What hardware to use for reliable and scalable services
Each Enswitch customer has different goals for their system, either in terms of features or sizing. Therefore, the specifications below are only recommendations.
These specifications will provide a good scalable system without excessive cost. Small or lightly-used systems can have lower specifications, while they may be inadequate for very large or non-standard ones. If in doubt, please discuss your plans with us.
Test and demo machines: Any modern server PC with 4GB or more RAM.
Single-machine systems: 2 processor cores, 4GB RAM, 2 disks of 40GB or more in RAID1.
Database machines for systems below 1000 concurrent calls: 4 or more processor cores, 16GB or more RAM, 4 disks of 120GB or more in 2 RAID1 pairs (one pair is for the operating system, and the other for the database). Database machines are normally RAM and disk bound, so use plenty of RAM and the fastest affordable disks
Database machines for systems of 1000 concurrent calls and above: 8 or more processor cores, 32GB or more RAM, 6 disks of 120GB or more in 3 RAID1 pairs (one pair is for the operating system, the other one is for database storage, and the third one is for logs). Database machines are normally RAM and disk bound, so use plenty of RAM and the fastest affordable disks.
NFS machines: 4 or more processor cores, 16GB or more RAM, 4 disks of 120GB or more in 2 RAID1 pairs (one pair is for the operating system, and the other for the file storage). NFS machines are normally disk bound, so the fastest affordable disks should be used. Having more RAM also helps.
Kamailio/OpenSIPS machines: 4 or more processor cores, 8GB RAM, 2 disks of 40GB or more in RAID1. Disk performance isn’t critical.
Asterisk machines: 4 or more processor cores, 4GB RAM, 2 disks of 40GB or more in RAID1. Asterisk machines are normally CPU bound, so if budget allows add more and faster processor cores. Disk performance isn’t critical.
The database machines will be the long-term limit on how many users the system can support, so, if you are on a budget, spend it on the database servers.
The easiest ways to increase the scalability of large clusters are to tune MySQL or to add more RAM to the database servers.
1 dual core processor is an acceptable substitute for dual processors; likewise, 2 dual-core processors are an acceptable substitute for quad-processors, and so on.
Hardware RAID controllers are preferred for the RAID1 pairs for performance.
For very large or busy systems, RAID10 can be used to increase disk performance on the database and NFS machines. Due to its complexity, please discuss it with your vendor.
If you are using NFS, please disable Spanning Tree Protocol (STP) on the Ethernet switch. Not doing so may lead to the NFS mounts failing to start automatically after rebooting NFS clients. STP is enabled by default on Cisco switches.
Do you need further help or information? Contact us