Cooling

All vacuum tubes produce heat while operating. Compared to semiconductor devices, larger tubes operate at higher power levels and hence dissipate more heat. The majority of the heat is dissipated at the anode, though some of the grids can also dissipate power. The tube's heater also contributes to the total, and is a source that semiconductors are free from. Caution should be used in handling heated tubes, as the temperature of the glass may be high enough to easily and quickly burn the skin, even with low-power miniature tubes.
In order to remove generated heat, various methods of cooling may be used. For low power dissipation devices, the heat is radiated from the anode—it often being blackened on the external surface to assist infrared radiation. Natural air circulation or convection is usually required to keep power tubes from overheating. For larger power dissipation, forced-air cooling (fans) may be required.
From the inception of this technology until the 1950s, the dominant approach to cooling low power tubes remained aimed at avoiding immediate or very short term failures. For noncritical consumer applications, and in absence of technological alternatives, tube failures did not create major problems for equipment manufacturers, as the cost of tube replacements was borne by end users long accustomed to the experience. Some tubes for the US defense market featured a metal casing, as opposed to glass, and an opaque, black finish that facilitated both heat conduction and radiative cooling. In some highly specialized professional applications where replacement was out of the question, such as undersea cable repeaters, no failures were acceptable. Moreover, as vacuum tube based defence systems became increasingly complex and deployed in ever increasing numbers, it became clear that point failures which were individually easy to diagnose and rectify had a devastating effect on the uptime of systems that contained tens, hundreds, and especially thousands of tubes. This resulted in both the creation of special long lasting tubes for projects such as Whirlwind and SAGE, and also in special tube shields that aided heat dispersal and could be retrofitted on existing equipment. These shields act by improving heat conduction from the surface of the tube to the shield itself by means of tens of copper tongues in contact with the glass tube, and have an opaque, black outside finish for improved heat radiation.
High-power tubes in older, large transmitters or power amplifiers are liquid cooled, usually with deionised water for heat transfer to an external radiator, similar to the cooling system of an internal combustion engine. Since the anode is usually the cooled element, the anode voltage appears directly on the cooling water surface, thus requiring the water to be an electrical insulator. Otherwise the high voltage can be conducted through the cooling water to the radiator system; hence the need for deionised water. Such systems usually have a built-in water-conductance monitor which will shut down the high tension supply (often tens of kilovolts) if the conductance becomes too high. Some very high-power transmitters, such as those used in shortwave broadcasting and VLF communications, use pressurized steam for cooling. Modern transmitters using tubes mainly in the PA section are now largely cooled by forced air through a radiator or other heat-sinking device.

Information from Wikipedia.