Affordable, cutting edge
Bare Metal Servers
for your most intensive workloads.

CONTACT US ORDER A SERVER

Frequently Asked Questions

General

How long have you been in business?

We've been in business since November 2011. We started with a single colocation server and we progressively scaled to hundreds of high-end servers in 3 different datacenter facilities.

Do you own the servers, IP addresses and datacenters?

We own all servers offered. We do not lease, rent or resell anyone else's servers (unless we are specially requested to do so by our customer). We own all access and distribution layer network equipment connecting our servers. We own all IP addresses assigned to our customers. We are a member of both ARIN and RIPE. We currently own and manage over 20,000 IP addresses. We are present in 3 datacenter facilities : NOCIX, UnReal Servers and Equinix (Amsterdam, via Psychz Networks). We do not own these datacenter facilities. We rent colocation space and we leverage the remote hands services offered in these datacenter facilities to successfully operate.

What operating systems can you install?

We can install any custom or off-the-shelf operating system provided that it is freely available for download or provided an ISO from you. This includes all Linux variants, Windows Server (trial), VMware ESXi, etc. You are also welcome to install your own operating system yourself using the IPMI remote console of your server.

Do you offer software licenses?

We do not offer any software (operating system, control panel, etc.) license. Such licenses must be purchased from your own channels.

Do you offer server management services?

We do not offer server management services; our servers are sold as unmanaged services. The responsibilities of Heymman Servers are limited to ensuring that your server is powered up, has healthy hardware and is connected to a functioning, healthy network. For issues related to the software being run on your server, we typically have to ask you to obtain assistance from your own channels. Our services are targetted at IT professionals or organizations with access to such resources. We can reinstall your operating system on request, if needed.

What are your policies regarding abuse?

All abuse reports (copyright infringement, phishing attacks, spam, etc.) are manually screened, then forwarded to the corresponding customer. Customers are generally allowed 24 hours to resolve the abuse before a null route is instated. In case the abuse isn't resolved and/or the customer isn't responsive or the severity of the abuse demands immediate termination, then the abusive IP address will be null routed. Customers who do not adequately manage their network or who are believed to be directly engaged in abusive activities will be terminated. For more details about our policies, please refer to our Acceptable Use Policy and our Terms of Service.

What are your policies regarding incoming network attacks?

Our Kansas City and Chicago datacenter facilities aren't designed to filter network attacks. In case of an incoming network attack, we are generally able to only null route the targeted IP address. The null route will be maintained until the attack has subsided. If this is a recurrent issue, then we may have to ask you to seek a solution from a third party provider, such a service offering tunneling of clean traffic. Services provided from our Amsterdam location are delivered with DDoS protection up to your allowed limit of 40 Gbps or 6 Mpps (shared). Past this limit, the same terms as defined for our other locations apply.

Does your network supports IPv6?

We fully support IPv6 in all our locations. You are welcome to submit a demand for a free IPv6 assignment.

What's the cost for additional IP addresses?

Additional IP addresses can be ordered in CIDR subnets for $0.50 per IP address per month, subject to availability.

Can you announce a subnet or provide a BGP session?

Demands for announcing subnets are evaluated on a case-by-case basis. We do not provide BGP sessions, so all subnets must originate from one of the ASN's we operate from.

Specific to the HP DL580 G7 systems

Tell us more about the quad E7-4870 offer?

Unlike E3 and E5 v1 processors, which are based on the Sandy Bridge architecture, the E7 v1 family of processors is based on the Westmere architecture. Based on PassMark's benchmark, a combination of 4 of these processors offer compute performance that is similar to dual flagship E5-2698 v4 processors.

[Dual CPU] Intel Xeon E7- 4870 @ 2.40GHz 17541 (<= multiply by 2)
[Dual CPU] Intel Xeon E5-2698 v4 @ 2.20GHz 33192

Here's another great comparative benchmark :

http://www.anandtech.com/show/7757/quad-ivy-brigde-ex-60-cores-120-threads/11

The Crucial MX300 SSD's, which we provision these servers with, are capable of up to 530 MB/s sequential reads, 510 MB/s sequential writes, 92K random 4K reads and 83K random 4K writes, per drive. To meet the expected level of performance from a RAID array of SSD's, we provision the servers by default with an LSI 9361-8i hardware RAID controller. This controller supports 12Gbps drives and 500+K IOPS throughput, allowing the SSD's to reach their full potential. This controller, based on the LSI 3108, is the latest and greatest controller from LSI and is the same controller integrated on all modern Supermicro motherboards with hardware RAID support.

Is there any bandwidth or port limitation?

