Welcome to our server buying guide! We’ve listed a variety of items for you to learn, review, and ensure your investment matches your company’s needs. Let’s get started!
Why do servers have a high range in price?
Server cost estimations have a wide range because servers are generally custom built to order. Pricing varies based on brand, specs, licenses, and warranties. Setup time also varies based on equipment quantity, complexity, age (new, used, refurbished) and functionality.
What is a good range for a new server (standalone hardware only budget)?
We’ve established 5 levels of stand alone server purchasing options. It’s important to note that these servers can be combined into replication functions as well as clustered servers. Combining servers creates redundancy, flexibility, and often better performance.
Level 1 (1-10 staff): Basic performance and minimal redundancy – $3k – 8k
Level 2 (1-20 staff): Good speed and recommended redundancy – $8k to $18k
Level 3 (20-50 staff): Very good performance and recommended redundancy – $18k to $25k
Level 4 (50-100 staff): Great performance and recommended redundancy – $25-$35k
Level 5 (100+ staff): Best performance and recommended redundancy – $35k-55k+
How often should I replace a server and how does this affect the end user/company experience?
1) Evergreen: 3-5 years to stay in front of old age failures. Hardware performance meets and exceeds software requirements delivering an excellent user experience.
2) Average: “Normal wear and tear” old age failures are common between 5-7 years. Hardware performance level meets software minimum requirements delivering an acceptable user experience.
3) HIGH RISK: 7+ years is borrowed time for hardware! Hardware performance is very lagging behind and hindering software delivering a poor user experience.
NOTE: Always look at performance monitoring with hardware. Replacing hardware will NOT fix poorly written software or databases. Poor performance can be created from both inadequate hardware or bad software (code or settings). If using VMWARE, you can easily see IO requests. Hard drives are generally the slowest part of ANY server. Even poorly performing software will improve with new hardware. However, real performance gains will occur when the software is healthy and the network connecting systems together is optimal. Always remember… you have several factors that dictate performance in an environment! This can be extremely challenging to narrow down, take your time, get a second (or third) opinion, and use the tools at your disposal to develop a well thought out plan.
Why should you consider adding server replication?
The main reason to add replication is to reduce downtime risk levels from hardware failure. Failover scenarios can be “cold/manual”, “warm/ready”, or “hot/ automatic/active”. Failover s using warm failover technology (manual start). To enable true replication, you will require two servers which generally carries double the cost. In some scenarios, you can use the existing (old server) server when appropriate, if compatible, and IF client expectations are extremely clear (generally this is a BAD IDEA… see “NOT RECOMMENDED” towards the end of this article).
If the servers have different hardware, live migrations are NOT possible. The benefit of a live migration is having the ability to move a virtual machine from one host server to another with ZERO downtime. This allows a business to perform hardware repairs and other maintenance without interrupting production. If the server hardware is different, production must stop while the virtual server is moved to another host. This can cause downtime of up to 30 minutes due to the operating system needing to adjust to new hardware which also increases the risk of boot issues. We highly recommend a thorough test before reusing existing different equipment for a replication fail-over host to ensure the least amount of problems.
=== Add Server High Availability and Load Balancing (Cluster) ===
Why?Businesses add this feature to ensure maximum up-time protection from hardware failures as your critical software infrastructure is shared across different hardware. This function requires 2 or more servers (minimum is two with limited functionality, 3 is the REAL minimum for proper business continuity). Clients can start with 2 and then grow as they go! Make sure that you have the correct licensing!
Cost?2-3+ Servers in a cluster with shared storage at a cost of 2-3X hardware costs. SAN hardware (aka shared storage) can also be used in a headless server scenario (servers do not use internal hard drives, only the SAN drives are used) = $20k-$30k per SAN depending on drive types
Go one step further?Add VMware ESXI HCI software (aka Hyper Converged Infrastructure VSAN) which shares local storage on all servers as a single virtual storage = $3k-$8k
Special Note: Even with high availability, it is still recommended to have a separate replication server as a warm fail-over option.
Zero downtime and data loss expectation: This can be complicated on several levels, even in a cluster who may need reboot time (1-2 min). Zero RPO and RTO is most easily accomplished at the software level meaning the software needs to support real-time replication to another copy of itself. Most software doesn’t support this. Ask the LOB vendor if this is possible or not. Set expectations with client based on this conversation.
=== VMWare ESXI hypervisor ===
Why? Get the best performance, management, hardware monitoring, flexibility, and reliability in your hypervisor. Most clients will fit into the “essentials” which is very economical and feature robust. VMware has a noticeable performance improvement over HyperV due to the OS (linux based) and the increased IO (resources) that can be controlled strictly.
Costs? Software license costs can range from $2k to $30k depending on licenses (does not include VSAN licenses and applies on a per server level) and costs are yearly which include software updates and hypervisor support (troubleshooting only). Licenses can be purchased up front and provide savings for the client (up to 5 years).
