Knowledge Base

The minimum hardware requirements are:

• x86_64 compatible processor
• 4 GB memory per core
• 18 GB disk space for software installation; 400-500 GB if databases (PDB, BLAST, etc) are also installed
• Network card with a configured network interface
• 16-bit color (for Maestro)

It is strongly recommended that you also have the following hardware:

• 60 GB minimum scratch disk space for running jobs
• Graphics card that supports hardware-accelerated OpenGL with 1GB onboard memory and an up-to-date vendor-supplied graphics driver.
• Monitor with a refresh rate of 60 Hz or more and 1280x1024 resolution.

For hardware stereo viewing, one of the following is required (see Article 1745 for details):

• A graphics card that supports quad-buffered OpenGL stereo, with appropriate glasses, and a monitor with a refresh rate of 120 Hz or more.
• A monitor that supports interlaced stereo.

For using Desmond, you must have a supported Nvidia graphics card on a Linux machine. See the Supported Platforms page for more information.

See the MD Compatible Systems page for information about purchasing preconfigured and preinstalled computers for Desmond GPGPU computing.

The minimum is not necessarily the optimum hardware configuration, particularly if you are planning to run large modeling jobs. The list below gives some guidelines and recommendations for the choice of hardware over the requirements listed above.

• Multiple CPUs. Most jobs can be distributed over multiple processors, so the use of a multi-core CPU or a cluster of CPUs, equipped with a queueing system, is recommended to increase throughput.
• Larger CPU cache is recommended, especially for jobs that process large amounts of data, such as Jaguar and QSite. Increasing the L2 or L3 cache size for these jobs is more important than a small increase in CPU speed.
• Faster local disk access is important for jobs that read a lot of data. For example, using SSD, a disk with a higher speed (e.g. 10000 rpm), or a disk array that uses multiple controllers and striping can be beneficial. Distributed Phase database searches are particularly limited by contention for disk access. Jaguar and QSite jobs also perform a lot of IO, so faster disk access is important. Local disks are preferred over networked disks for temporary storage (or for data that is used often) because networked disks are affected by network access, bandwidth, and network traffic.
• Adequate cooling and power supply. Laptops or notebooks are not usually a good choice for large jobs because the cooling and power supply can be inadequate. If you run jobs for a long time, you should consider a power supply with backup, in case of outages.

For large jobs, computing on a cluster with a queueing system is recommended, with the following hardware components:

• A highly capable file server for the external network.
• Shared storage for the intra-cluster network, to reduce traffic to and from the external network.
• Fast processors, large memory, and high-quality motherboards and network interfaces, especially on the management nodes.

Note: Using a networked file share mounted via CIFS (Samba) is not recommended, as Maestro projects use SQLite databases that have locking dependencies not typically available on them.

If you want to use a laptop, you should ensure that you buy a commercial-grade laptop that meets the requirements above. Consumer-grade laptops may not have enough memory or dedicated memory, or adequate graphics cards.

Once you have the hardware, you should install one of the supported operating systems, which are listed on the Supported Platforms page. You might also have to install additional libraries, which are listed in the Installation Guide.