None whatsoever. We do not block any port, we do not filter traffic and we do not throttle ports. These servers are offered with an unmetered 1Gbps or 10Gbps network port. You can expect to be able to use your port at its full capacity 24x7x365. Our 10Gbps network ports are equivalent to FDC Server's unmetered, dedicated 10Gbps ports. Since the ports are single-homed to Cogent, expect 7-8 Gbps available 24/7.

Tell us more about the disks performance?

We tested the LSI 9361-8i hardware RAID controller with CentOS 6.8 and the ext4 filesystem. For the following command :

# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync

We obtain the following results :

RAID-10 (4x 2TB Crucial MX300) :
665 to 680 MB/s with Write Back and 540 to 560 MB/s with Write Through.

RAID-10 (4x 2TB Samsung 850 EVO) :
650 to 660 MB/s with Write Back and 530 to 545 MB/s with Write Through.

RAID-5 (4x 2TB Crucial MX300) :
580 to 680 MB/s with Write Back and 480 to 760 MB/s with Write Through.

RAID-5 (4x 2TB Samsung 850 EVO) :
580 to 680 MB/s with Write Back and 645 to 655 MB/s with Write Through.

We ran disk benchmarks on Windows Server 2012 R2, using CrystalDiskMark and ATTO Disk Benchmark. The CrystalDiskMark results are as follows :

CrystlaDiskMark benchmarks for the HP DL580 G7

The ATTO Disk Benchmark results are as follows :

ATTO Disk Benchmark benchmarks for the HP DL580 G7

ATTO Disk Benchmark results for a 1 drive configuration follow below :

ATTO Disk Benchmark benchmarks for the HP DL580 G7

These results go to show the importance of the cache memory of the controller, for sequential writes, and a significant performance gain with the Samsung 850 EVO, when the controller's cache runs out. We equip our LSI 9361-8i controllers with the matching CacheVault module, allowing Write Back to be safely enabled.

Do you offer any upgrade for the quad E7-4870 servers, beyond what's shown on the order page?

Deploying standardized servers allows us to reduce our overhead costs. In some cases, we can offer additional customization options. Please contact us regarding your requirements. Also, we do not downgrade servers after they were ordered.

How quickly can you deploy these quad E7-4870 servers?

The setup time varies dependently of several factors. The time required to deploy one of these servers can vary between few hours to approximately 8 business days. Please contact us for a more exact ETA.

The server takes a long time to boot (5-7 minutes), why?

Most modern dual socket motherboards support up to 16 memory modules, while most modern quad socket motherboards support up to 48 memory modules. Typically, motherboards are made of a single PCB, into which memory modules are directly inserted. To elevate the maximum amount of memory supported, some very high end systems employ memory expansion modules (or cartridges), which are inserted into the motherboard's main PCB. The HP DL580 G7, but also the G8, G9 and later versions as well as Dell R930 and multiple Supermicro systems, for example, employ memory expansion modules to increase memory density. Due to the complexity of these systems, an additional 3-4 minutes is required at POST time to detect all memory modules. The presence of an hardware RAID controller also adds approximately a minute to the boot process.

What operating systems are the quad E7-4870 servers compatible with?

The quad E7-4870 servers are compatible with all commonly used operating systems, including Windows Server 2012 R2 and Windows Server 2016 (which might require the installation of additional drivers for the NIC). We are happy to assist with the installation of your operating system.

Can we disable CPU cores to reduce our licensing expenditures?

The quad E7-4870 servers support disabling an even number of CPU cores, meaning that as little as 2 cores per CPU may be enabled (8 cores in total). This can translate into dramatic savings, for workloads requiring a huge amount of memory, but little processing power.

Specific to the HP SL230s systems

Tell us more about the disks performance and the HP B320i RAID controller?

The SP SL230s comes standard with an HP H220i HBA (host bus adapter), which is a rebranded LSI 2308. These systems are also equipped with an HP B320i RAID controller (with no cache memory).

Note : Before enabling the HP B320i RAID controller, the first and last sectors of each drive must be erased (there must be no metadata present on the drives). Before exiting the HP ACU, the newly created virtual drive must be selected as the primary and secondary boot drive in the "Set Bootable Logical Drive/Volume". Failure to follow these instructions will result in the system failing to POST until the drives are physically detached.

To wipe out the first and last sectors of the drives, a proposed method is to boot the server into a rescue environment (e.g. from a CentOS installation disc) and to run :

# dd bs=512 count=10000 if=/dev/zero of=/dev/sda seek=0
# dd bs=512 count=10000 if=/dev/zero of=/dev/sda seek={last_sector - 10000}

You can find the number of sectors using :

# blockdev --getsz /dev/sda

Subsract 10,000 to this number and write 10,000 sectors to overwrite the last 10,000 sectors. Repeat for each drive, replacing /dev/sda accordingly.