=== Hyper-V hypervisor ===
Why?Cost is lower than VMWARE but at the cost of management, control, performance, and flexibility. The other downside is that Microsoft provides ZERO “real” support for their software. Microsoft can be a rabbit hole, try to avoid it from a hypervisor standpoint. This can lead to downtime risk.
Costs? Software license costs can range from $1k-$2K per host server. Datacenter versions allow for unlimited VM’s but can carry a cost of $5k to $30k. It’s critical to pay attention to CPU and Core allocations per license.
=== Server OS Work ===
PRO TIP: Do not add to much to a workscope. Adding a server OS software replacement to a new server hardware workscope is a bad idea. Too many moving parts and variables. If anything, you should recommend multiple stages in this process to ensure a control change in the environment. Each stage should have a CMR created to ensure proper execution and communication.
NO TRANSFERS OR UPGRADE OPTIONS (FRESH REBUILD): This can be very time consuming and detailed depending on the prior setup. The details of this process are beyond the scope of this KB. However, it’s not uncommon for our labor to range from 10-30 hours to setup a new server depending on software, features, gpo, and client environmental specifics with testing/research. REMEMBER!!! The servers need a migration and the workstations need updating as well. I’d also recommend adding 30 minutes per workstation as well. This is complicated and requires many steps. Careful here and work with others to ensure nothing is missed.
SERVER OS UPGRADE: This can be a smooth process (sometimes). Always perform a test upgrade prior to performing in a production environment. Always have a CMR created. Allow 3-6 hours for the total process per server depending on server roles (50% testing, 50% production move). Workstations generally do not need any adjustments. Testing a workstation or two is highly recommended to ensure LOB (line of businses) applications and normal user functionality are operating as expected (redirected folders, network shares, printers, group policy, dns, etc…). Add 30 minutes per workstation for testing. Two to three workstations is enough. ALWAYS have the client confirm after you’ve done your check.
=== NOT RECOMMENDED ===
Reusing Old Equipment – We do not encourage the reuse of old equipment whenever possible. Worst case, it’s possible to use the old system as a backup if they have an OS license, rack space, and power for it (you can’t use the same server license key on two physical servers, each need their own license key to not be illegal). Keep in mind that old systems may have the wrong OS, compatibility issues, and may cause more heartaches than they are worth.
Note: Remember, if they have support agreement with us, they need to also pay for this additional server as it’s STILL a server that we need to monitor/manage. The only difference is that it’s high risk vs the newer equipment due to failure rates. Not always worth it for anyone involved. Always CYA and let the client know it’s not a good idea and that proper business continuity planning equipment should be used.
=== Software Licensing Considerations ===
COMPLICATED – This is not a joke, software licensing is complicated. Seek a team member who has a lot of experience to ensure we do not misquote or set the incorrect expectations with the client. It is not a good situation for anyone when we have to tell a client they need to spend another $10K on licensing after they’ve received their new server. Always make sure the client knows that we do our best to ensure accurate licensing and that we cannot be held liable if we miss something. However, we will do our best to always provide the correct quote the first time without surprises.
Considerations: As you build out a new server quote, it is important to understand their current software licensing infrastructure with OS licenses (RDS Cals, User Cals, Device Cals, SQL Cals). The reason this is important is because when you go to new server hardware, you may exceed the permitted licensing assigned to the prior system since the CPU has changed (more cores, more physical processors, etc…). Newer OS’s tend to have more strict requirements and higher pricing since they are optimized and generally contain more features.
Penalties: It’s important to remember that Host Hypervisors like Vmware and HyperV has host, core, and physical CPU requirements that must be met. If the correct license is not purchased, either the software may have issues operating correctly, fail to continue operating, or be in violation of licensing agreements with the software vendor which may add large fines plus ongoing audit requirements (Audit Costs + List pricing + 5% penalty + settlement – In other words… CAREFUL!!!). Since software licenses are governed by contracts with vendors and U.S. copyright law, both civil and criminal penalties are a possibility. Fines can be as much as $150,000 for each illegally installed copy of software, which includes legally purchased software that is not properly licensed. As an individual, fines can be up to $250,000 and up to five years in jail. These costs are in addition to legal fees, negative publicity (BSA loves publicity) and time spent away from the company to deal with the lawsuit.
Auditing Documentation: Information they will likely ask for includes:
1) Proofs of purchase
2) Invoices and sales receipts from vendors
3) Manuals and certificates of authenticity
4) Purchase records for any software and licenses listed in the audit report
5) Comparison of your purchased licenses to the installations found — be sure to consider free and paid upgrades
Source #1: License Compliance Verification FAQ | Microsoft Volume Licensing
Source #2: Software License Compliance: Myth vs. Reality | Business | E-Commerce Times (ecommercetimes.com)