HomeDatacenterNew Datacenter! Move date is 21st. Advantages of performance data
New Datacenter! Move date is 21st. Advantages of performance data
March 13, 2016
Pulsed Media March Newsletter
Oh do we have big news this month! We’ve been hard at work to make new datacenter a reality and to increase performance!
Woo! We are moving to new datacenter this month. The new location has more than 3 times the floor space, couple times the electrical power, many times the cooling capacity, raised floor, high end rittal rack cabinets etc. and in a better location!
The new location is very good; most major network operators are present in this building, it is a nexus for networking and is one of the most important on-net locations in Finland. Not only all of these advantages, but this is also much more economical place. This is a good place for a growing company! 🙂
We have been working really hard for the past month to make the move in happen this month so we can start leveraging the advantages as early as possible!
After the move we finally have the cooling capacity required to grow our operations. Especially sweet is the economizer systems for cooling, we have direct access to outside air for cooling needs, this combined with the raised floor gives us opportunity to use outside air to economize cooling for most of the year, and only run the CRAC units during the hot period of the year, likely only a few hours each day! Finland is a cool place to live in, and we finally get to take advantage of this! Helsinki July average temps are only at around 22C – So even during the hot months of July and August we can mostly operate using economizers solely, for large portion of the night at least.
We have also focused on power usage metrics quite a lot, and we can monitor each rack power utilization for both redundant power feed and non-redundant direct power network access, as well as the total consumption. This gives us means to calculate PUE precisely and to make informed decisions when it comes to cooling. Each rack gets a total of 64A of power feeds for total usable of 51AMPS or 11 730W for 24/7 utilization.
All of this information and cost efficiency will allow us to dedicate a larger portion to hardware investments, and eventually pass down the benefits to You! 🙂
Datacenter Move: 21st of March
Current estimate move date is 21st of March, around 10:00 to 24:00 GMT. Service will be interrupted as we are moving servers from old datacenter to new datacenter.
This will cause a large portion of servers to be down during this period, also provisioning will be temporarily disabled for this period. Some servers will experience only a short 60 minute downtime, while others might experience as long as 12-16hours.
Unaffected services are: MDS Series, Almost all dedicated servers. Affected services are: SSD Seedboxes (all), Dragon/2012 (most), Super50/Super50+ (some), Super250 Series (some), Value Series (Most), Storage Seedboxes (some)
You can always ask from support if your server is affected. We will try to do our best to answer all support requests during this period, but we will be hard at work to bring servers back online as fast as possible.
Since we are physically moving the servers, there is potential for disk failures and other hardware failures. Every time a server’s electricity is cut and it needs to be started from cold there is a real danger of hardware faults. So we highly recommend that you backup everything important right now. Sorting through the disk failures may take quite a bit of time after the move period. We have reserved hard drives and complete servers w/ spare parts for replacements just in case. This is only precautionary measure.
If your service is still down 22nd of March ~06:00 GMT, please open a ticket and we will let you know what is going on and estimated service recovery time.
SLA Compensations: SLA Compensations are done for prolonged downtime lasting more than 24 hours. We will be proactive about compensations.
We started gathering and analyzing precise performance data from servers sometime ago, and the benefits have been huge.
We have noticed several times that performance degrades just prior to disk failure, hence we can migrate all data and users before disk failure in those cases. We have identified servers which simply will not perform, and have upgraded these servers. We have identified misconfigurations of servers as well, where tuning of the subsystems were simply forgotten during server setup, and could fix those as well.
This has also allowed us to enhance our provisioning algorithms based on performance, now a server which has heavy activity gets penalized in provisioning and hence a new user likely gets to a server with less load, as long as there are choices for which server to use for provisioning.
This data has also shown us that we can increase some settings for more aggressive seeding 🙂
Espoo Network Upgrades
We updated our edge switch more than a month ago in Espoo. The end result: Problems. But the kind of problems You want to have!
Our transit usage nearly doubled overnight after the upgrade! We were hitting hard on our transit commit levels for the whole month of February. A large portion of February the bandwidth usage was sitting flat at the maximum capacity!
We have since upgraded by 30% and we are still hitting the limit often. So some work is required on QoS etc. and another transit upgrade for April.
We have a few Try This Out type of specials to offer this month! First month 50% off! SSD, Super250 and Dragon offers are valid until end of the Month. MDS Offer is valid until the 18th.