The following screenshot shows how to set the bootable drive :

Setting the bootable drive for the HP SL230s

We tested the HP H220i HBA and HP B320i RAID controller with CentOS 7.6 and the XFS filesystem. For the following command :

# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync

We obtained the following results :

- No RAID on HP H220i (1x 500GB Samsung 860 EVO) : 200 to 220 MB/s.
- Software (mdadm) RAID-10 on HP H220i (4x 500GB Samsung 860 EVO) : 350 to 380 MB/s.
- No RAID on HP B320i (1x 500GB Samsung 860 EVO) : 400 to 450 MB/s.
- Hardware RAID-10 on HP B320i (4x 500GB Samsung 860 EVO) : 550 to 590 MB/s (~730 MB/s on a VM with VirtIO enabled).

We ran disk benchmarks on Windows Server 2016, using CrystalDiskMark and ATTO Disk Benchmark. The CrystalDiskMark results are as follows :

CrystlaDiskMark benchmarks for the HP SL230s

The ATTO Disk Benchmark results are as follows :

ATTO Disk Benchmark benchmarks for the HP SL230s

These results show a tremendous improvement when enabling the HP B320i RAID controller. In fact, in the single drive scenario, the drive can roughly reach the full manufacturer's specification of 550 MB/s sequential reads and 520 MB/s sequential writes. In the RAID-10 scenario, we observe that the HP B320i performs roughly 15% less (both in reads and writes) than a modern, top-of-the-line RAID controller in Write Through mode (e.g. an LSI 9361, see the same article in the "Specific to the quad E7-4870 offer" section of this FAQ for additional benchmarks). We therefor conclude that the HP B320i controller does not constitute a substantial bottleneck to modern SSD's, in comparison to modern, top-of-the-line RAID controllers in Write Through mode. On the other hand, the HP H220i HBA significantly bottlenecks the write performance of the SSD's. We therefor strongly recommend to enable to HP B320i RAID controller.

The HP B320i RAID controller can be enabled from the "HP Dynamic Smart Array B320i RAID Controller" menu in the system's BIOS. The RAID array can be created from the HP Smart Storage Administrator (ACU) utility. Our tests show that it is crucial to enable the "Physical Drive Write Cache State" setting in the "Modify Controller Settings" menu. We recommend to disable the controller cache setting. We did not measure any significant difference between a 128 KiB / 265 KiB and a 256 KiB / 512 KiB Stripe Size.

Creating a logical drive for the HP SL230s

Changing drive write cache setting for the HP SL230s

All conventional operating systems have built-in drivers for the HP H220i HBA. HP provides closed source drivers for the HP B320i RAID controller for a number of operating systems, including and limited to Windows Server 2008, 2012, 2012 R2, 2016 and (unofficially) 2019 as well as VMware ESXi 6.5 and older and Red Hat / CentOS 5, 6 and 7. If installing a supplemental, custom kernel (e.g. Proxmox or CloudLinux), compatibility will be lost. For Windows operating systems, the drivers have to be slipstreamed in the installation media. For VMware ESXi, a custom ISO, freely available on VMware's website, has to be used. For Red Hat based operating systems, a driver disc has to be created and the drivers have to be loaded using, for instance, the inst.dd boot option. Here are articles to assist with these tasks :

- Tool to slipstream drivers to Windows 10 : https://www.ntlite.com/
- Creating a CentOS 7 driver disk : https://www.marcanoonline.com/post/2016/07/red-hat-enterprise-linux-and-centos-7-driver-disk/
- Loading drivers to the CentOS 7 installer : https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/7/html/installation_guide/sect-driver-updates-performing-ppc

Heymman Servers will install all Windows, Red Hat based and VMware ESXi operating systems with the HP B320i RAID controller enabled to ensure that our customers benefit from the highest level of performance.

Video tutorials

How to use the IPMI management interface of my HP G7 server (iLO 3)?

To access the iLO3 IPMI management interface of your HP server, please direct your web browser to the URL ("IPMI address") that was given to you and use the username and password that were given to you when your server has been provisioned. From this interface, in the "Remote Console" tab, in the left panel, and the "Remote Console" menu element, you may launch the IPMI remote console by clicking on "Launch" under "Java Integrated Remote Console", which will provide you with KVM access to the server. Please make sure that your Java security settings allow the Java application to run. To be able to run unsigned Java applets, the IPMI IP address has to be added to your Java "Exception Site List", in the Java Control Panel. From the Java applet, you may also emulate an ISO of an operating system, in the "Virtual Drives" tab and "Image File CD/DVD-ROM" menu element, should you want to install your own operating system. Should it ever be needed, you may also power cycle your server from the "Power Management" menu, under "Server Power".

If you need further assistance, please consult the following video tutorial or contact us.