Last Updated on April 27, 2014 by Ruchir Shastri
It is very difficult to fall in the trap of signing up for cheapest dedicated server through tempting offers that are flowing all over the place. We always think of getting the lowest cost offers to cut our costs – however, it is very important to understand the ‘gotchas’ of the so called cheap servers. I’ve seen many customers complaining online that they are losing thousands and hundreds of thousands of dollars due to their server being down with a provider from whom they have obtained the server at a very cheapish rate. I’m not sure how someone can trust placing their high revenue business website on $100 non-redundant server?
I’ve compiled a check list that will be helpful to you in selecting correct dedicated server for your business / personal requirement:
Ensure that your provider is using enterprise class server components
Most low cost providers will be using desktop class server components (RAM / HDD / Processor) that is not made to run on 24×7 server environment. They will fail more often than enterprise class hardware which is prepared to run on 24×7 high intensity environment.
In case of RAM, you need to ensure that your provider is using ECC RAM which can auto-detect and auto-correct most of the internal data corruption which is not done by desktop class non-ECC RAM.
In case of HDD, it is important that your provider is using hard drive that is marked for server environment. Some of the popular brands of enterprise class SATA HDD’s are Western Digital RE4, Seagate Constellation, Toshiba Enterprise Class, etc. If you opt for SSD, I will recommend considering the same for it. Samsung 840 Pro, Intel 520, Intel DC S3500 are some of the popular enterprise class solid state drives available in market currently.
Processors are the key to operation of dedicated server and it is highly important that you don’t fall for any low cost desktop class processor. For eg, all ‘Intel i’ series of processors are meant for desktop use and not for servers. Server processors are marked ‘Intel Xeon’ or ‘AMD Opteron’.
It is always advisable to check on redundancy available with your provider at network level since this is going to be the pipe through which data / website will flow from your servers. You will need to ensure that your prospective provider has multiple connectivity backbones from some of the popular transit providers in US like Level3, NTT, nLayer, AboveNet, AT&T, etc. Having multiple providers will also ensure that your visitors will be routed through the shortest path between your server and their place of access from available different routes from multiple providers. It is not a good idea to rely on just one or two providers since they can go down randomly or have some sort of maintenance which can take your server / website / application out of production.
In addition to network provider, you should also ensure that your provider uses top end networking hardware (routers, switches, firewalls, etc.) which is meant for enterprise class usage. Some of the popular providers of networking equipment are Cisco, Juniper, Brocade/Foundry.
This is one area where most providers fall as they are mostly manned during ‘Business Hours’. Hence, if your server goes down during night time, it will be difficult to get hold of anyone to look into it and get it back in working state. Providers require to pay more for techs working overnight and hence they are either not staffed during night time or they have level 1 tech available who is not permitted to enter data center premises. It is very necessary to ensure that your provider has 24×7 knowledgeable staff available who can fix any kind of issue for you at any time of day.
Choosing Correct Server
Once you have identified a provider for your requirement, it will be down to selecting appropriate hardware for your server. In today’s world, processors are very powerful and most generally you will not experience slowness in terms of processing power – however, it is important that you get a powerful processor that is more than enough for your processing requirements.
When it comes to choosing correct hard drive for your server, most clients end up just looking at the amount of hard drive space (500GB, 1TB, 2TB, 4TB, etc.). However, you will also need to consider the type of HDD used. The most commonly used is SATA HDD. If you need more reliable hard drive, you can consider SAS drives. The fastest of all are Solid State Drives which do not have any sort of mechanical device / platters as found in traditional SATA / SAS drives. Hence, they are very fast in operation and provides you with best IOPs. However, as you may have guessed, they are most costly and are available in smaller size compared to SATA / SAS drives. Hence, if you have an application that is performing lot of random writes (database / transactional), it is recommend to consider SSD’s for best disk I/O results. In case you have sequential writes/reads (like streaming, etc.), you will not be benefited much by investing more in SSD’s and traditional SATA / SAS drives should be just fine for your requirement.
Once you have chosen correct hard drive and have also come up with amount of disk space that you will need, it is important to take into consideration redundancy at hard drive level. It is a well known fact that hard drives do fail at some point in time and it is necessary to plan for it right at the time of deployment. For redundancy at hard drive level, you will need to consider one of the RAID solution – hardware or software. A very basic is RAID 1 solution whereby you will need to have minimum two hard drives whereby one of the drives will be used for mirroring. Hence, you can afford to lose one of your drive and still have your server up and working from second drive. Once detected, you can replace failed hard disk and allow the RAID array to rebuild to be back at full redundancy level. It is utmost important to have some kind of RAID implemented for your server.
After the recent issue with Sandy Storm which knocked out power from majority of the providers on East Coast, customers have become more aware / concerned about redundancy at power level. Hence, it is important to ensure that your prospective host is having appropriate power backup options to fight against utility power outage and/or any other natural calamities. Ideally, data centers should have UPS battery backup (of around 5-15 minutes) to allow all equipment to operate on it by the time it switch overs to generator power backup. It is also important that there is redundancy at generator level along with appropriate amount of fuel on site (most generally providers keep fuel that allows to run the generators for around 12-24 hours.
It is utmost necessary to ensure that appropriate temperature is maintained in data center in order to allow servers to function appropriately. The data center should also be designed in a way that there is hot aisle / cold aisle to allow maintaining sufficient cooling in data center. There should not be any “hot spots” in data center else it can result in severely degraded server performance along with blowing up some of the important server components. Most data centers maintain N+1 level redundancy at cooling level as well.
I hope the above points will help you in choosing a correct provider for your business to ensure that your server / website operates smoothly without any downtime and minimal failures.
We, at Softsys Hosting, have carefully chosen our data centers in Denver & Chicago where we colocate all of our servers (along with network switches, power strips, etc.) so as to maintain high level of services which allows us to offer unmatched uptime and performance from our servers. We do not compromise on hosting quality and this is the only reason why we have had customers staying with us for years together. We are committed to provide excellent services at all times to our valued customer.
President – SoftSys Hosting