunately, SSDs as we know it have been around only for a couple of years, so no one really knows. The life cycle is limited by the number of write cycles. This is why there are a lot of server applications for SSDs where information is only to be read from them.

Let’s not forget they do the job faster. This means that they complete tasks faster than traditional storage devices. Ideally, this could reduce the amount of disks required in an installation. This is highly unlikely before larger SSDs become available, but it is one of those features that will make a difference once the technology improves.

You get higher performance, high reliability, power savings, more than a reasonable lifespan, and a hefty price tag. Depending on the scale of implementation, the last one might not be true too, considering the lower power bill.

If you plan to upgrade your installation it might be wise to wait for a while. Prices are said to go down by 50% by the end of the year. Early adopters, who have chosen to use SSDs in their web and database servers rarely complain and speed is never the issue.

Raspberry Pi

Thursday, June 30th, 2011

I thought that after the SheevaPlug some time will pass before we see even smaller computers intended for light IT tasks. This time the project is even smaller. It is called the Raspberry Pi. At its current state it looks like this. You might think the product has a long way to go before it hits the market, but you are wrong. One of the ideas of the creator is to bring computing closer to students and inspire them to learn more about the hardware and how it works. That is why it will most likely look like this, when the first units head out to public schools.

At the moment, it is marketed as the 25 USD PC. It has a 700Mhz ARM CPU, 128MB of RAM, HDMI slot and a USB port. You can attach a keyboard, mouse and any other USB device. You can connect a hub. You can see that there is also an image sensor in the middle. It is a 12mp CMOS sensor. You can also insert a memory card in order to have some place to install the OS.

Despite its size, this little gizmo is fully capable to play HD video. For this feature I love and hate the device at the same time. I love the fact that such a small contraption can become part of my home cinema installation, and that is exactly the reason for me to hate it.

This ultra small PC is intended for educational purpose. It is for hobbyists, tinkering with it to power a weekend project, to run some kind of server. It is mostly for people willing to build up on top of it. I will hate seeing this be treated as a cheap HD player.

I can’t wait to get my hands on it. It is cheap enough for anyone to go ahead and play around with, without thinking too much about not being too great with the soldering iron. The CMOS sensor cam can probably have a wide variety of implementations – light detector, face recognition, webcam. By being that small, the Raspberry Pi enables enthusiasts to try out many new projects.

I will definitely try to run something light, such as a telnet server, or a webcam, with the cam being the server. The latter’s retail price is at about $100+. Here is a thought – if the device keeps the promised price, it will cost less than my personal hosting plan. I can move everything there and just pay for the IP address I’m already using.

How would you use or mod this device?